init

README.md

@@ -1 +1,1402 @@
-We are still uploading and the upload is completed today
+---
+pipeline_tag: any-to-any
+datasets:
+- openbmb/RLAIF-V-Dataset
+library_name: transformers
+language:
+- multilingual
+tags:
+- minicpm-o
+- omni
+- vision
+- ocr
+- multi-image
+- video
+- custom_code
+- audio
+- speech
+- voice cloning
+- live Streaming
+- realtime speech conversation
+- asr
+- tts
+---
+
+<h1>A GPT-4o Level MLLM for Vision, Speech and Multimodal Live Streaming on Your Phone</h1>
+
+[GitHub](https://github.com/OpenBMB/MiniCPM-o) | [Online Demo](https://minicpm-omni-webdemo-us.modelbest.cn)</a>
+
+
+## MiniCPM-o 2.6
+
+**MiniCPM-o 2.6** is the latest and most capable model in the MiniCPM-o series. The model is built in an end-to-end fashion based on SigLip-400M, Whisper-medium-300M, ChatTTS-200M, and Qwen2.5-7B with a total of 8B parameters. It exhibits a significant performance improvement over MiniCPM-V 2.6, and introduces new features for real-time speech conversation and multimodal live streaming. Notable features of MiniCPM-o 2.6 include:
+
+- 🔥 **Leading Visual Capability.**
+  MiniCPM-o 2.6 achieves an average score of 70.2 on OpenCompass, a comprehensive evaluation over 8 popular benchmarks. **With only 8B parameters, it surpasses widely used proprietary models like GPT-4o-202405, Gemini 1.5 Pro, and Claude 3.5 Sonnet** for single image understanding. It also **outperforms GPT-4V and Claude 3.5 Sonnet** in mutli-image and video understanding, and shows promising in-context learning capability.
+
+- 🎙 **State-of-the-art Speech Capability.** MiniCPM-o 2.6 supports **bilingual real-time speech conversation with configurable voices** in English and Chinese. It **outperforms GPT-4o-realtime on audio understanding tasks** such as ASR and STT translation, and shows **state-of-the-art performance on speech conversation in both semantic and acoustic evaluations in the open-source community**. It also allows for fun features such as emotion/speed/style control, end-to-end voice cloning, role play, etc.
+
+- 🎬 **Strong Multimodal Live Streaming Capability.** As a new feature, MiniCPM-o 2.6 can **accept continous video and audio streams independent of user queries, and support real-time speech interaction**. It **outperforms GPT-4o-202408 and Claude 3.5 Sonnet and shows state-of-art performance in open-source community on StreamingBench**, a comprehensive benchmark for real-time video understanding, omni-source (video & audio) understanding, and multimodal contextual understanding.										
+
+- 💪 **Strong OCR Capability and Others.**
+Advancing popular visual capabilites from MiniCPM-V series, MiniCPM-o 2.6 can process images with any aspect ratio and up to 1.8 million pixels (e.g., 1344x1344). It achieves **state-of-the-art performance on OCRBench for models under 25B, surpassing proprietary models such as GPT-4o-202405**.
+  Based on the the latest [RLAIF-V](https://github.com/RLHF-V/RLAIF-V/) and [VisCPM](https://github.com/OpenBMB/VisCPM) techniques, it features **trustworthy behaviors**, outperforming GPT-4o and Claude 3.5 Sonnet on MMHal-Bench, and supports **multilingual capabilities** on more than 30 languages.
+
+
+- 🚀 **Superior Efficiency.**
+  In addition to its friendly size, MiniCPM-o 2.6 also shows **state-of-the-art token density** (i.e., number of pixels encoded into each visual token). **It produces only 640 tokens when processing a 1.8M pixel image, which is 75% fewer than most models**. This directly improves the inference speed, first-token latency, memory usage, and power consumption. As a result, MiniCPM-o 2.6 can efficiently support **multimodal live streaming** on end-side devices such as iPad.
+
+-  💫  **Easy Usage.**
+MiniCPM-o 2.6 can be easily used in various ways: (1) [llama.cpp](https://github.com/OpenBMB/llama.cpp/blob/minicpm-omni/examples/llava/README-minicpmo2.6.md) support for efficient CPU inference on local devices, (2) [int4](https://huggingface.co/openbmb/MiniCPM-o-2_6-int4) and [GGUF](https://huggingface.co/openbmb/MiniCPM-o-2_6-gguf) format quantized models in 16 sizes, (3) [vLLM](#efficient-inference-with-llamacpp-ollama-vllm) support for high-throughput and memory-efficient inference, (4) fine-tuning on new domains and tasks with [LLaMA-Factory](./docs/llamafactory_train.md), (5) quick local WebUI demo setup with [Gradio](#chat-with-our-demo-on-gradio), and (6) online web demo on [server](https://minicpm-omni-webdemo-us.modelbest.cn/).
+
+
+**Model Architecture.**
+
+- **End-to-end Omni-modal Architecture.** Different modality encoder/decoders are connected and trained in an **end-to-end** fashion to fully exploit rich multimodal knowledge.
+- **Omni-modal Live Streaming Mechanism.** (1) We change the offline modality encoder/decoders into online ones for **streaminig inputs/outputs.** (2) We devise a **time-division multiplexing (TDM) mechanism** for omni-modality streaminig processing in the LLM backbone. It divides parallel omni-modality streams into sequential info within small periodic time slices. 
+- **Configurable Speech Modeling Design.** We devise a multimodal system prompt, including traditional text system prompt, and **a new audio system prompt to determine the assistant voice**. This enables flexible voice configurations in inference time, and also facilitates end-to-end voice cloning and description-based voice creation.
+
+<div align="center">
+<img src="https://github.com/OpenBMB/MiniCPM-o/raw/main/assets/minicpm-o-26-framework.png" , width=80%>
+</div>
+
+
+### Evaluation  <!-- omit in toc -->
+
+<div align="center">
+    <img src="https://github.com/OpenBMB/MiniCPM-o/raw/main/assets/radar.jpg" width=90% />
+</div>
+
+<details>
+<summary>Click to view visual understanding results.</summary>
+
+**Image Understanding**
+
+<div align="center">
+<table style="margin: 0px auto;">
+    <thead>
+        <tr>
+            <th align="left">Model</th>
+            <th>Size</th>
+            <th>Token Density<sup>+</sup></th>
+            <th>OpenCompass</th>
+            <th>OCRBench</th>
+            <th>MathVista mini</th>
+            <th>ChartQA</th>
+            <th>MMVet</th>
+            <th>MMStar</th>
+            <th>MME</th>
+            <th>MMB1.1 test</th>
+            <th>AI2D</th>
+            <th>MMMU val</th>
+            <th>HallusionBench</th>
+            <th>TextVQA val</th>
+            <th>DocVQA test</th>
+            <th>MathVerse mini</th>
+            <th>MathVision</th>
+            <th>MMHal Score</th>
+        </tr>
+    </thead>
+    <tbody align="center">
+        <tr>
+            <td colspan="19" align="left"><strong>Proprietary</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">GPT-4o-20240513</td>
+            <td>-</td>
+            <td>1088</td>
+            <td><u>69.9</u></td>
+            <td>736</td>
+            <td>61.3</td>
+            <td>85.7</td>
+            <td><strong>69.1</strong></td>
+            <td>63.9</td>
+            <td>2328.7</td>
+            <td>82.2</td>
+            <td>84.6</td>
+            <td><strong>69.2</strong></td>
+            <td><strong>55.0</strong></td>
+            <td>-</td>
+            <td>92.8</td>
+            <td><strong>50.2</strong></td>
+            <td><strong>30.4</strong></td>
+            <td><u>3.6</u></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Claude3.5-Sonnet</td>
+            <td>-</td>
+            <td>750</td>
+            <td>67.9</td>
+            <td>788</td>
+            <td>61.6</td>
+            <td><strong>90.8</strong></td>
+            <td>66.0</td>
+            <td>62.2</td>
+            <td>1920.0</td>
+            <td>78.5</td>
+            <td>80.2</td>
+            <td><u>65.9</u></td>
+            <td>49.9</td>
+            <td>-</td>
+            <td><strong>95.2</strong></td>
+            <td>-</td>
+            <td>-</td>
+            <td>3.4</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Gemini 1.5 Pro</td>
+            <td>-</td>
+            <td>-</td>
+            <td>64.4</td>
+            <td>754</td>
+            <td>57.7</td>
+            <td>81.3</td>
+            <td>64.0</td>
+            <td>59.1</td>
+            <td>2110.6</td>
+            <td>73.9</td>
+            <td>79.1</td>
+            <td>60.6</td>
+            <td>45.6</td>
+            <td>73.5</td>
+            <td>86.5</td>
+            <td>-</td>
+            <td>19.2</td>
+            <td>-</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">GPT-4o-mini-20240718</td>
+            <td>-</td>
+            <td>1088</td>
+            <td>64.1</td>
+            <td>785</td>
+            <td>52.4</td>
+            <td>-</td>
+            <td>66.9</td>
+            <td>54.8</td>
+            <td>2003.4</td>
+            <td>76.0</td>
+            <td>77.8</td>
+            <td>60.0</td>
+            <td>46.1</td>
+            <td>-</td>
+            <td>-</td>
+            <td>-</td>
+            <td>-</td>
+            <td>3.3</td>
+        </tr>
+        <tr>
+            <td colspan="19" align="left"><strong>Open Source</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Cambrian-34B</td>
+            <td>34B</td>
+            <td><u>1820</u></td>
+            <td>58.3</td>
+            <td>591</td>
+            <td>50.3</td>
+            <td>75.6</td>
+            <td>53.2</td>
+            <td>54.2</td>
+            <td>2049.9</td>
+            <td>77.8</td>
+            <td>79.5</td>
+            <td>50.4</td>
+            <td>41.6</td>
+            <td>76.7</td>
+            <td>75.5</td>
+            <td>-</td>
+            <td>-</td>
+            <td>-</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">GLM-4V-9B</td>
+            <td>13B</td>
+            <td>784</td>
+            <td>59.1</td>
+            <td>776</td>
+            <td>51.1</td>
+            <td>-</td>
+            <td>58.0</td>
+            <td>54.8</td>
+            <td>2018.8</td>
+            <td>67.9</td>
+            <td>71.2</td>
+            <td>46.9</td>
+            <td>45.0</td>
+            <td>-</td>
+            <td>-</td>
+            <td>-</td>
+            <td>-</td>
+            <td>-</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Pixtral-12B</td>
+            <td>12B</td>
+            <td>256</td>
+            <td>61.0</td>
+            <td>685</td>
+            <td>56.9</td>
+            <td>81.8</td>
+            <td>58.5</td>
+            <td>54.5</td>
+            <td>-</td>
+            <td>72.7</td>
+            <td>79.0</td>
+            <td>51.1</td>
+            <td>47.0</td>
+            <td>75.7</td>
+            <td>90.7</td>
+            <td>-</td>
+            <td>-</td>
+            <td>-</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">DeepSeek-VL2-27B (4B)</td>
+            <td>27B</td>
+            <td>672</td>
+            <td>66.4</td>
+            <td>809</td>
+            <td>63.9</td>
+            <td>86.0</td>
+            <td>60.0</td>
+            <td>61.9</td>
+            <td>2253.0</td>
+            <td>81.2</td>
+            <td>83.8</td>
+            <td>54.0</td>
+            <td>45.3</td>
+            <td><u>84.2</u></td>
+            <td>93.3</td>
+            <td>-</td>
+            <td>-</td>
+            <td>3.0</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Qwen2-VL-7B</td>
+            <td>8B</td>
+            <td>784</td>
+            <td>67.1</td>
+            <td><u>866</u></td>
+            <td>58.2</td>
+            <td>83.0</td>
+            <td>62.0</td>
+            <td>60.7</td>
+            <td>2326.0</td>
+            <td>81.8</td>
+            <td>83.0</td>
+            <td>54.1</td>
+            <td>50.6</td>
+            <td><strong>84.3</strong></td>
+            <td><u>94.5</u></td>
+            <td>31.9</td>
+            <td>16.3</td>
+            <td>3.2</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">LLaVA-OneVision-72B</td>
+            <td>72B</td>
+            <td>182</td>
+            <td>68.1</td>
+            <td>741</td>
+            <td>67.5</td>
+            <td>83.7</td>
+            <td>60.6</td>
+            <td><strong>65.8</strong></td>
+            <td>2261.0</td>
+            <td><strong>85.0</strong></td>
+            <td><u>85.6</u></td>
+            <td>56.8</td>
+            <td>49.0</td>
+            <td>80.5</td>
+            <td>91.3</td>
+            <td>39.1</td>
+            <td>-</td>
+            <td>3.5</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">InternVL2.5-8B</td>
+            <td>8B</td>
+            <td>706</td>
+            <td>68.3</td>
+            <td>822</td>
+            <td><u>64.4</u></td>
+            <td>84.8</td>
+            <td>62.8</td>
+            <td>62.8</td>
+            <td>2344.0</td>
+            <td><u>83.6</u></td>
+            <td>84.5</td>
+            <td>56.0</td>
+            <td>50.1</td>
+            <td>79.1</td>
+            <td>93.0</td>
+            <td>39.5</td>
+            <td>19.7</td>
+            <td>3.4</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">MiniCPM-V 2.6</td>
+            <td>8B</td>
+            <td><strong>2822</strong></td>
+            <td>65.2</td>
+            <td>852*</td>
+            <td>60.6</td>
+            <td>79.4</td>
+            <td>60.0</td>
+            <td>57.5</td>
+            <td><u>2348.4*</u></td>
+            <td>78.0</td>
+            <td>82.1</td>
+            <td>49.8*</td>
+            <td>48.1*</td>
+            <td>80.1</td>
+            <td>90.8</td>
+            <td>25.7</td>
+            <td>18.3</td>
+            <td>3.6</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">MiniCPM-o 2.6</td>
+            <td>8B</td>
+            <td><strong>2822</strong></td>
+            <td><strong>70.2</strong></td>
+            <td><strong>897*</strong></td>
+            <td><strong>71.9*</strong></td>
+            <td><u>86.9*</u></td>
+            <td><u>67.5</u></td>
+            <td><u>64.0</u></td>
+            <td><strong>2372.0*</strong></td>
+            <td>80.5</td>
+            <td><strong>85.8</strong></td>
+            <td>50.4*</td>
+            <td><u>51.9</u></td>
+            <td>82.0</td>
+            <td>93.5</td>
+            <td><u>41.4*</u></td>
+            <td><u>23.1*</u></td>
+            <td><strong>3.8</strong></td>
+        </tr>
+    </tbody>
+</table>
+</div>
+* We evaluate this benchmark using chain-of-thought prompting. Specifically, for MME, we used this technique only for the Cognition set.
+
+
+<sup>+</sup> Token Density: number of pixels encoded into each visual token at maximum resolution, i.e., # pixels at maximum resolution / # visual tokens.
+
+Note: For proprietary models, we calculate token density based on the image encoding charging strategy defined in the official API documentation, which provides an upper-bound estimation.
+
+
+**Multi-image and Video Understanding**
+
+<div align="center">
+ 
+<table style="margin: 0px auto;">
+    <thead>
+        <tr>
+            <th align="left">Model</th>
+            <th>Size</th>
+            <th>BLINK val</th>
+            <th>Mantis Eval</th>
+            <th>MIRB</th>
+            <th>Video-MME (wo / w subs)</th>
+        </tr>
+    </thead>
+    <tbody align="center">
+        <tr>
+            <td colspan="6" align="left"><strong>Proprietary</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">GPT-4o-20240513</td>
+            <td>-</td>
+            <td><strong>68.0</strong></td>
+            <td>-</td>
+            <td>-</td>
+            <td><strong>71.9/77.2<strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">GPT4V</td>
+            <td>-</td>
+            <td>54.6</td>
+            <td>62.7</td>
+            <td>53.1</td>
+            <td>59.9/63.3</td>
+        </tr>
+        <tr>
+            <td colspan="6" align="left"><strong>Open-source</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">LLaVA-NeXT-Interleave 14B</td>
+            <td>14B</td>
+            <td>52.6</td>
+            <td>66.4</td>
+            <td>30.2</td>
+            <td>-</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">LLaVA-OneVision-72B</td>
+            <td>72B</td>
+            <td>55.4</td>
+            <td><strong>77.6</strong></td>
+            <td>-</td>
+            <td><u>66.2/69.5</u></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">MANTIS 8B</td>
+            <td>8B</td>
+            <td>49.1</td>
+            <td>59.5</td>
+            <td>34.8</td>
+            <td>-</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Qwen2-VL-7B</td>
+            <td>8B</td>
+            <td>53.2</td>
+            <td>69.6*</td>
+            <td><strong>67.6*</strong></td>
+            <td>63.3/69.0</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">InternVL2.5-8B</td>
+            <td>8B</td>
+            <td>54.8</td>
+            <td>67.7</td>
+            <td>52.5</td>
+            <td>64.2/66.9</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">MiniCPM-V 2.6</td>
+            <td>8B</td>
+            <td>53.0</td>
+            <td>69.1</td>
+            <td>53.8</td>
+            <td>60.9/63.6</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">MiniCPM-o 2.6</td>
+            <td>8B</td>
+            <td><u>56.7</u></td>
+            <td><u>71.9</u></td>
+            <td><u>58.6</u></td>
+            <td>63.9/67.9</td>
+        </tr>
+    </tbody>
+</table>
+
+</div>
+* We evaluate officially released checkpoints by ourselves.
+
+</details>
+
+
+<details>
+<summary>Click to view audio understanding and speech conversation results.</summary>
+
+**Audio Understanding**
+
+<div align="center">
+<table style="margin: 0px auto;">
+    <thead>
+        <tr>
+            <th align="left">Task</th>
+            <th>Size</th>
+            <th colspan="3">ASR (zh)</th>
+            <th colspan="3">ASR (en)</th>
+            <th colspan="2">AST</th>
+            <th>Emotion</th>
+        </tr>
+        <tr>
+            <th align="left">Metric</th>
+            <td></td>
+            <th colspan="3">CER↓</th>
+            <th colspan="3">WER↓</th>
+            <th colspan="2">BLEU↑</th>
+            <th>ACC↑</th>
+        </tr>
+        <tr>
+            <th align="left">Dataset</th>
+            <td></td>
+            <th>AISHELL-1</th>
+            <th>Fleurs zh</th>
+            <th>WenetSpeech test-net</th>
+            <th>LibriSpeech test-clean</th>
+            <th>GigaSpeech</th>
+            <th>TED-LIUM</th>
+            <th>CoVoST en2zh</th>
+            <th>CoVoST zh2en</th>
+            <th>MELD emotion</th>
+        </tr>
+    </thead>
+    <tbody align="center">
+        <tr>
+            <td colspan="11" align="left"><strong>Proprietary</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">GPT-4o-Realtime</td>
+            <td>-</td>
+            <td>7.3*</td>
+            <td><u>5.4*</u></td>
+            <td>28.9*</td>
+            <td>2.6*</td>
+            <td>12.9*</td>
+            <td>4.8*</td>
+            <td>37.1*</td>
+            <td>15.7*</td>
+            <td>33.2*</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Gemini 1.5 Pro</td>
+            <td>-</td>
+            <td>4.5*</td>
+            <td>5.9*</td>
+            <td>14.3*</td>
+            <td>2.9*</td>
+            <td>10.6*</td>
+            <td><strong>3.0*</strong></td>
+            <td><u>47.3*</u></td>
+            <td>22.6*</td>
+            <td>48.4*</td>
+        </tr>
+        <tr>
+            <td colspan="11" align="left"><strong>Open-Source</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Qwen2-Audio-Base</td>
+            <td>8B</td>
+            <td>-</td>
+            <td>7.5</td>
+            <td>-</td>
+            <td><strong>1.6</strong></td>
+            <td>-</td>
+            <td>-</td>
+            <td>45.2</td>
+            <td><u>24.4</u></td>
+            <td><strong>55.3</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Qwen2-Audio-Instruction</td>
+            <td>8B</td>
+            <td>2.6*</td>
+            <td>6.9*</td>
+            <td><u>10.3*</u></td>
+            <td>3.1*</td>
+            <td><u>9.7</u>*</td>
+            <td>5.9*</td>
+            <td>39.5*</td>
+            <td>22.9*</td>
+            <td>17.4*</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">GLM-4-Voice-Base</td>
+            <td>9B</td>
+            <td><u>2.5</u></td>
+            <td>-</td>
+            <td>-</td>
+            <td>2.8</td>
+            <td>-</td>
+            <td>-</td>
+            <td>-</td>
+            <td>-</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">MiniCPM-o 2.6</td>
+            <td>8B</td>
+            <td><strong>1.6</strong></td>
+            <td><strong>4.4</strong></td>
+            <td><strong>6.9</strong></td>
+            <td><u>1.7</u></td>
+            <td><strong>8.7</strong></td>
+            <td><strong>3.0</strong></td>
+            <td><strong>48.2</strong></td>
+            <td><strong>27.2</strong></td>
+            <td><u>52.4</u></td>
+        </tr>
+    </tbody>
+</table>
+</div>
+* We evaluate officially released checkpoints by ourselves.<br><br>
+
+**Speech Generation**
+
+<div align="center">
+<table style="margin: 0px auto;">
+    <thead>
+        <tr>
+            <th align="left">Task</th>
+            <th>Size</th>
+            <th colspan="9">SpeechQA</th>
+        </tr>
+        <tr>
+            <th align="left">Metric</th>
+            <th></th>
+            <th colspan="3">ACC↑</th>
+            <th>G-Eval (10 point)↑</th>
+            <th>Semantic ELO score↑</th>
+            <th>Acoustic ELO score↑</th>
+            <th>Overall ELO score↑</th>
+            <th>UTMOS↑</th>
+            <th>ASR-WER↓</th>
+        </tr>
+        <tr>
+            <th align="left">Dataset</th>
+            <th></th>
+            <th>Speech Llama Q.</th>
+            <th>Speech Web Q.</th>
+            <th>Speech Trivia QA</th>
+            <th>Speech AlpacaEval</th>
+            <th colspan="5">AudioArena</th>
+        </tr>
+    </thead>
+    <tbody align="center">
+        <tr>
+            <td colspan="11" align="left"><strong>Proprietary</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">GPT-4o-Realtime</td>
+            <td></td>
+            <td><strong>71.7</strong></td>
+            <td><strong>51.6</strong></td>
+            <td><strong>69.7</strong></td>
+            <td><strong>7.4</strong></td>
+            <td><strong>1157</strong></td>
+            <td><strong>1203</strong></td>
+            <td><strong>1200</strong></td>
+            <td><strong>4.2</strong></td>
+            <td><strong>2.3</strong></td>
+        </tr>
+        <tr>
+            <td colspan="11" align="left"><strong>Open-Source</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">GLM-4-Voice</td>
+            <td>9B</td>
+            <td>50.0</td>
+            <td>32.0</td>
+            <td>36.4</td>
+            <td><u>5.1</u></td>
+            <td>999</td>
+            <td>1147</td>
+            <td>1035</td>
+            <td><u>4.1</u></td>
+            <td><u>11.7</u></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Llama-Omni</td>
+            <td>8B</td>
+            <td>45.3</td>
+            <td>22.9</td>
+            <td>10.7</td>
+            <td>3.9</td>
+            <td>960</td>
+            <td>878</td>
+            <td>897</td>
+            <td>3.2</td>
+            <td>24.3</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Moshi</td>
+            <td>7B</td>
+            <td>43.7</td>
+            <td>23.8</td>
+            <td>16.7</td>
+            <td>2.4</td>
+            <td>871</td>
+            <td>808</td>
+            <td>875</td>
+            <td>2.8</td>
+            <td>8.2</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Mini-Omni</td>
+            <td>1B</td>
+            <td>22.0</td>
+            <td>12.8</td>
+            <td>6.9</td>
+            <td>2.5</td>
+            <td>926</td>
+            <td>803</td>
+            <td>865</td>
+            <td>3.4</td>
+            <td>10.0</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">MiniCPM-o 2.6</td>
+            <td>8B</td>
+            <td><u>61.0</u></td>
+            <td><u>40.0</u></td>
+            <td><u>40.2</u></td>
+            <td><u>5.1</u></td>
+            <td><u>1088</u></td>
+            <td><u>1163</u></td>
+            <td><u>1131</u></td>
+            <td><strong>4.2</strong></td>
+            <td>9.8</td>
+        </tr>
+    </tbody>
+</table>
+</div>
+All results are from AudioEvals, and the evaluation methods along with further details can be found in <a href="https://github.com/OpenBMB/UltraEval-Audio" target="_blank">AudioEvals</a>.<br><br>
+
+**End-to-end Voice Cloning**
+
+<div align="center">
+<table style="margin: 0px auto;">
+    <thead>
+        <tr>
+            <th align="left">Task</th>
+            <th colspan="2">Voice cloning</th>
+        </tr>
+        <tr>
+            <th align="left">Metric</th>
+            <th>SIMO↑</th>
+            <th>SIMO↑</th>
+        </tr>
+        <tr>
+            <th align="left">Dataset</th>
+            <th>Seed-TTS test-zh</th>
+            <th>Seed-TTS test-en</th>
+        </tr>
+    </thead>
+    <tbody align="center">
+        <tr>
+            <td nowrap="nowrap" align="left">F5-TTS</td>
+            <td><strong>76</strong></td>
+            <td><strong>67</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">CosyVoice</td>
+            <td><u>75</u></td>
+            <td><u>64</u></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">FireRedTTS</td>
+            <td>63</td>
+            <td>46</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">MiniCPM-o 2.6</td>
+            <td>57</td>
+            <td>47</td>
+        </tr>
+    </tbody>
+</table>
+</div>
+
+</details>
+
+<details>
+<summary>Click to view multimodal live streaming results.</summary>
+  
+**Multimodal Live Streaming**: results on StreamingBench
+
+<table style="margin: 0px auto;">
+    <thead>
+        <tr>
+            <th align="left">Model</th>
+            <th>Size</th>
+            <th>Real-Time Video Understanding</th>
+            <th>Omni-Source Understanding</th>
+            <th>Contextual Understanding</th>
+            <th>Overall</th>
+        </tr>
+    </thead>
+    <tbody align="center">
+        <tr>
+            <td colspan="7" align="left"><strong>Proprietary</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Gemini 1.5 Pro</td>
+            <td>-</td>
+            <td><u>77.4</u></td>
+            <td><strong>67.8</strong></td>
+            <td><strong>51.1</strong></td>
+            <td><strong>70.3</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">GPT-4o-202408</td>
+            <td>-</td>
+            <td>74.5</td>
+            <td>51.0</td>
+            <td><u>48.0</u></td>
+            <td>64.1</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Claude-3.5-Sonnet</td>
+            <td>-</td>
+            <td>74.0</td>
+            <td>41.4</td>
+            <td>37.8</td>
+            <td>59.7</td>
+        </tr>
+        <tr>
+            <td colspan="9" align="left"><strong>Open-source</strong></td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">VILA-1.5</td>
+            <td>8B</td>
+            <td>61.5</td>
+            <td>37.5</td>
+            <td>26.7</td>
+            <td>49.5</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">LongVA</td>
+            <td>7B</td>
+            <td>63.1</td>
+            <td>35.9</td>
+            <td>30.2</td>
+            <td>50.7</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">LLaVA-Next-Video-34B</td>
+            <td>34B</td>
+            <td>69.8</td>
+            <td>41.7</td>
+            <td>34.3</td>
+            <td>56.7</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">Qwen2-VL-7B</td>
+            <td>8B</td>
+            <td>71.2</td>
+            <td>40.7</td>
+            <td>33.1</td>
+            <td>57.0</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">InternVL2-8B</td>
+            <td>8B</td>
+            <td>70.1</td>
+            <td>42.7</td>
+            <td>34.1</td>
+            <td>57.0</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">VITA-1.5</td>
+            <td>8B</td>
+            <td>70.9</td>
+            <td>40.8</td>
+            <td>35.8</td>
+            <td>57.4</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">LLaVA-OneVision-7B</td>
+            <td>8B</td>
+            <td>74.3</td>
+            <td>40.8</td>
+            <td>31.0</td>
+            <td>58.4</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">InternLM-XC2.5-OL-7B</td>
+            <td>8B</td>
+            <td>75.4</td>
+            <td>46.2</td>
+            <td>33.6</td>
+            <td>60.8</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">MiniCPM-V 2.6</td>
+            <td>8B</td>
+            <td>72.4</td>
+            <td>40.2</td>
+            <td>33.4</td>
+            <td>57.7</td>
+        </tr>
+        <tr>
+            <td nowrap="nowrap" align="left">MiniCPM-o 2.6</td>
+            <td>8B</td>
+            <td><strong>79.9</strong></td>
+            <td><u>53.4</u></td>
+            <td>38.5</td>
+            <td><u>66.0</u></td>
+        </tr>
+    </tbody>
+</table>
+
+</details>
+
+
+### Examples <!-- omit in toc -->
+
+We deploy MiniCPM-o 2.6 on end devices. The demo video is the raw-speed recording on an iPad Pro and a Web demo.
+
+<div align="center">
+  <a href="https://youtu.be/JFJg9KZ_iZk"><img src="https://github.com/OpenBMB/MiniCPM-o/raw/main/assets/o-2dot6-demo-video-preview.png", width=70%></a>
+</div>
+
+<br>
+
+
+<div style="display: flex; flex-direction: column; align-items: center;">
+  <img src="https://github.com/OpenBMB/MiniCPM-o/raw/main/assets/minicpmo2_6/minicpmo2_6_math_intersect.png" alt="math" style="margin-bottom: 5px;">
+  <img src="https://github.com/OpenBMB/MiniCPM-o/raw/main/assets/minicpmo2_6/minicpmo2_6_diagram_train_NN.png" alt="diagram" style="margin-bottom: 5px;">
+  <img src="https://github.com/OpenBMB/MiniCPM-o/raw/main/assets/minicpmo2_6/minicpmo2_6_multi-image_bike.png" alt="bike" style="margin-bottom: 5px;">
+</div>
+
+
+
+
+## Online Demo
+Click here to try the online demo of [MiniCPM-o 2.6](https://minicpm-omni-webdemo-us.modelbest.cn).
+
+
+## Usage
+Inference using Huggingface transformers on NVIDIA GPUs. Requirements tested on python 3.10：
+```
+Pillow==10.1.0
+torch==2.2.0
+torchaudio==2.2.0
+torchvision==0.17.0
+transformers==4.44.2
+librosa==0.9.0
+soundfile==0.12.1
+vector-quantize-pytorch==1.18.5
+vocos==0.1.0
+decord
+moviepy
+```
+
+
+### Model initialization
+```python
+
+import torch
+from PIL import Image
+from transformers import AutoModel, AutoTokenizer
+
+# load omni model default, the default init_vision/init_audio/init_tts is True
+# if load vision-only model, please set init_audio=False and init_tts=False
+# if load audio-only model, please set init_vision=False
+model = AutoModel.from_pretrained(
+    'openbmb/MiniCPM-o-2_6',
+    trust_remote_code=True,
+    attn_implementation='sdpa', # sdpa or flash_attention_2
+    torch_dtype=torch.bfloat16,
+    init_vision=True,
+    init_audio=True,
+    init_tts=True
+)
+
+
+model = model.eval().cuda()
+tokenizer = AutoTokenizer.from_pretrained('openbmb/MiniCPM-o-2_6', trust_remote_code=True)
+
+# In addition to vision-only mode, tts processor and vocos also needs to be initialized
+model.init_tts()
+model.tts.float()
+```
+### Omni mode
+we provide two inference modes: chat and streaming
+
+#### Chat inference
+```python
+import math
+import numpy as np
+from PIL import Image
+from moviepy.editor import VideoFileClip
+import tempfile
+import librosa
+import soundfile as sf
+
+def get_video_chunk_content(video_path, flatten=True):
+    video = VideoFileClip(video_path)
+    print('video_duration:', video.duration)
+    
+    with tempfile.NamedTemporaryFile(suffix=".wav", delete=True) as temp_audio_file:
+        temp_audio_file_path = temp_audio_file.name
+        video.audio.write_audiofile(temp_audio_file_path, codec="pcm_s16le", fps=16000)
+        audio_np, sr = librosa.load(temp_audio_file_path, sr=16000, mono=True)
+    num_units = math.ceil(video.duration)
+    
+    # 1 frame + 1s audio chunk
+    contents= []
+    for i in range(num_units):
+        frame = video.get_frame(i+1)
+        image = Image.fromarray((frame).astype(np.uint8))
+        audio = audio_np[sr*i:sr*(i+1)]
+        if flatten:
+            contents.extend(["<unit>", image, audio])
+        else:
+            contents.append(["<unit>", image, audio])
+    
+    return contents
+
+video_path="/path/to/video"
+# if use voice clone prompt, please set ref_audio
+ref_audio_path = 'assets/demo.wav'
+ref_audio, _ = librosa.load(ref_audio_path, sr=16000, mono=True)
+sys_msg = model.get_sys_prompt(ref_audio=ref_audio, mode='omni', language='en')
+# or use default prompt
+# sys_msg = model.get_sys_prompt(mode='omni', language='en')
+
+contents = get_video_chunk_content(video_path)
+msg = {"role":"user", "content": contents}
+msgs = [sys_msg, msg]
+
+# please set generate_audio=True and output_audio_path to save the tts result
+generate_audio = True
+output_audio_path = 'output.wav'
+
+res = model.chat(
+    msgs=msgs,
+    tokenizer=tokenizer,
+    sampling=True,
+    temperature=0.5,
+    max_new_tokens=4096,
+    omni_input=True, # please set omni_input=True when omni inference
+    use_tts_template=True,
+    generate_audio=generate_audio,
+    output_audio_path=output_audio_path,
+    max_slice_nums=1,
+    use_image_id=False,
+    return_dict=True
+)
+print(res)
+```
+#### Streaming inference
+```python
+# a new conversation need reset session first, it will reset the kv-cache
+model.reset_session()
+
+contents = get_video_chunk_content(video_path, flatten=False)
+session_id = '123'
+generate_audio = True
+
+# 1. prefill system prompt
+res = model.streaming_prefill(
+    session_id=session_id,
+    msgs=[sys_msg], 
+    tokenizer=tokenizer
+)
+
+# 2. prefill video/audio chunks
+for content in contents:
+    msgs = [{"role":"user", "content": content}]
+    res = model.streaming_prefill(
+        session_id=session_id,
+        msgs=msgs, 
+        tokenizer=tokenizer
+    )
+
+# 3. generate
+res = model.streaming_generate(
+    session_id=session_id,
+    tokenizer=tokenizer,
+    temperature=0.5,
+    generate_audio=generate_audio
+)
+
+audios = []
+text = ""
+
+if generate_audio:
+    for r in res:
+        audio_wav = r.audio_wav
+        sampling_rate = r.sampling_rate
+        txt = r.text
+
+        audios.append(audio_wav)
+        text += txt
+        
+    res = np.concatenate(audios)
+    sf.write("output.wav", res, samplerate=sampling_rate)
+    print("text:", text)
+    print("audio saved to output.wav")
+else:
+    for r in res:
+        text += r['text']
+    print("text:", text)
+
+```
+
+### Audio-Only mode
+#### Mimick
+`Mimick` task reflects a model's end-to-end speech modeling capability. The model takes audio input, and outputs an ASR transcription and subsequently reconstructs the original audio with high similarity. The higher the similarity between the reconstructed audio and the original audio, the stronger the model's foundational capability in end-to-end speech modeling.
+```python
+mimick_prompt = "Please repeat each user's speech, including voice style and speech content."
+audio_input, _ = librosa.load('xxx.wav', sr=16000, mono=True)
+msgs = [{'role': 'user', 'content': [mimick_prompt,audio_input]}]
+
+res = model.chat(
+    msgs=msgs,
+    tokenizer=tokenizer,
+    sampling=True,
+    max_new_tokens=128,
+    use_tts_template=True,
+    temperature=0.3,
+    generate_audio=True,
+    output_audio_path='output.wav', # save the tts result to output_audio_path
+)
+```
+
+#### General Speech Conversation with Configurable Voices
+<details> <summary>Click to view the Python code for enabling MiniCPM-o 2.6 to interact with you in a specified voice.</summary>
+
+```python
+ref_audio, _ = librosa.load('assets/demo.wav', sr=16000, mono=True) # load the reference audio
+
+# Choose the mode you want to use
+# Audio RolePlay:  # With this mode, model will role-play the character based on the audio prompt. (More human-like conversation but unstable)
+# sys_prompt = model.get_sys_prompt(ref_audio=ref_audio, mode='audio_roleplay', language='en')
+# user_question = {'role': 'user', 'content': [librosa.load('xxx.wav', sr=16000, mono=True)[0]]}
+
+Audio Assistant: # With this mode, model will speak with the voice in ref_audio as a AI assistant. (Stable and more suitable for general conversation)
+sys_prompt = model.get_sys_prompt(ref_audio=ref_audio, mode='audio_assistant', language='en') 
+user_question = {'role': 'user', 'content': [librosa.load('xxx.wav', sr=16000, mono=True)[0]]} # Try to ask something by recording it in 'xxx.wav'!!!
+```
+```python
+msgs = [sys_prompt, user_question]
+# round one
+res = model.chat(
+    msgs=msgs,
+    tokenizer=tokenizer,
+    sampling=True,
+    max_new_tokens=128,
+    use_tts_template=True,
+    generate_audio=True,
+    temperature=0.3,
+    output_audio_path='result.wav',
+)
+
+# round two
+history = msgs.append({'role': 'assistant', 'content': res})
+user_question = {'role': 'user', 'content': [librosa.load('xxx.wav', sr=16000, mono=True)[0]]}
+msgs = history.append(user_question)
+res = model.chat(
+    msgs=msgs,
+    tokenizer=tokenizer,
+    sampling=True,
+    max_new_tokens=128,
+    use_tts_template=True,
+    generate_audio=True,
+    temperature=0.3,
+    output_audio_path='result_round_2.wav',
+)
+print(res)
+```
+
+</details>
+
+#### Addressing various audio tasks
+<details>
+<summary> Click to show Python code running MiniCPM-o 2.6 with specific audioQA task. </summary>
+
+```python
+'''
+Audio Understanding Task Prompt:
+Speech:
+    ASR with ZH(same as AST en2zh): 请仔细听这段音频片段，并将其内容逐字记录。
+    ASR with EN(same as AST zh2en): Please listen to the audio snippet carefully and transcribe the content.
+    Speaker Analysis: Based on the speaker's content, speculate on their gender, condition, age range, and health status.
+General Audio:
+    Audio Caption: Summarize the main content of the audio.
+    Sound Scene Tagging: Utilize one keyword to convey the audio's content or the associated scene.
+'''
+task_prompt = "" # Choose the task prompt above
+audio_input, _ = librosa.load('xxx.wav', sr=16000, mono=True)
+
+msgs = [{'role': 'user', 'content': [task_prompt,audio_input]}]
+
+res = model.chat(
+    msgs=msgs,
+    tokenizer=tokenizer,
+    sampling=True,
+    max_new_tokens=128,
+    use_tts_template=True,
+    generate_audio=True,
+    temperature=0.3,
+    output_audio_path='result.wav',
+)
+print(res)
+```
+```python
+'''
+Speech Generation Task Prompt:
+    Human Instruction-to-Speech: see https://voxinstruct.github.io/VoxInstruct/
+    Example:
+        # 在新闻中，一个年轻男性兴致勃勃地说：“祝福亲爱的祖国母亲美丽富强！”他用低音调和低音量，慢慢地说出了这句话。
+        # Delighting in a surprised tone, an adult male with low pitch and low volume comments:"One even gave my little dog a biscuit" This dialogue takes place at a leisurely pace, delivering a sense of excitement and surprise in the context. 
+
+    Voice Cloning or Voice Conversion: With this mode, model will act like a TTS model. 
+'''
+# Human Instruction-to-Speech:
+task_prompt = '' #Try to make some Human Instruction-to-Speech prompt (Voice Creation)
+msgs = [{'role': 'user', 'content': [task_prompt]}] # you can also try to ask the same audio question
+
+# Voice Cloning mode: 
+# sys_prompt = model.get_sys_prompt(ref_audio=ref_audio, mode='voice_cloning', language='en')
+# text_prompt = f"Please read the text below."
+# user_question = {'role': 'user', 'content': [text_prompt, "content that you want to read"]} # using same voice in sys_prompt to read the text. (Voice Cloning)
+# user_question = {'role': 'user', 'content': [text_prompt, librosa.load('xxx.wav', sr=16000, mono=True)[0]]} # using same voice in sys_prompt to read 'xxx.wav'. (Voice Conversion)
+# msgs = [sys_prompt, user_question]
+
+res = model.chat(
+    msgs=msgs,
+    tokenizer=tokenizer,
+    sampling=True,
+    max_new_tokens=128,
+    use_tts_template=True,
+    generate_audio=True,
+    temperature=0.3,
+    output_audio_path='result.wav',
+)
+
+
+```
+
+</details>
+
+### Vision-Only mode
+
+`MiniCPM-o-2_6` has the same inference methods as `MiniCPM-V-2_6`
+
+#### Chat with single image
+```python
+# test.py
+image = Image.open('xx.jpg').convert('RGB')
+question = 'What is in the image?'
+msgs = [{'role': 'user', 'content': [image, question]}]
+res = model.chat(
+    image=None,
+    msgs=msgs,
+    tokenizer=tokenizer
+)
+print(res)
+
+## if you want to use streaming, please make sure sampling=True and stream=True
+## the model.chat will return a generator
+res = model.chat(
+    msgs=msgs,
+    tokenizer=tokenizer,
+    sampling=True,
+    stream=True
+)
+generated_text = ""
+for new_text in res:
+    generated_text += new_text
+    print(new_text, flush=True, end='')
+```
+
+#### Chat with multiple images
+<details>
+<summary> Click to show Python code running MiniCPM-o 2.6 with multiple images input. </summary>
+  
+```python
+image1 = Image.open('image1.jpg').convert('RGB')
+image2 = Image.open('image2.jpg').convert('RGB')
+question = 'Compare image 1 and image 2, tell me about the differences between image 1 and image 2.'
+msgs = [{'role': 'user', 'content': [image1, image2, question]}]
+answer = model.chat(
+    msgs=msgs,
+    tokenizer=tokenizer
+)
+print(answer)
+```
+</details>
+
+#### In-context few-shot learning
+<details>
+<summary> Click to view Python code running MiniCPM-o 2.6 with few-shot input. </summary>
+
+```python
+question = "production date" 
+image1 = Image.open('example1.jpg').convert('RGB')
+answer1 = "2023.08.04"
+image2 = Image.open('example2.jpg').convert('RGB')
+answer2 = "2007.04.24"
+image_test = Image.open('test.jpg').convert('RGB')
+msgs = [
+    {'role': 'user', 'content': [image1, question]}, {'role': 'assistant', 'content': [answer1]},
+    {'role': 'user', 'content': [image2, question]}, {'role': 'assistant', 'content': [answer2]},
+    {'role': 'user', 'content': [image_test, question]}
+]
+answer = model.chat(
+    msgs=msgs,
+    tokenizer=tokenizer
+)
+print(answer)
+```
+</details>
+
+#### Chat with video
+<details>
+<summary> Click to view Python code running MiniCPM-o 2.6 with video input. </summary>
+
+```python
+MAX_NUM_FRAMES=64 # if cuda OOM set a smaller number
+def encode_video(video_path):
+    def uniform_sample(l, n):
+        gap = len(l) / n
+        idxs = [int(i * gap + gap / 2) for i in range(n)]
+        return [l[i] for i in idxs]
+    vr = VideoReader(video_path, ctx=cpu(0))
+    sample_fps = round(vr.get_avg_fps() / 1)  # FPS
+    frame_idx = [i for i in range(0, len(vr), sample_fps)]
+    if len(frame_idx) > MAX_NUM_FRAMES:
+        frame_idx = uniform_sample(frame_idx, MAX_NUM_FRAMES)
+    frames = vr.get_batch(frame_idx).asnumpy()
+    frames = [Image.fromarray(v.astype('uint8')) for v in frames]
+    print('num frames:', len(frames))
+    return frames
+video_path ="video_test.mp4"
+frames = encode_video(video_path)
+question = "Describe the video"
+msgs = [
+    {'role': 'user', 'content': frames + [question]}, 
+]
+# Set decode params for video
+params={}
+params["use_image_id"] = False
+params["max_slice_nums"] = 2 # use 1 if cuda OOM and video resolution >  448*448
+answer = model.chat(
+    msgs=msgs,
+    tokenizer=tokenizer,
+    **params
+)
+print(answer)
+```
+</details>
+
+Please look at [GitHub](https://github.com/OpenBMB/MiniCPM-o) for more detail about usage.
+
+
+## Inference with llama.cpp<a id="llamacpp"></a>
+MiniCPM-o 2.6 (vision-only mode) can run with llama.cpp. See our fork of [llama.cpp](https://github.com/OpenBMB/llama.cpp/tree/minicpm-omni) and [readme](https://github.com/OpenBMB/llama.cpp/blob/minicpm-omni/examples/llava/README-minicpmo2.6.md) for more detail.
+
+
+## Int4 quantized version
+Download the int4 quantized version for lower GPU memory (7GB) usage:  [MiniCPM-o-2_6-int4](https://huggingface.co/openbmb/MiniCPM-o-2_6-int4).
+
+
+## License
+#### Model License
+* The code in this repo is released under the [Apache-2.0](https://github.com/OpenBMB/MiniCPM/blob/main/LICENSE) License. 
+* The usage of MiniCPM-o and MiniCPM-V series model weights must strictly follow [MiniCPM Model License.md](https://github.com/OpenBMB/MiniCPM/blob/main/MiniCPM%20Model%20License.md).
+* The models and weights of MiniCPM are completely free for academic research. After filling out a ["questionnaire"](https://modelbest.feishu.cn/share/base/form/shrcnpV5ZT9EJ6xYjh3Kx0J6v8g) for registration, MiniCPM-o 2.6 weights are also available for free commercial use.
+
+
+#### Statement
+* As an LMM, MiniCPM-o 2.6 generates contents by learning a large mount of multimodal corpora, but it cannot comprehend, express personal opinions or make value judgement. Anything generated by MiniCPM-o 2.6 does not represent the views and positions of the model developers
+* We will not be liable for any problems arising from the use of the MinCPM-V models, including but not limited to data security issues, risk of public opinion, or any risks and problems arising from the misdirection, misuse, dissemination or misuse of the model.
+
+## Key Techniques and Other Multimodal Projects
+
+👏 Welcome to explore key techniques of MiniCPM-o 2.6 and other multimodal projects of our team:
+
+[VisCPM](https://github.com/OpenBMB/VisCPM/tree/main) | [RLHF-V](https://github.com/RLHF-V/RLHF-V) | [LLaVA-UHD](https://github.com/thunlp/LLaVA-UHD)  | [RLAIF-V](https://github.com/RLHF-V/RLAIF-V)
+
+## Citation
+
+If you find our work helpful, please consider citing our papers 📝 and liking this project ❤️！
+
+```bib
+@article{yao2024minicpm,
+  title={MiniCPM-V: A GPT-4V Level MLLM on Your Phone},
+  author={Yao, Yuan and Yu, Tianyu and Zhang, Ao and Wang, Chongyi and Cui, Junbo and Zhu, Hongji and Cai, Tianchi and Li, Haoyu and Zhao, Weilin and He, Zhihui and others},
+  journal={arXiv preprint arXiv:2408.01800},
+  year={2024}
+}
+```

added_tokens.json

@@ -0,0 +1,59 @@
+{
+  "</asr>": 151682,
+  "</box>": 151670,
+  "</image>": 151666,
+  "</image_id>": 151678,
+  "</point>": 151674,
+  "</quad>": 151672,
+  "</query>": 151684,
+  "</ref>": 151668,
+  "</slice>": 151676,
+  "</tool_call>": 151658,
+  "</unit>": 151680,
+  "<asr>": 151681,
+  "<box>": 151669,
+  "<image>": 151665,
+  "<image_id>": 151677,
+  "<point>": 151673,
+  "<quad>": 151671,
+  "<query>": 151683,
+  "<ref>": 151667,
+  "<reserved_43>": 151698,
+  "<reserved_53>": 151699,
+  "<slice>": 151675,
+  "<tool_call>": 151657,
+  "<unit>": 151679,
+  "<|audio_end|>": 151687,
+  "<|audio_start|>": 151685,
+  "<|audio|>": 151686,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|interrupt|>": 151695,
+  "<|listen|>": 151693,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|speak|>": 151694,
+  "<|spk_bos|>": 151688,
+  "<|spk_eos|>": 151690,
+  "<|spk|>": 151689,
+  "<|tts_bos|>": 151691,
+  "<|tts_eos|>": 151692,
+  "<|vad_end|>": 151697,
+  "<|vad_start|>": 151696,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

config.json

@@ -0,0 +1,208 @@
+{
+  "_name_or_path": "openbmb/MiniCPM-o-2_6",
+  "architectures": [
+    "MiniCPMO"
+  ],
+
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 3584,
+  "initializer_range": 0.02,
+  "intermediate_size": 18944,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 28,
+  "num_attention_heads": 28,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 4,
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 1000000.0,
+  "sliding_window": 131072,
+  "tie_word_embeddings": false,
+  "use_sliding_window": false,
+  "vocab_size": 151700,
+  "batch_vision_input": true,
+  "drop_vision_last_layer": false,
+  "image_size": 448,
+
+  "audio_chunk_length": 1.0,
+  "audio_config": {
+    "_name_or_path": "openai/whisper-medium",
+    "architectures": [
+      "MiniCPMWhisperEncoder"
+    ],
+    "begin_suppress_tokens": [
+      220,
+      50257
+    ],
+    "bos_token_id": 50257,
+    "d_model": 1024,
+    "decoder_attention_heads": 16,
+    "decoder_ffn_dim": 4096,
+    "decoder_layers": 24,
+    "decoder_start_token_id": 50258,
+    "encoder_attention_heads": 16,
+    "encoder_ffn_dim": 4096,
+    "encoder_layers": 24,
+    "eos_token_id": 50257,
+    "forced_decoder_ids": [
+      [
+        1,
+        50259
+      ],
+      [
+        2,
+        50359
+      ],
+      [
+        3,
+        50363
+      ]
+    ],
+    "max_length": 448,
+    "model_type": "whisper",
+    "num_hidden_layers": 24,
+    "pad_token_id": 50257,
+    "suppress_tokens": [
+      1,
+      2,
+      7,
+      8,
+      9,
+      10,
+      14,
+      25,
+      26,
+      27,
+      28,
+      29,
+      31,
+      58,
+      59,
+      60,
+      61,
+      62,
+      63,
+      90,
+      91,
+      92,
+      93,
+      359,
+      503,
+      522,
+      542,
+      873,
+      893,
+      902,
+      918,
+      922,
+      931,
+      1350,
+      1853,
+      1982,
+      2460,
+      2627,
+      3246,
+      3253,
+      3268,
+      3536,
+      3846,
+      3961,
+      4183,
+      4667,
+      6585,
+      6647,
+      7273,
+      9061,
+      9383,
+      10428,
+      10929,
+      11938,
+      12033,
+      12331,
+      12562,
+      13793,
+      14157,
+      14635,
+      15265,
+      15618,
+      16553,
+      16604,
+      18362,
+      18956,
+      20075,
+      21675,
+      22520,
+      26130,
+      26161,
+      26435,
+      28279,
+      29464,
+      31650,
+      32302,
+      32470,
+      36865,
+      42863,
+      47425,
+      49870,
+      50254,
+      50258,
+      50358,
+      50359,
+      50360,
+      50361,
+      50362
+    ],
+    "torch_dtype": "float32"
+  },
+  "audio_pool_step": 2,
+  "auto_map": {
+    "AutoConfig": "configuration_minicpm.MiniCPMOConfig",
+    "AutoModel": "modeling_minicpmo.MiniCPMO",
+    "AutoModelForCausalLM": "modeling_minicpmo.MiniCPMO"
+  },
+  "chunk_input": true,
+  "listen_speak_type": "asr",
+  "model_type": "minicpmo",
+  "patch_size": 14,
+  "query_num": 64,
+  "slice_config": {
+    "max_slice_nums": 9,
+    "model_type": "minicpmv"
+  },
+  "slice_mode": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.44.2",
+  "tts_config": {
+    "model_type": "conditional_chattts",
+    "llm_dim": 3584
+  },
+  "use_cache": true,
+  "use_image_id": true,
+  "version": 2.6,
+  "vision_batch_size": 16,
+  "vision_config": {
+    "hidden_size": 1152,
+    "image_size": 980,
+    "intermediate_size": 4304,
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 27,
+    "patch_size": 14
+  },
+  "quantization_config": {
+    "bits": 4,
+    "checkpoint_format": "gptq",
+    "damp_percent": 0.01,
+    "desc_act": true,
+    "group_size": -1,
+    "lm_head": false,
+    "model_file_base_name": null,
+    "model_name_or_path": null,
+    "quant_method": "gptq",
+    "static_groups": false,
+    "sym": true,
+    "true_sequential": true
+  }
+}

configuration_minicpm.py

@@ -0,0 +1,210 @@
+# coding=utf-8
+# Copyright 2025 The OpenBMB Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+from typing import Union
+
+from transformers import PretrainedConfig
+from transformers import Qwen2Config
+from transformers import WhisperConfig
+from transformers.utils import logging
+
+from .modeling_navit_siglip import SiglipVisionConfig
+
+logger = logging.get_logger(__name__)
+
+
+class MiniCPMVSliceConfig(PretrainedConfig):
+    model_type = "minicpmv"
+
+    def __init__(
+        self,
+        patch_size=14,
+        max_slice_nums=9,
+        scale_resolution=448,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+        self.patch_size = patch_size
+        self.max_slice_nums = max_slice_nums
+        self.scale_resolution = scale_resolution
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> "PretrainedConfig":
+        cls._set_token_in_kwargs(kwargs)
+
+        config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
+
+        if config_dict.get("model_type") == "minicpmv":
+            config_dict = config_dict["slice_config"]
+
+        if "model_type" in config_dict and hasattr(cls, "model_type") and config_dict["model_type"] != cls.model_type:
+            logger.warning(
+                f"You are using a model of type {config_dict['model_type']} to instantiate a model of type "
+                f"{cls.model_type}. This is not supported for all configurations of models and can yield errors."
+            )
+
+        return cls.from_dict(config_dict, **kwargs)
+
+
+class ConditionalChatTTSConfig(PretrainedConfig):
+    model_type = "conditional_chattts"
+
+    def __init__(
+        self,
+        llm_dim: int = 2560,
+        hidden_size: int = 768,
+        intermediate_size: int = 3072,
+        num_attention_heads: int = 12,
+        num_hidden_layers: int = 20,
+        max_position_embeddings: int = 4096,
+        num_audio_tokens: int = 626,
+        num_text_tokens: int = 21178,
+        num_mel_bins: int = 100,
+        num_vq: int = 4,
+        use_speaker_embedding: bool = True,
+        use_llm_hidden_state: bool = False,
+        spk_emb_token_id: int = 21143,
+        num_spk_embs: int = 1,
+        audio_bos_token_id: int = 21132,
+        text_eos_token_id: int = 21133,
+        use_text: bool = True,
+        streaming: bool = True,
+        streaming_text_chunk_size: int = 10,
+        streaming_text_reserved_len: int = 300,
+        streaming_audio_chunk_size: int = 50,
+        attn_implementation: str = "sdpa",
+        use_mlp: bool = True,
+        aug_loss_weight: bool = True,
+        do_sample: bool = True,
+        top_p: float = 0.7,
+        top_k: int = 20,
+        repetition_penalty: float = 1.0,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.llm_dim = llm_dim
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_attention_heads = num_attention_heads
+        self.num_hidden_layers = num_hidden_layers
+        self.max_position_embeddings = max_position_embeddings
+        self.num_audio_tokens = num_audio_tokens
+        self.num_text_tokens = num_text_tokens
+        self.num_mel_bins = num_mel_bins
+        self.num_vq = num_vq
+        self.use_speaker_embedding = use_speaker_embedding
+        self.use_llm_hidden_state = use_llm_hidden_state
+        self.spk_emb_token_id = spk_emb_token_id
+        self.num_spk_embs = num_spk_embs
+        self.audio_bos_token_id = audio_bos_token_id
+        self.text_eos_token_id = text_eos_token_id
+        self.use_text = use_text
+        self.streaming = streaming
+        self.streaming_text_chunk_size = streaming_text_chunk_size
+        self.streaming_text_reserved_len = streaming_text_reserved_len
+        self.streaming_audio_chunk_size = streaming_audio_chunk_size
+        self.attn_implementation = attn_implementation
+        self.use_mlp = use_mlp
+        self.aug_loss_weight = aug_loss_weight
+        self.do_sample = do_sample
+        self.top_p = top_p
+        self.top_k = top_k
+        self.repetition_penalty = repetition_penalty
+
+
+class MiniCPMOConfig(Qwen2Config):
+    model_type = "minicpmo"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    default_vision_config = {
+        "hidden_size": 1152,
+        "image_size": 980,
+        "intermediate_size": 4304,
+        "model_type": "siglip",
+        "num_attention_heads": 16,
+        "num_hidden_layers": 27,
+        "patch_size": 14,
+    }
+
+    def __init__(
+        self,
+        use_cache=True,
+        query_num=64,
+        image_size=448,
+        drop_vision_last_layer=True,
+        batch_vision_input=True,
+        slice_config=None,
+        vision_config=None,
+        audio_config=None,
+        tts_config=None,
+        use_image_id=True,
+        vision_batch_size=16,
+        audio_pool_step=2,
+        audio_chunk_length=1.0,
+        stream_input=False,
+        init_vision=True,
+        init_audio=True,
+        init_tts=True,
+        **kwargs,
+    ):
+        self.use_cache = use_cache
+        self.query_num = query_num
+        self.image_size = image_size
+        self.drop_vision_last_layer = drop_vision_last_layer
+        self.batch_vision_input = batch_vision_input
+        self.use_image_id = use_image_id
+        self.vision_batch_size = vision_batch_size
+        self.audio_pool_step = audio_pool_step
+        self.audio_chunk_length = audio_chunk_length
+        self.stream_input = stream_input
+        self.init_vision = init_vision
+        self.init_audio = init_audio
+        self.init_tts = init_tts
+
+        if slice_config is None:
+            self.slice_config = MiniCPMVSliceConfig(max_slice_nums=1)
+        else:
+            self.slice_config = MiniCPMVSliceConfig(**slice_config)
+        self.slice_mode = True
+
+        # same as HuggingFaceM4/siglip-so400m-14-980-flash-attn2-navit add tgt_sizes
+        if vision_config is None:
+            self.vision_config = SiglipVisionConfig(**self.default_vision_config)
+            logger.info("vision_config is None, using default vision config")
+        elif isinstance(vision_config, dict):
+            self.vision_config = SiglipVisionConfig(**vision_config)
+        elif isinstance(vision_config, SiglipVisionConfig):
+            self.vision_config = vision_config
+
+        # same as openai/whisper-medium add use_cache
+        if audio_config is None:
+            self.audio_config = WhisperConfig()
+        elif isinstance(audio_config, dict):
+            self.audio_config = WhisperConfig(**audio_config)
+        elif isinstance(audio_config, WhisperConfig):
+            self.audio_config = audio_config
+
+        if tts_config is None:
+            self.tts_config = ConditionalChatTTSConfig()
+        elif isinstance(tts_config, dict):
+            self.tts_config = ConditionalChatTTSConfig(**tts_config)
+        elif isinstance(tts_config, ConditionalChatTTSConfig):
+            self.tts_config = tts_config
+
+        self.patch_size = self.vision_config.patch_size
+
+        super().__init__(**kwargs)

gitattributes

@@ -0,0 +1,39 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+*png filter=lfs diff=lfs merge=lfs -text
+*jpg filter=lfs diff=lfs merge=lfs -text
+*gif filter=lfs diff=lfs merge=lfs -text
+*.wav filter=lfs diff=lfs merge=lfs -text

image_processing_minicpmv.py

@@ -0,0 +1,407 @@
+# coding=utf-8
+# Copyright 2025 The OpenBMB Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import math
+from typing import Any
+from typing import Dict
+from typing import List
+from typing import Optional
+from typing import Union
+
+import numpy as np
+import PIL
+import PIL.Image
+import PIL.ImageSequence
+import torch
+from PIL import Image
+from transformers import AutoImageProcessor
+from transformers.image_processing_utils import BaseImageProcessor
+from transformers.image_processing_utils import BatchFeature
+from transformers.image_transforms import to_channel_dimension_format
+from transformers.image_utils import ChannelDimension
+from transformers.image_utils import infer_channel_dimension_format
+from transformers.image_utils import is_torch_tensor
+from transformers.image_utils import to_numpy_array
+from transformers.image_utils import valid_images
+from transformers.utils import is_torch_device
+from transformers.utils import is_torch_dtype
+from transformers.utils import requires_backends
+from transformers.utils import TensorType
+
+
+def recursive_converter(converter, value):
+    if isinstance(value, list):
+        new_value = []
+        for v in value:
+            new_value += [recursive_converter(converter, v)]
+        return new_value
+    else:
+        return converter(value)
+
+
+class MiniCPMOBatchFeature(BatchFeature):
+    r"""
+    Extend from BatchFeature for supporting various image size
+    """
+
+    def __init__(self, data: Optional[Dict[str, Any]] = None, tensor_type: Union[None, str, TensorType] = None):
+        super().__init__(data)
+        self.convert_to_tensors(tensor_type=tensor_type)
+
+    def convert_to_tensors(self, tensor_type: Optional[Union[str, TensorType]] = None):
+        if tensor_type is None:
+            return self
+
+        is_tensor, as_tensor = self._get_is_as_tensor_fns(tensor_type)
+
+        def converter(value):
+            try:
+                if not is_tensor(value):
+                    tensor = as_tensor(value)
+                    return tensor
+            except:  # noqa E722
+                if key == "overflowing_values":
+                    raise ValueError("Unable to create tensor returning overflowing values of different lengths. ")
+                raise ValueError(
+                    "Unable to create tensor, you should probably activate padding "
+                    "with 'padding=True' to have batched tensors with the same length."
+                )
+
+        for key, value in self.items():
+            self[key] = recursive_converter(converter, value)
+        return self
+
+    def to(self, *args, **kwargs) -> "MiniCPMOBatchFeature":
+        requires_backends(self, ["torch"])
+        import torch
+
+        def cast_tensor(v):
+            # check if v is a floating point
+            if torch.is_floating_point(v):
+                # cast and send to device
+                return v.to(*args, **kwargs)
+            elif device is not None:
+                return v.to(device=device)
+            else:
+                return v
+
+        new_data = {}
+        device = kwargs.get("device")
+        # Check if the args are a device or a dtype
+        if device is None and len(args) > 0:
+            # device should be always the first argument
+            arg = args[0]
+            if is_torch_dtype(arg):
+                # The first argument is a dtype
+                pass
+            elif isinstance(arg, str) or is_torch_device(arg) or isinstance(arg, int):
+                device = arg
+            else:
+                # it's something else
+                raise ValueError(f"Attempting to cast a BatchFeature to type {str(arg)}. This is not supported.")
+        # We cast only floating point tensors to avoid issues with tokenizers casting `LongTensor` to `FloatTensor`
+        for k, v in self.items():
+            new_data[k] = recursive_converter(cast_tensor, v)
+        self.data = new_data
+        return self
+
+
+class MiniCPMVImageProcessor(BaseImageProcessor):
+    model_input_names = ["pixel_values"]
+
+    def __init__(self, max_slice_nums=9, scale_resolution=448, patch_size=14, **kwargs):
+        super().__init__(**kwargs)
+        self.max_slice_nums = max_slice_nums
+        self.scale_resolution = scale_resolution
+        self.patch_size = patch_size
+        self.use_image_id = kwargs.pop("use_image_id", False)
+        self.image_feature_size = kwargs.pop("image_feature_size", 64)
+        self.im_start_token = kwargs.pop("im_start", "<image>")
+        self.im_end_token = kwargs.pop("im_end", "</image>")
+        self.slice_start_token = kwargs.pop("slice_start", "<slice>")
+        self.slice_end_token = kwargs.pop("slice_end", "</slice>")
+        self.unk_token = kwargs.pop("unk", "<unk>")
+        self.im_id_start = kwargs.pop("im_id_start", "<image_id>")
+        self.im_id_end = kwargs.pop("im_id_end", "</image_id>")
+        self.slice_mode = kwargs.pop("slice_mode", True)
+
+        self.mean = np.array(kwargs.pop("norm_mean", [0.5, 0.5, 0.5]))
+        self.std = np.array(kwargs.pop("norm_std", [0.5, 0.5, 0.5]))
+        self.version = kwargs.pop("version", 2.0)
+
+    def ensure_divide(self, length, patch_size):
+        return max(round(length / patch_size) * patch_size, patch_size)
+
+    def find_best_resize(self, original_size, scale_resolution, patch_size, allow_upscale=False):
+        width, height = original_size
+        if (width * height > scale_resolution * scale_resolution) or allow_upscale:
+            r = width / height
+            height = int(scale_resolution / math.sqrt(r))
+            width = int(height * r)
+        best_width = self.ensure_divide(width, patch_size)
+        best_height = self.ensure_divide(height, patch_size)
+        return (best_width, best_height)
+
+    def get_refine_size(self, original_size, grid, scale_resolution, patch_size, allow_upscale=False):
+        width, height = original_size
+        grid_x, grid_y = grid
+
+        refine_width = self.ensure_divide(width, grid_x)
+        refine_height = self.ensure_divide(height, grid_y)
+
+        grid_width = refine_width / grid_x
+        grid_height = refine_height / grid_y
+
+        best_grid_size = self.find_best_resize(
+            (grid_width, grid_height), scale_resolution, patch_size, allow_upscale=allow_upscale
+        )
+        refine_size = (best_grid_size[0] * grid_x, best_grid_size[1] * grid_y)
+        return refine_size
+
+    def split_to_patches(self, image, grid):
+        patches = []
+        width, height = image.size
+        grid_x = int(width / grid[0])
+        grid_y = int(height / grid[1])
+        for i in range(0, height, grid_y):
+            images = []
+            for j in range(0, width, grid_x):
+                box = (j, i, j + grid_x, i + grid_y)
+                patch = image.crop(box)
+                images.append(patch)
+            patches.append(images)
+        return patches
+
+    def slice_image(self, image, max_slice_nums=9, scale_resolution=448, patch_size=14, never_split=False):
+        original_size = image.size
+        source_image = None
+        best_grid = self.get_sliced_grid(original_size, max_slice_nums, never_split)
+        patches = []
+
+        if best_grid is None:
+            # dont need to slice, upsample
+            best_size = self.find_best_resize(original_size, scale_resolution, patch_size, allow_upscale=True)
+            source_image = image.resize(best_size, resample=Image.Resampling.BICUBIC)
+        else:
+            # source image, down-sampling and ensure divided by patch_size
+            best_resize = self.find_best_resize(original_size, scale_resolution, patch_size)
+            source_image = image.copy().resize(best_resize, resample=Image.Resampling.BICUBIC)
+            refine_size = self.get_refine_size(
+                original_size, best_grid, scale_resolution, patch_size, allow_upscale=True
+            )
+            refine_image = image.resize(refine_size, resample=Image.Resampling.BICUBIC)
+            patches = self.split_to_patches(refine_image, best_grid)
+
+        return source_image, patches, best_grid
+
+    def get_grid_placeholder(self, grid):
+        if grid is None:
+            return ""
+        slice_image_placeholder = (
+            self.slice_start_token + self.unk_token * self.image_feature_size + self.slice_end_token
+        )
+
+        cols = grid[0]
+        rows = grid[1]
+        slices = []
+        for i in range(rows):
+            lines = []
+            for j in range(cols):
+                lines.append(slice_image_placeholder)
+            slices.append("".join(lines))
+
+        slice_placeholder = "\n".join(slices)
+        return slice_placeholder
+
+    def get_image_id_placeholder(self, idx=0):
+        return f"{self.im_id_start}{idx}{self.im_id_end}"
+
+    def get_sliced_images(self, image, max_slice_nums=None):
+        slice_images = []
+
+        if not self.slice_mode:
+            return [image]
+
+        max_slice_nums = self.max_slice_nums if max_slice_nums is None else int(max_slice_nums)
+        assert max_slice_nums > 0
+        source_image, patches, sliced_grid = self.slice_image(
+            image, max_slice_nums, self.scale_resolution, self.patch_size  # default: 9  # default: 448  # default: 14
+        )
+
+        slice_images.append(source_image)
+        if len(patches) > 0:
+            for i in range(len(patches)):
+                for j in range(len(patches[0])):
+                    slice_images.append(patches[i][j])
+        return slice_images
+
+    def get_sliced_grid(self, image_size, max_slice_nums, nerver_split=False):
+        original_width, original_height = image_size
+        log_ratio = math.log(original_width / original_height)
+        ratio = original_width * original_height / (self.scale_resolution * self.scale_resolution)
+        multiple = min(math.ceil(ratio), max_slice_nums)
+        if multiple <= 1 or nerver_split:
+            return None
+        candidate_split_grids_nums = []
+        for i in [multiple - 1, multiple, multiple + 1]:
+            if i == 1 or i > max_slice_nums:
+                continue
+            candidate_split_grids_nums.append(i)
+
+        candidate_grids = []
+        for split_grids_nums in candidate_split_grids_nums:
+            m = 1
+            while m <= split_grids_nums:
+                if split_grids_nums % m == 0:
+                    candidate_grids.append([m, split_grids_nums // m])
+                m += 1
+
+        best_grid = [1, 1]
+        min_error = float("inf")
+        for grid in candidate_grids:
+            error = abs(log_ratio - math.log(grid[0] / grid[1]))
+            if error < min_error:
+                best_grid = grid
+                min_error = error
+
+        return best_grid
+
+    def get_slice_image_placeholder(self, image_size, image_idx=0, max_slice_nums=None, use_image_id=None):
+        max_slice_nums = self.max_slice_nums if max_slice_nums is None else int(max_slice_nums)
+        assert max_slice_nums > 0
+        grid = self.get_sliced_grid(image_size=image_size, max_slice_nums=max_slice_nums)
+
+        image_placeholder = self.im_start_token + self.unk_token * self.image_feature_size + self.im_end_token
+        use_image_id = self.use_image_id if use_image_id is None else bool(use_image_id)
+        if use_image_id:
+            final_placeholder = self.get_image_id_placeholder(image_idx) + image_placeholder
+        else:
+            final_placeholder = image_placeholder
+
+        if self.slice_mode:
+            final_placeholder = final_placeholder + self.get_grid_placeholder(grid=grid)
+        return final_placeholder
+
+    def to_pil_image(self, image, rescale=None) -> PIL.Image.Image:
+        """
+        Converts `image` to a PIL Image. Optionally rescales it and puts the channel dimension back as the last axis if
+        needed.
+
+        Args:
+            image (`PIL.Image.Image` or `numpy.ndarray` or `torch.Tensor`):
+                The image to convert to the PIL Image format.
+            rescale (`bool`, *optional*):
+                Whether or not to apply the scaling factor (to make pixel values integers between 0 and 255). Will
+                default to `True` if the image type is a floating type, `False` otherwise.
+        """
+        if isinstance(image, PIL.Image.Image):
+            return image
+        if is_torch_tensor(image):
+            image = image.numpy()
+
+        if isinstance(image, np.ndarray):
+            if rescale is None:
+                # rescale default to the array being of floating type.
+                rescale = isinstance(image.flat[0], np.floating)
+            # If the channel as been moved to first dim, we put it back at the end.
+            if image.ndim == 3 and image.shape[0] in [1, 3]:
+                image = image.transpose(1, 2, 0)
+            if rescale:
+                image = image * 255
+            image = image.astype(np.uint8)
+            return PIL.Image.fromarray(image)
+        return image
+
+    def reshape_by_patch(self, image):
+        """
+        :param image: shape [3, H, W]
+        :param patch_size:
+        :return: [3, patch_size, HW/patch_size]
+        """
+        image = torch.from_numpy(image)
+        patch_size = self.patch_size
+        patches = torch.nn.functional.unfold(image, (patch_size, patch_size), stride=(patch_size, patch_size))
+
+        patches = patches.reshape(image.size(0), patch_size, patch_size, -1)
+        patches = patches.permute(0, 1, 3, 2).reshape(image.size(0), patch_size, -1)
+        return patches.numpy()
+
+    def preprocess(
+        self,
+        images: Union[Image.Image, List[Image.Image], List[List[Image.Image]]],
+        do_pad: Optional[bool] = True,
+        max_slice_nums: int = None,
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        **kwargs,
+    ) -> MiniCPMOBatchFeature:
+        if isinstance(images, Image.Image):
+            images_list = [[images]]
+        elif isinstance(images[0], Image.Image):
+            images_list = [images]
+        else:
+            images_list = images
+
+        new_images_list = []
+        image_sizes_list = []
+        tgt_sizes_list = []
+
+        for _images in images_list:
+            if _images is None or len(_images) == 0:
+                new_images_list.append([])
+                image_sizes_list.append([])
+                tgt_sizes_list.append([])
+                continue
+            if not valid_images(_images):
+                raise ValueError(
+                    "Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, "
+                    "torch.Tensor, tf.Tensor or jax.ndarray."
+                )
+
+            _images = [self.to_pil_image(image).convert("RGB") for image in _images]
+            input_data_format = infer_channel_dimension_format(np.array(_images[0]))
+
+            new_images = []
+            image_sizes = [image.size for image in _images]
+            tgt_sizes = []
+            for image in _images:
+                image_patches = self.get_sliced_images(image, max_slice_nums)
+                image_patches = [to_numpy_array(image).astype(np.float32) / 255 for image in image_patches]
+                image_patches = [
+                    self.normalize(image=image, mean=self.mean, std=self.std, input_data_format=input_data_format)
+                    for image in image_patches
+                ]
+                image_patches = [
+                    to_channel_dimension_format(image, ChannelDimension.FIRST, input_channel_dim=input_data_format)
+                    for image in image_patches
+                ]
+                for slice_image in image_patches:
+                    new_images.append(self.reshape_by_patch(slice_image))
+                    tgt_sizes.append(
+                        np.array((slice_image.shape[1] // self.patch_size, slice_image.shape[2] // self.patch_size))
+                    )
+
+            if tgt_sizes:
+                tgt_sizes = np.vstack(tgt_sizes)
+
+            new_images_list.append(new_images)
+            image_sizes_list.append(image_sizes)
+            tgt_sizes_list.append(tgt_sizes)
+        return MiniCPMOBatchFeature(
+            data={"pixel_values": new_images_list, "image_sizes": image_sizes_list, "tgt_sizes": tgt_sizes_list},
+            tensor_type=return_tensors,
+        )
+
+
+AutoImageProcessor.register("MiniCPMVImageProcessor", MiniCPMVImageProcessor)

modeling_navit_siglip.py

@@ -0,0 +1,940 @@
+# coding=utf-8
+# Copyright 2024 Google AI and The HuggingFace Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+""" PyTorch Siglip model. """
+# Copied from  HuggingFaceM4/siglip-so400m-14-980-flash-attn2-navit and add tgt_sizes
+
+
+import math
+import os
+import warnings
+from dataclasses import dataclass
+from typing import Optional
+from typing import Tuple
+from typing import Union
+
+import numpy as np
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn.init import _calculate_fan_in_and_fan_out
+from transformers.activations import ACT2FN
+from transformers.configuration_utils import PretrainedConfig
+from transformers.modeling_attn_mask_utils import _prepare_4d_attention_mask
+from transformers.modeling_outputs import BaseModelOutput
+from transformers.modeling_outputs import BaseModelOutputWithPooling
+from transformers.modeling_utils import PreTrainedModel
+from transformers.utils import add_start_docstrings
+from transformers.utils import add_start_docstrings_to_model_forward
+from transformers.utils import is_flash_attn_2_available
+from transformers.utils import logging
+from transformers.utils import ModelOutput
+from transformers.utils import replace_return_docstrings
+
+logger = logging.get_logger(__name__)
+
+
+class SiglipVisionConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`SiglipVisionModel`]. It is used to instantiate a
+    Siglip vision encoder according to the specified arguments, defining the model architecture. Instantiating a
+    configuration with the defaults will yield a similar configuration to that of the vision encoder of the Siglip
+    [google/siglip-base-patch16-224](https://huggingface.co/google/siglip-base-patch16-224) architecture.
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+    Args:
+        hidden_size (`int`, *optional*, defaults to 768):
+            Dimensionality of the encoder layers and the pooler layer.
+        intermediate_size (`int`, *optional*, defaults to 3072):
+            Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder.
+        num_hidden_layers (`int`, *optional*, defaults to 12):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 12):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_channels (`int`, *optional*, defaults to 3):
+            Number of channels in the input images.
+        image_size (`int`, *optional*, defaults to 224):
+            The size (resolution) of each image.
+        patch_size (`int`, *optional*, defaults to 16):
+            The size (resolution) of each patch.
+        hidden_act (`str` or `function`, *optional*, defaults to `"gelu_pytorch_tanh"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
+            `"relu"`, `"selu"` and `"gelu_new"` ``"quick_gelu"` are supported.
+        layer_norm_eps (`float`, *optional*, defaults to 1e-06):
+            The epsilon used by the layer normalization layers.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+    Example:
+    ```python
+    >>> from transformers import SiglipVisionConfig, SiglipVisionModel
+    >>> # Initializing a SiglipVisionConfig with google/siglip-base-patch16-224 style configuration
+    >>> configuration = SiglipVisionConfig()
+    >>> # Initializing a SiglipVisionModel (with random weights) from the google/siglip-base-patch16-224 style configuration
+    >>> model = SiglipVisionModel(configuration)
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "siglip_vision_model"
+
+    def __init__(
+        self,
+        hidden_size=768,
+        intermediate_size=3072,
+        num_hidden_layers=12,
+        num_attention_heads=12,
+        num_channels=3,
+        image_size=224,
+        patch_size=16,
+        hidden_act="gelu_pytorch_tanh",
+        layer_norm_eps=1e-6,
+        attention_dropout=0.0,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_channels = num_channels
+        self.patch_size = patch_size
+        self.image_size = image_size
+        self.attention_dropout = attention_dropout
+        self.layer_norm_eps = layer_norm_eps
+        self.hidden_act = hidden_act
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> "PretrainedConfig":
+        cls._set_token_in_kwargs(kwargs)
+
+        config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
+
+        # get the vision config dict if we are loading from SiglipConfig
+        if config_dict.get("model_type") == "siglip":
+            config_dict = config_dict["vision_config"]
+
+        if "model_type" in config_dict and hasattr(cls, "model_type") and config_dict["model_type"] != cls.model_type:
+            logger.warning(
+                f"You are using a model of type {config_dict['model_type']} to instantiate a model of type "
+                f"{cls.model_type}. This is not supported for all configurations of models and can yield errors."
+            )
+
+        return cls.from_dict(config_dict, **kwargs)
+
+
+_CHECKPOINT_FOR_DOC = "google/siglip-base-patch16-224"
+
+SIGLIP_PRETRAINED_MODEL_ARCHIVE_LIST = [
+    "google/siglip-base-patch16-224",
+    # See all SigLIP models at https://huggingface.co/models?filter=siglip
+]
+
+if is_flash_attn_2_available():
+    from flash_attn import flash_attn_func
+    from flash_attn import flash_attn_varlen_func
+    from flash_attn.bert_padding import index_first_axis  # noqa
+    from flash_attn.bert_padding import pad_input
+    from flash_attn.bert_padding import unpad_input
+
+
+# Copied from transformers.models.llama.modeling_llama._get_unpad_data
+def _get_unpad_data(attention_mask):
+    seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
+    indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
+    max_seqlen_in_batch = seqlens_in_batch.max().item()
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0))
+    return (
+        indices,
+        cu_seqlens,
+        max_seqlen_in_batch,
+    )
+
+
+def _trunc_normal_(tensor, mean, std, a, b):
+    # Cut & paste from PyTorch official master until it's in a few official releases - RW
+    # Method based on https://people.sc.fsu.edu/~jburkardt/presentations/truncated_normal.pdf
+    def norm_cdf(x):
+        # Computes standard normal cumulative distribution function
+        return (1.0 + math.erf(x / math.sqrt(2.0))) / 2.0
+
+    if (mean < a - 2 * std) or (mean > b + 2 * std):
+        warnings.warn(
+            "mean is more than 2 std from [a, b] in nn.init.trunc_normal_. "
+            "The distribution of values may be incorrect.",
+            stacklevel=2,
+        )
+
+    # Values are generated by using a truncated uniform distribution and
+    # then using the inverse CDF for the normal distribution.
+    # Get upper and lower cdf values
+    l = norm_cdf((a - mean) / std)
+    u = norm_cdf((b - mean) / std)
+
+    # Uniformly fill tensor with values from [l, u], then translate to
+    # [2l-1, 2u-1].
+    tensor.uniform_(2 * l - 1, 2 * u - 1)
+
+    # Use inverse cdf transform for normal distribution to get truncated
+    # standard normal
+    if tensor.dtype in [torch.float16, torch.bfloat16]:
+        # The `erfinv_` op is not (yet?) defined in float16+cpu, bfloat16+gpu
+        og_dtype = tensor.dtype
+        tensor = tensor.to(torch.float32)
+        tensor.erfinv_()
+        tensor = tensor.to(og_dtype)
+    else:
+        tensor.erfinv_()
+
+    # Transform to proper mean, std
+    tensor.mul_(std * math.sqrt(2.0))
+    tensor.add_(mean)
+
+    # Clamp to ensure it's in the proper range
+    if tensor.dtype == torch.float16:
+        # The `clamp_` op is not (yet?) defined in float16+cpu
+        tensor = tensor.to(torch.float32)
+        tensor.clamp_(min=a, max=b)
+        tensor = tensor.to(torch.float16)
+    else:
+        tensor.clamp_(min=a, max=b)
+
+
+def trunc_normal_tf_(
+    tensor: torch.Tensor, mean: float = 0.0, std: float = 1.0, a: float = -2.0, b: float = 2.0
+) -> torch.Tensor:
+    """Fills the input Tensor with values drawn from a truncated
+    normal distribution. The values are effectively drawn from the
+    normal distribution :math:`\\mathcal{N}(\text{mean}, \text{std}^2)`
+    with values outside :math:`[a, b]` redrawn until they are within
+    the bounds. The method used for generating the random values works
+    best when :math:`a \\leq \text{mean} \\leq b`.
+    NOTE: this 'tf' variant behaves closer to Tensorflow / JAX impl where the
+    bounds [a, b] are applied when sampling the normal distribution with mean=0, std=1.0
+    and the result is subsquently scaled and shifted by the mean and std args.
+    Args:
+        tensor: an n-dimensional `torch.Tensor`
+        mean: the mean of the normal distribution
+        std: the standard deviation of the normal distribution
+        a: the minimum cutoff value
+        b: the maximum cutoff value
+    """
+    with torch.no_grad():
+        _trunc_normal_(tensor, 0, 1.0, a, b)
+        tensor.mul_(std).add_(mean)
+
+
+def variance_scaling_(tensor, scale=1.0, mode="fan_in", distribution="normal"):
+    fan_in, fan_out = _calculate_fan_in_and_fan_out(tensor)
+    if mode == "fan_in":
+        denom = fan_in
+    elif mode == "fan_out":
+        denom = fan_out
+    elif mode == "fan_avg":
+        denom = (fan_in + fan_out) / 2
+
+    variance = scale / denom
+
+    if distribution == "truncated_normal":
+        # constant is stddev of standard normal truncated to (-2, 2)
+        trunc_normal_tf_(tensor, std=math.sqrt(variance) / 0.87962566103423978)
+    elif distribution == "normal":
+        with torch.no_grad():
+            tensor.normal_(std=math.sqrt(variance))
+    elif distribution == "uniform":
+        bound = math.sqrt(3 * variance)
+        with torch.no_grad():
+            tensor.uniform_(-bound, bound)
+    else:
+        raise ValueError(f"invalid distribution {distribution}")
+
+
+def lecun_normal_(tensor):
+    variance_scaling_(tensor, mode="fan_in", distribution="truncated_normal")
+
+
+def default_flax_embed_init(tensor):
+    variance_scaling_(tensor, mode="fan_in", distribution="normal")
+
+
+@dataclass
+# Copied from transformers.models.clip.modeling_clip.CLIPVisionModelOutput with CLIP->Siglip
+class SiglipVisionModelOutput(ModelOutput):
+    """
+    Base class for vision model's outputs that also contains image embeddings of the pooling of the last hidden states.
+    Args:
+        image_embeds (`torch.FloatTensor` of shape `(batch_size, output_dim)` *optional* returned when model is initialized with `with_projection=True`):
+            The image embeddings obtained by applying the projection layer to the pooler_output.
+        last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
+            Sequence of hidden-states at the output of the last layer of the model.
+        hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
+            Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
+            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
+            Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.
+        attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
+            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
+            sequence_length)`.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
+    """
+
+    image_embeds: Optional[torch.FloatTensor] = None
+    last_hidden_state: torch.FloatTensor = None
+    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    attentions: Optional[Tuple[torch.FloatTensor]] = None
+
+
+class SiglipVisionEmbeddings(nn.Module):
+    def __init__(self, config: SiglipVisionConfig):
+        super().__init__()
+        self.config = config
+        self.embed_dim = config.hidden_size
+        self.image_size = config.image_size
+        self.patch_size = config.patch_size
+
+        self.patch_embedding = nn.Conv2d(
+            in_channels=config.num_channels,
+            out_channels=self.embed_dim,
+            kernel_size=self.patch_size,
+            stride=self.patch_size,
+            padding="valid",
+        )
+
+        self.num_patches_per_side = self.image_size // self.patch_size
+        self.num_patches = self.num_patches_per_side**2
+        self.num_positions = self.num_patches
+        self.position_embedding = nn.Embedding(self.num_positions, self.embed_dim)
+
+    def forward(
+        self,
+        pixel_values: torch.FloatTensor,
+        patch_attention_mask: torch.BoolTensor,
+        tgt_sizes: Optional[torch.IntTensor] = None,
+    ) -> torch.Tensor:
+        batch_size = pixel_values.size(0)
+
+        patch_embeds = self.patch_embedding(pixel_values)
+        embeddings = patch_embeds.flatten(2).transpose(1, 2)
+
+        max_im_h, max_im_w = pixel_values.size(2), pixel_values.size(3)
+        max_nb_patches_h, max_nb_patches_w = max_im_h // self.patch_size, max_im_w // self.patch_size
+        boundaries = torch.arange(1 / self.num_patches_per_side, 1.0, 1 / self.num_patches_per_side)
+        position_ids = torch.full(
+            size=(
+                batch_size,
+                max_nb_patches_h * max_nb_patches_w,
+            ),
+            fill_value=0,
+        )
+
+        for batch_idx, p_attn_mask in enumerate(patch_attention_mask):
+            if tgt_sizes is not None:
+                nb_patches_h = tgt_sizes[batch_idx][0]
+                nb_patches_w = tgt_sizes[batch_idx][1]
+            else:
+                nb_patches_h = p_attn_mask[:, 0].sum()
+                nb_patches_w = p_attn_mask[0].sum()
+
+            fractional_coords_h = torch.arange(0, 1 - 1e-6, 1 / nb_patches_h)
+            fractional_coords_w = torch.arange(0, 1 - 1e-6, 1 / nb_patches_w)
+
+            bucket_coords_h = torch.bucketize(fractional_coords_h, boundaries, right=True)
+            bucket_coords_w = torch.bucketize(fractional_coords_w, boundaries, right=True)
+
+            pos_ids = (bucket_coords_h[:, None] * self.num_patches_per_side + bucket_coords_w).flatten()
+            position_ids[batch_idx][p_attn_mask.view(-1).cpu()] = pos_ids
+
+        position_ids = position_ids.to(self.position_embedding.weight.device)
+
+        embeddings = embeddings + self.position_embedding(position_ids)
+        return embeddings
+
+
+class SiglipAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    # Copied from transformers.models.clip.modeling_clip.CLIPAttention.__init__
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.embed_dim = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.embed_dim // self.num_heads
+        if self.head_dim * self.num_heads != self.embed_dim:
+            raise ValueError(
+                f"embed_dim must be divisible by num_heads (got `embed_dim`: {self.embed_dim} and `num_heads`:"
+                f" {self.num_heads})."
+            )
+        self.scale = self.head_dim**-0.5
+        self.dropout = config.attention_dropout
+
+        self.k_proj = nn.Linear(self.embed_dim, self.embed_dim)
+        self.v_proj = nn.Linear(self.embed_dim, self.embed_dim)
+        self.q_proj = nn.Linear(self.embed_dim, self.embed_dim)
+        self.out_proj = nn.Linear(self.embed_dim, self.embed_dim)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = False,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        """Input shape: Batch x Time x Channel"""
+
+        batch_size, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(batch_size, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(batch_size, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(batch_size, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+
+        k_v_seq_len = key_states.shape[-2]
+        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) * self.scale
+
+        if attn_weights.size() != (batch_size, self.num_heads, q_len, k_v_seq_len):
+            raise ValueError(
+                f"Attention weights should be of size {(batch_size, self.num_heads, q_len, k_v_seq_len)}, but is"
+                f" {attn_weights.size()}"
+            )
+
+        if attention_mask is not None:
+            if attention_mask.size() != (batch_size, 1, q_len, k_v_seq_len):
+                raise ValueError(
+                    f"Attention mask should be of size {(batch_size, 1, q_len, k_v_seq_len)}, but is {attention_mask.size()}"
+                )
+            attn_weights = attn_weights + attention_mask
+
+        # upcast attention to fp32
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
+        attn_weights = nn.functional.dropout(attn_weights, p=self.dropout, training=self.training)
+        attn_output = torch.matmul(attn_weights, value_states)
+
+        if attn_output.size() != (batch_size, self.num_heads, q_len, self.head_dim):
+            raise ValueError(
+                f"`attn_output` should be of size {(batch_size, self.num_heads, q_len, self.head_dim)}, but is"
+                f" {attn_output.size()}"
+            )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.reshape(batch_size, q_len, self.embed_dim)
+
+        attn_output = self.out_proj(attn_output)
+
+        return attn_output, attn_weights
+
+
+class SiglipFlashAttention2(SiglipAttention):
+    """
+    Llama flash attention module. This module inherits from `LlamaAttention` as the weights of the module stays
+    untouched. The only required change would be on the forward pass where it needs to correctly call the public API of
+    flash attention and deal with padding tokens in case the input contains any of them.
+    """
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.is_causal = False  # Hack to make sure we don't use a causal mask
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        **kwargs,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        output_attentions = False
+
+        bsz, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        # Flash attention requires the input to have the shape
+        # batch_size x seq_length x head_dim x hidden_dim
+        # therefore we just need to keep the original shape
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        # cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+        # query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+
+        # if past_key_value is not None:
+        #     cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
+        #     key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        # TODO: These transpose are quite inefficient but Flash Attention requires the layout [batch_size, sequence_length, num_heads, head_dim]. We would need to refactor the KV cache
+        # to be able to avoid many of these transpose/reshape/view.
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+
+        dropout_rate = self.dropout if self.training else 0.0
+
+        # In PEFT, usually we cast the layer norms in float32 for training stability reasons
+        # therefore the input hidden states gets silently casted in float32. Hence, we need
+        # cast them back in the correct dtype just to be sure everything works as expected.
+        # This might slowdown training & inference so it is recommended to not cast the LayerNorms
+        # in fp32. (LlamaRMSNorm handles it correctly)
+
+        input_dtype = query_states.dtype
+        if input_dtype == torch.float32:
+            if torch.is_autocast_enabled():
+                target_dtype = torch.get_autocast_gpu_dtype()
+            # Handle the case where the model is quantized
+            elif hasattr(self.config, "_pre_quantization_dtype"):
+                target_dtype = self.config._pre_quantization_dtype
+            else:
+                target_dtype = self.q_proj.weight.dtype
+
+            logger.warning_once(
+                "The input hidden states seems to be silently casted in float32, this might be related to the fact"
+                " you have upcasted embedding or layer norm layers in float32. We will cast back the input in"
+                f" {target_dtype}."
+            )
+
+            query_states = query_states.to(target_dtype)
+            key_states = key_states.to(target_dtype)
+            value_states = value_states.to(target_dtype)
+
+        attn_output = self._flash_attention_forward(
+            query_states, key_states, value_states, attention_mask, q_len, dropout=dropout_rate
+        )
+
+        attn_output = attn_output.reshape(bsz, q_len, self.embed_dim).contiguous()
+        attn_output = self.out_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights
+
+    def _flash_attention_forward(
+        self, query_states, key_states, value_states, attention_mask, query_length, dropout=0.0, softmax_scale=None
+    ):
+        """
+        Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
+        first unpad the input, then computes the attention scores and pad the final attention scores.
+        Args:
+            query_states (`torch.Tensor`):
+                Input query states to be passed to Flash Attention API
+            key_states (`torch.Tensor`):
+                Input key states to be passed to Flash Attention API
+            value_states (`torch.Tensor`):
+                Input value states to be passed to Flash Attention API
+            attention_mask (`torch.Tensor`):
+                The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
+                position of padding tokens and 1 for the position of non-padding tokens.
+            dropout (`int`, *optional*):
+                Attention dropout
+            softmax_scale (`float`, *optional*):
+                The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
+        """
+
+        # TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in LlamaFlashAttention2 __init__.
+        causal = self.is_causal and query_length != 1
+
+        # Contains at least one padding token in the sequence
+        if attention_mask is not None:
+            batch_size = query_states.shape[0]
+            query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
+                query_states, key_states, value_states, attention_mask, query_length
+            )
+
+            cu_seqlens_q, cu_seqlens_k = cu_seq_lens
+            max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens
+
+            attn_output_unpad = flash_attn_varlen_func(
+                query_states,
+                key_states,
+                value_states,
+                cu_seqlens_q=cu_seqlens_q,
+                cu_seqlens_k=cu_seqlens_k,
+                max_seqlen_q=max_seqlen_in_batch_q,
+                max_seqlen_k=max_seqlen_in_batch_k,
+                dropout_p=dropout,
+                softmax_scale=softmax_scale,
+                causal=causal,
+            )
+
+            attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
+        else:
+            attn_output = flash_attn_func(
+                query_states, key_states, value_states, dropout, softmax_scale=softmax_scale, causal=causal
+            )
+
+        return attn_output
+
+    def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
+        indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
+        batch_size, kv_seq_len, num_key_value_heads, head_dim = key_layer.shape
+
+        key_layer = index_first_axis(
+            key_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
+        )
+        value_layer = index_first_axis(
+            value_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
+        )
+        if query_length == kv_seq_len:
+            query_layer = index_first_axis(
+                query_layer.reshape(batch_size * kv_seq_len, self.num_heads, head_dim), indices_k
+            )
+            cu_seqlens_q = cu_seqlens_k
+            max_seqlen_in_batch_q = max_seqlen_in_batch_k
+            indices_q = indices_k
+        elif query_length == 1:
+            max_seqlen_in_batch_q = 1
+            cu_seqlens_q = torch.arange(
+                batch_size + 1, dtype=torch.int32, device=query_layer.device
+            )  # There is a memcpy here, that is very bad.
+            indices_q = cu_seqlens_q[:-1]
+            query_layer = query_layer.squeeze(1)
+        else:
+            # The -q_len: slice assumes left padding.
+            attention_mask = attention_mask[:, -query_length:]
+            query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)
+
+        return (
+            query_layer,
+            key_layer,
+            value_layer,
+            indices_q,
+            (cu_seqlens_q, cu_seqlens_k),
+            (max_seqlen_in_batch_q, max_seqlen_in_batch_k),
+        )
+
+
+# Copied from transformers.models.clip.modeling_clip.CLIPMLP with CLIP->Siglip
+class SiglipMLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.activation_fn = ACT2FN[config.hidden_act]
+        self.fc1 = nn.Linear(config.hidden_size, config.intermediate_size)
+        self.fc2 = nn.Linear(config.intermediate_size, config.hidden_size)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.fc1(hidden_states)
+        hidden_states = self.activation_fn(hidden_states)
+        hidden_states = self.fc2(hidden_states)
+        return hidden_states
+
+
+# Copied from transformers.models.clip.modeling_clip.CLIPEncoderLayer with CLIP->Siglip
+class SiglipEncoderLayer(nn.Module):
+    def __init__(self, config: SiglipVisionConfig):
+        super().__init__()
+        self.embed_dim = config.hidden_size
+        self._use_flash_attention_2 = config._attn_implementation == "flash_attention_2"
+        self.self_attn = SiglipAttention(config) if not self._use_flash_attention_2 else SiglipFlashAttention2(config)
+        self.layer_norm1 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
+        self.mlp = SiglipMLP(config)
+        self.layer_norm2 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: torch.Tensor,
+        output_attentions: Optional[bool] = False,
+    ) -> Tuple[torch.FloatTensor]:
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`):
+                Input to the layer of shape `(batch, seq_len, embed_dim)`.
+            attention_mask (`torch.FloatTensor`):
+                Attention mask of shape `(batch, 1, q_len, k_v_seq_len)` where padding elements are indicated by very large negative values.
+            output_attentions (`bool`, *optional*, defaults to `False`):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+        """
+        residual = hidden_states
+
+        hidden_states = self.layer_norm1(hidden_states)
+        hidden_states, attn_weights = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            output_attentions=output_attentions,
+        )
+        hidden_states = residual + hidden_states
+
+        residual = hidden_states
+        hidden_states = self.layer_norm2(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        outputs = (hidden_states,)
+
+        if output_attentions:
+            outputs += (attn_weights,)
+
+        return outputs
+
+
+class SiglipPreTrainedModel(PreTrainedModel):
+    """
+    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
+    models.
+    """
+
+    config_class = SiglipVisionConfig
+    base_model_prefix = "siglip"
+    supports_gradient_checkpointing = True
+
+    def _init_weights(self, module):
+        """Initialize the weights"""
+
+        if isinstance(module, SiglipVisionEmbeddings):
+            width = self.config.hidden_size
+            nn.init.normal_(module.position_embedding.weight, std=1 / np.sqrt(width))
+        elif isinstance(module, nn.Embedding):
+            default_flax_embed_init(module.weight)
+        elif isinstance(module, SiglipAttention):
+            nn.init.normal_(module.q_proj.weight)
+            nn.init.normal_(module.k_proj.weight)
+            nn.init.normal_(module.v_proj.weight)
+            nn.init.normal_(module.out_proj.weight)
+            nn.init.zeros_(module.q_proj.bias)
+            nn.init.zeros_(module.k_proj.bias)
+            nn.init.zeros_(module.v_proj.bias)
+            nn.init.zeros_(module.out_proj.bias)
+        elif isinstance(module, SiglipMLP):
+            nn.init.normal_(module.fc1.weight)
+            nn.init.normal_(module.fc2.weight)
+            nn.init.normal_(module.fc1.bias, std=1e-6)
+            nn.init.normal_(module.fc2.bias, std=1e-6)
+        elif isinstance(module, (nn.Linear, nn.Conv2d)):
+            lecun_normal_(module.weight)
+            if module.bias is not None:
+                nn.init.zeros_(module.bias)
+        elif isinstance(module, nn.LayerNorm):
+            module.bias.data.zero_()
+            module.weight.data.fill_(1.0)
+
+
+SIGLIP_START_DOCSTRING = r"""
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+    Parameters:
+        config ([`SiglipVisionConfig`]): Model configuration class with all the parameters of the model.
+            Initializing with a config file does not load the weights associated with the model, only the
+            configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+
+SIGLIP_VISION_INPUTS_DOCSTRING = r"""
+    Args:
+        pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):
+            Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using
+            [`AutoImageProcessor`]. See [`CLIPImageProcessor.__call__`] for details.
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+"""
+
+
+# Copied from transformers.models.clip.modeling_clip.CLIPEncoder with CLIP->Siglip
+class SiglipEncoder(nn.Module):
+    """
+    Transformer encoder consisting of `config.num_hidden_layers` self attention layers. Each layer is a
+    [`SiglipEncoderLayer`].
+    Args:
+        config: SiglipConfig
+    """
+
+    def __init__(self, config: SiglipVisionConfig):
+        super().__init__()
+        self.config = config
+        self.layers = nn.ModuleList([SiglipEncoderLayer(config) for _ in range(config.num_hidden_layers)])
+        self.gradient_checkpointing = False
+
+    # Ignore copy
+    def forward(
+        self,
+        inputs_embeds,
+        attention_mask: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, BaseModelOutput]:
+        r"""
+        Args:
+            inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
+                Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation.
+                This is useful if you want more control over how to convert `input_ids` indices into associated vectors
+                than the model's internal embedding lookup matrix.
+            attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+                - 1 for tokens that are **not masked**,
+                - 0 for tokens that are **masked**.
+                [What are attention masks?](../glossary#attention-mask)
+            output_attentions (`bool`, *optional*):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            output_hidden_states (`bool`, *optional*):
+                Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors
+                for more detail.
+            return_dict (`bool`, *optional*):
+                Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+        """
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        encoder_states = () if output_hidden_states else None
+        all_attentions = () if output_attentions else None
+
+        hidden_states = inputs_embeds
+        for encoder_layer in self.layers:
+            if output_hidden_states:
+                encoder_states = encoder_states + (hidden_states,)
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    encoder_layer.__call__,
+                    hidden_states,
+                    attention_mask,
+                    output_attentions,
+                )
+            else:
+                layer_outputs = encoder_layer(
+                    hidden_states,
+                    attention_mask,
+                    output_attentions=output_attentions,
+                )
+
+            hidden_states = layer_outputs[0]
+
+            if output_attentions:
+                all_attentions = all_attentions + (layer_outputs[1],)
+
+        if output_hidden_states:
+            encoder_states = encoder_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(v for v in [hidden_states, encoder_states, all_attentions] if v is not None)
+        return BaseModelOutput(last_hidden_state=hidden_states, hidden_states=encoder_states, attentions=all_attentions)
+
+
+@add_start_docstrings("""The vision model from SigLIP without any head or projection on top.""", SIGLIP_START_DOCSTRING)
+class SiglipVisionTransformer(SiglipPreTrainedModel):
+    config_class = SiglipVisionConfig
+    main_input_name = "pixel_values"
+    _supports_flash_attn_2 = True
+    _no_split_modules = []
+
+    def __init__(self, config: SiglipVisionConfig):
+        super().__init__(config)
+        self.config = config
+        embed_dim = config.hidden_size
+
+        self.embeddings = SiglipVisionEmbeddings(config)
+        self.encoder = SiglipEncoder(config)
+        self.post_layernorm = nn.LayerNorm(embed_dim, eps=config.layer_norm_eps)
+        self._use_flash_attention_2 = config._attn_implementation == "flash_attention_2"
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self) -> nn.Module:
+        return self.embeddings.patch_embedding
+
+    @add_start_docstrings_to_model_forward(SIGLIP_VISION_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=BaseModelOutputWithPooling, config_class=SiglipVisionConfig)
+    def forward(
+        self,
+        pixel_values,
+        patch_attention_mask: Optional[torch.BoolTensor] = None,
+        tgt_sizes: Optional[torch.IntTensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, BaseModelOutputWithPooling]:
+        r"""
+        Returns:
+        """
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        batch_size = pixel_values.size(0)
+        if patch_attention_mask is None:
+            patch_attention_mask = torch.ones(
+                size=(
+                    batch_size,
+                    pixel_values.size(2) // self.config.patch_size,
+                    pixel_values.size(3) // self.config.patch_size,
+                ),
+                dtype=torch.bool,
+                device=pixel_values.device,
+            )
+
+        hidden_states = self.embeddings(
+            pixel_values=pixel_values, patch_attention_mask=patch_attention_mask, tgt_sizes=tgt_sizes
+        )
+
+        patch_attention_mask = patch_attention_mask.view(batch_size, -1)
+        # The call to `_upad_input` in `_flash_attention_forward` is expensive
+        # So when the `patch_attention_mask` is full of 1s (i.e. attending to the whole sequence),
+        # avoiding passing the attention_mask, which is equivalent to attending to the full sequence
+        if not torch.any(~patch_attention_mask):
+            attention_mask = None
+        else:
+            attention_mask = (
+                _prepare_4d_attention_mask(patch_attention_mask, hidden_states.dtype)
+                if not self._use_flash_attention_2
+                else patch_attention_mask
+            )
+
+        encoder_outputs = self.encoder(
+            inputs_embeds=hidden_states,
+            attention_mask=attention_mask,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        last_hidden_state = encoder_outputs[0]
+        last_hidden_state = self.post_layernorm(last_hidden_state)
+
+        if not return_dict:
+            return (last_hidden_state, None) + encoder_outputs[1:]
+
+        return BaseModelOutputWithPooling(
+            last_hidden_state=last_hidden_state,
+            pooler_output=None,
+            hidden_states=encoder_outputs.hidden_states,
+            attentions=encoder_outputs.attentions,
+        )

preprocessor_config.json

@@ -0,0 +1,24 @@
+{
+    "image_processor_type": "MiniCPMVImageProcessor",
+    "auto_map": {
+        "AutoProcessor": "processing_minicpmo.MiniCPMOProcessor",
+        "AutoImageProcessor": "image_processing_minicpmv.MiniCPMVImageProcessor"
+      },
+    "processor_class": "MiniCPMOProcessor",
+    "max_slice_nums": 9,
+    "scale_resolution": 448,
+    "patch_size": 14,
+    "use_image_id": true,
+    "image_feature_size": 64,
+    "im_start": "<image>",
+    "im_end": "</image>",
+    "slice_start": "<slice>",
+    "slice_end": "</slice>",
+    "unk": "<unk>",
+    "im_id_start": "<image_id>",
+    "im_id_end": "</image_id>",
+    "slice_mode": true,
+    "norm_mean": [0.5, 0.5, 0.5],
+    "norm_std": [0.5, 0.5, 0.5],
+    "version": 2.6
+}

processing_minicpmo.py

@@ -0,0 +1,505 @@
+# coding=utf-8
+# Copyright 2025 The OpenBMB Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Processor class for MiniCPMO.
+"""
+
+import math
+import re
+from typing import List
+from typing import Literal
+from typing import Optional
+from typing import Union
+
+import numpy as np
+import torch
+import torchaudio
+from transformers.image_utils import ImageInput
+from transformers.processing_utils import ProcessorMixin
+from transformers.tokenization_utils_base import PreTokenizedInput
+from transformers.tokenization_utils_base import TextInput
+from transformers.utils import TensorType
+
+from .image_processing_minicpmv import MiniCPMOBatchFeature
+
+
+class MiniCPMOProcessor(ProcessorMixin):
+    r"""
+    Constructs a MiniCPMV processor which wraps a MiniCPMV image processor and a MiniCPMV tokenizer into a single processor.
+
+    [`MiniCPMVProcessor`] offers all the functionalities of [`MiniCPMVImageProcessor`] and [`LlamaTokenizerWrapper`]. See the
+    [`~MiniCPMVProcessor.__call__`] and [`~MiniCPMVProcessor.decode`] for more information.
+
+    Args:
+        image_processor ([`MiniCPMVImageProcessor`], *optional*):
+            The image processor is a required input.
+        tokenizer ([`LlamaTokenizerWrapper`], *optional*):
+            The tokenizer is a required input.
+    """
+
+    attributes = ["image_processor", "feature_extractor", "tokenizer"]
+    feature_extractor_class = "WhisperFeatureExtractor"
+    image_processor_class = "AutoImageProcessor"
+    tokenizer_class = "AutoTokenizer"
+
+    def __init__(self, image_processor=None, feature_extractor=None, tokenizer=None):
+        super().__init__(image_processor, feature_extractor, tokenizer)
+        self.version = image_processor.version
+
+    def __call__(
+        self,
+        text: Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]],
+        images: ImageInput = None,
+        audios: Union[np.ndarray, List[np.ndarray], List[List[np.ndarray]]] = None,
+        audio_parts: Optional[list] = None,
+        max_length: Optional[int] = None,
+        do_pad: Optional[bool] = True,
+        max_slice_nums: int = None,
+        use_image_id: bool = True,
+        chunk_input: bool = False,
+        return_tensors: Optional[Union[str, TensorType]] = TensorType.PYTORCH,
+        sampling_rate: Optional[int] = 16000,
+        **kwargs,
+    ) -> MiniCPMOBatchFeature:
+        if images is not None:
+            image_inputs = self.image_processor(
+                images, do_pad=do_pad, max_slice_nums=max_slice_nums, return_tensors=return_tensors
+            )
+        else:
+            image_inputs = None
+
+        if audios is not None:
+            audio_features, audio_feature_lens, audio_phs = self.audio_feature_extract(
+                audios, audio_parts, chunk_input, sampling_rate
+            )
+        else:
+            audio_features, audio_feature_lens, audio_phs = [], [], []
+
+        model_inputs = self._convert_omni_to_inputs(
+            image_inputs,
+            audio_phs,
+            text,
+            max_slice_nums=max_slice_nums,
+            use_image_id=use_image_id,
+            max_length=max_length,
+            **kwargs,
+        )
+
+        model_inputs["audio_features"] = audio_features
+        model_inputs["audio_feature_lens"] = audio_feature_lens
+
+        return MiniCPMOBatchFeature(data={**model_inputs})
+
+    def audio_feature_extract(
+        self,
+        audios: Union[np.ndarray, List[np.ndarray], List[List[np.ndarray]]],
+        audio_parts: Optional[list] = None,
+        chunk_input: Optional[bool] = False,
+        sampling_rate: Optional[int] = None,
+        chunk_length: Optional[int] = 1,
+        **kwargs,
+    ):
+        def get_audio_placeholder(audio_lens, chunk_input):
+            pool_step = 2
+            feature_lens = math.ceil(audio_lens / self.feature_extractor.hop_length)
+
+            feature_lens = (feature_lens - 1) // 2 + 1
+            output_lens = (feature_lens - pool_step) // pool_step + 1
+
+            if chunk_input:
+                fbank_feat_in_chunk = int(chunk_length * 100)
+                cnn_feat_in_chunk = (fbank_feat_in_chunk - 1) // 2 + 1
+                audio_embeds_in_chunk = (cnn_feat_in_chunk - pool_step) // pool_step + 1
+                num_audio_chunks = (output_lens + audio_embeds_in_chunk - 1) // audio_embeds_in_chunk
+
+                place_holders = ""
+                total_unk_len = 0
+                for _ in range(num_audio_chunks):
+                    unk_len = min(audio_embeds_in_chunk, output_lens - total_unk_len)
+                    place_holders += self.tokenizer.audio_start + "<unk>" * unk_len + self.tokenizer.audio_end
+                    total_unk_len += unk_len
+                audio_placeholder = place_holders
+            else:
+                audio_placeholder = self.tokenizer.audio_start + "<unk>" * output_lens + self.tokenizer.audio_end
+
+            return audio_placeholder
+
+        if isinstance(audios, np.ndarray):
+            audios_list = [[audios]]
+        elif isinstance(audios[0], np.ndarray):
+            audios_list = [audios]
+        else:
+            audios_list = audios
+
+        if audio_parts is not None:
+            assert len(audio_parts) == len(audios_list)
+            for parts, audios in zip(audio_parts, audios_list):
+                assert len(parts) == len(audios)
+
+        audio_feature_lens_list = []
+        audio_ph_list = []
+
+        audio_features_all = []
+
+        # audio placeholder not dependent on audio_parts
+        for audios in audios_list:
+            if audios:
+                audio_ph_list.append([get_audio_placeholder(len(a), chunk_input) for a in audios])
+            else:
+                audio_ph_list.append([])
+
+        for idx, audios in enumerate(audios_list):
+            if audio_parts is not None:
+                # same audio part merge
+                audio_part = audio_parts[idx]
+                merge_audio = []
+                cur_audio = []
+                for aid, (part, audio) in enumerate(zip(audio_part, audios)):
+                    if aid == 0 or audio_part[aid] == audio_part[aid - 1]:
+                        cur_audio.append(audio)
+                    else:
+                        merge_audio.append(np.hstack(cur_audio))
+                        cur_audio = [audio]
+                if cur_audio:
+                    merge_audio.append(np.hstack(cur_audio))
+
+            else:
+                merge_audio = audios
+
+            audio_feature_lens = []
+
+            # If the audio exceeds 30 seconds, split it into chunks every 30 seconds.
+            final_merge_audio = []
+            max_audio_inp_len = 30 * sampling_rate
+            for audio in merge_audio:
+                if len(audio) <= max_audio_inp_len:
+                    final_merge_audio.append(audio)
+                else:
+                    for i in range(math.ceil(len(audio) / max_audio_inp_len)):
+                        final_merge_audio.append(audio[i * max_audio_inp_len : (i + 1) * max_audio_inp_len])
+
+            if audios:
+                audio_inputs = self.feature_extractor(
+                    final_merge_audio,
+                    sampling_rate=sampling_rate,
+                    return_attention_mask=True,
+                    padding="max_length",
+                    return_tensors="pt",
+                    **kwargs,
+                )
+                audio_feature = audio_inputs["input_features"]
+                actual_lens = audio_inputs["attention_mask"].sum(dim=1)
+
+                for feat, lens in zip(audio_feature, actual_lens):
+                    audio_features_all.append(feat[:, :lens])
+                    audio_feature_lens.append(lens)
+
+                audio_feature_lens = torch.hstack(audio_feature_lens)
+                audio_feature_lens_list.append(audio_feature_lens)
+            else:
+                audio_feature_lens_list.append([])
+
+        if audio_features_all:
+            audio_features = [i.permute(1, 0) for i in audio_features_all]
+            audio_features = torch.nn.utils.rnn.pad_sequence(
+                audio_features, batch_first=True, padding_value=0.0
+            ).permute(0, 2, 1)
+        else:
+            audio_features = []
+
+        return audio_features, audio_feature_lens_list, audio_ph_list
+
+    # Copied from transformers.models.clip.processing_clip.CLIPProcessor.batch_decode with CLIP->Llama
+    def batch_decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to LlamaTokenizerFast's [`~PreTrainedTokenizer.batch_decode`]. Please
+        refer to the docstring of this method for more information.
+        """
+        output_ids = args[0]
+        result_text = []
+        for result in output_ids:
+            result = result[result != 0]
+            if result[0] == self.tokenizer.bos_id:
+                result = result[1:]
+            if result[-1] == self.tokenizer.eos_id:
+                result = result[:-1]
+            result_text.append(self.tokenizer.decode(result, *args[1:], **kwargs).strip())
+        return result_text
+        # return self.tokenizer.batch_decode(*args, **kwargs)
+
+    # Copied from transformers.models.clip.processing_clip.CLIPProcessor.decode with CLIP->Llama
+    def decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to LlamaTokenizerFast's [`~PreTrainedTokenizer.decode`]. Please refer to
+        the docstring of this method for more information.
+        """
+        result = args[0]
+        result = result[result != 0]
+        if result[0] == self.tokenizer.bos_id:
+            result = result[1:]
+        if result[-1] == self.tokenizer.eos_id or (
+            hasattr(self.tokenizer, "eot_id") and result[-1] == self.tokenizer.eot_id
+        ):
+            result = result[:-1]
+        return self.tokenizer.decode(result, *args[1:], **kwargs).strip()
+
+    def _convert(self, input_str, max_inp_length: Optional[int] = None, **kwargs):
+        input_ids = self.tokenizer.encode(input_str, **kwargs)
+        if max_inp_length is not None:
+            input_ids = input_ids[:max_inp_length]
+        input_ids = torch.tensor(input_ids, dtype=torch.int32)
+
+        ## image bound
+        start_cond = (input_ids == self.tokenizer.im_start_id) | (input_ids == self.tokenizer.slice_start_id)
+        end_cond = (input_ids == self.tokenizer.im_end_id) | (input_ids == self.tokenizer.slice_end_id)
+
+        image_start_idx = torch.where(start_cond)[0]
+        image_start_idx += 1
+        image_end_idx = torch.where(end_cond)[0]
+
+        valid_image_nums = max(len(image_start_idx), len(image_end_idx))
+
+        image_bounds = torch.hstack(
+            [
+                image_start_idx[:valid_image_nums].unsqueeze(-1),
+                image_end_idx[:valid_image_nums].unsqueeze(-1),
+            ]
+        )
+
+        ##  audio bound
+        audio_start_idx = torch.where(input_ids == self.tokenizer.audio_start_id)[0]
+        audio_end_idx = torch.where(input_ids == self.tokenizer.audio_end_id)[0]
+        assert len(audio_start_idx) == len(audio_end_idx)
+        audio_bounds = torch.hstack([(audio_start_idx + 1).unsqueeze(-1), audio_end_idx.unsqueeze(-1)])
+
+        spk_start_idx = torch.where(input_ids == self.tokenizer.spk_start_id)[0]
+        spk_end_idx = torch.where(input_ids == self.tokenizer.spk_end_id)[0]
+        assert len(spk_start_idx) == len(spk_end_idx)
+        spk_bounds = torch.hstack([(spk_start_idx + 1).unsqueeze(-1), spk_end_idx.unsqueeze(-1)])
+
+        return input_ids, image_bounds, audio_bounds, spk_bounds
+
+    def _convert_omni_to_inputs(
+        self,
+        images,
+        audio_phs,
+        texts: Union[str, List[str]],
+        truncation=None,
+        max_length=None,
+        max_slice_nums=None,
+        use_image_id=None,
+        return_tensors=None,
+        **kwargs,
+    ):
+        if images is None and audio_phs is None:
+            model_inputs = self.tokenizer(
+                texts, return_tensors=return_tensors, truncation=truncation, max_length=max_length, **kwargs
+            )
+            return MiniCPMOBatchFeature(data={**model_inputs})
+
+        image_tag = "(<image>./</image>)"
+        image_pattern = "\(<image>./</image>\)"
+        audio_tag = "(<audio>./</audio>)"
+        audio_pattern = "\(<audio>./</audio>\)"
+        split_pattern = f"({image_pattern}|{audio_pattern})"
+
+        if isinstance(texts, str):
+            texts = [texts]
+
+        bs = len(texts)
+        if images is not None:
+            images, image_sizes, tgt_sizes = images["pixel_values"], images["image_sizes"], images["tgt_sizes"]
+        else:
+            images, image_sizes, tgt_sizes = [[]] * bs, [[]] * bs, [[]] * bs
+
+        input_ids_list = []
+        image_bounds_list = []
+        audio_bounds_list = []
+        spk_bounds_list = []
+
+        for index, text in enumerate(texts):
+            text_chunks = re.split(split_pattern, text)
+
+            image_tags = re.findall(image_pattern, text)
+            audio_tags = re.findall(audio_pattern, text)
+
+            if image_tags:
+                assert images is not None
+                assert len(image_tags) == len(image_sizes[index])
+            if audio_tags:
+                assert audio_phs is not None
+                assert len(audio_tags) == len(audio_phs[index])
+
+            image_id = 0
+            audio_id = 0
+            for i, chunk in enumerate(text_chunks):
+                if chunk == image_tag:
+                    image_placeholder = self.image_processor.get_slice_image_placeholder(
+                        image_sizes[index][image_id], image_id, max_slice_nums, use_image_id
+                    )
+                    image_id += 1
+                    text_chunks[i] = image_placeholder
+                elif chunk == audio_tag:
+                    audio_placeholder = audio_phs[index][audio_id]
+                    audio_id += 1
+                    text_chunks[i] = audio_placeholder
+
+            final_text = "".join(text_chunks)
+            input_ids, image_bounds, audio_bounds, spk_bounds = self._convert(final_text, max_length, **kwargs)
+
+            input_ids_list.append(input_ids)
+            image_bounds_list.append(image_bounds)
+            audio_bounds_list.append(audio_bounds)
+            spk_bounds_list.append(spk_bounds)
+
+        padded_input_ids, padding_lengths = self.pad(input_ids_list, padding_side="left")
+        attention_mask = torch.ones_like(padded_input_ids, dtype=torch.bool)
+        for i, length in enumerate(padding_lengths):
+            image_bounds_list[i] = image_bounds_list[i] + length
+            audio_bounds_list[i] = audio_bounds_list[i] + length
+            spk_bounds_list[i] = spk_bounds_list[i] + length
+            attention_mask[i, :length] = False
+
+        data = {
+            "input_ids": padded_input_ids,
+            "attention_mask": attention_mask,
+            "pixel_values": images,
+            "image_sizes": image_sizes,
+            "image_bound": image_bounds_list,
+            "tgt_sizes": tgt_sizes,
+            "audio_bounds": audio_bounds_list,
+            "spk_bounds": spk_bounds_list,
+        }
+
+        return data
+
+    @property
+    # Copied from transformers.models.clip.processing_clip.CLIPProcessor.model_input_names
+    def model_input_names(self):
+        tokenizer_input_names = self.tokenizer.model_input_names
+        image_processor_input_names = self.image_processor.model_input_names
+        feature_extractor_input_names = self.feature_extractor.model_input_names
+        return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names + feature_extractor_input_names))
+
+    def pad(self, inputs, max_length=None, padding_value=0, padding_side="left"):
+        items = []
+        if isinstance(inputs[0], list):
+            assert isinstance(inputs[0][0], torch.Tensor)
+            for it in inputs:
+                for tr in it:
+                    items.append(tr)
+        else:
+            assert isinstance(inputs[0], torch.Tensor)
+            items = inputs
+
+        batch_size = len(items)
+        shape = items[0].shape
+        dim = len(shape)
+        assert dim <= 2
+        if max_length is None:
+            max_length = 0
+        max_length = max(max_length, max(item.shape[-1] for item in items))
+        min_length = min(item.shape[-1] for item in items)
+        dtype = items[0].dtype
+
+        if dim == 0:
+            return torch.stack([item for item in items], dim=0), [0]
+        elif dim == 1:
+            if max_length == min_length:
+                return torch.stack([item for item in items], dim=0), [0] * batch_size
+            tensor = torch.zeros((batch_size, max_length), dtype=dtype) + padding_value
+        else:
+            tensor = torch.zeros((batch_size, max_length, shape[-1]), dtype=dtype) + padding_value
+
+        padding_length = []
+        for i, item in enumerate(items):
+            if dim == 1:
+                if padding_side == "left":
+                    tensor[i, -len(item) :] = item.clone()
+                else:
+                    tensor[i, : len(item)] = item.clone()
+            elif dim == 2:
+                if padding_side == "left":
+                    tensor[i, -len(item) :, :] = item.clone()
+                else:
+                    tensor[i, : len(item), :] = item.clone()
+            padding_length.append(tensor.shape[-1] - len(item))
+
+        return tensor, padding_length
+
+
+class MelSpectrogramFeatures(torch.nn.Module):
+    def __init__(
+        self,
+        sample_rate=24000,
+        n_fft=1024,
+        hop_length=256,
+        n_mels=100,
+        padding: Literal["center", "same"] = "center",
+    ):
+        super().__init__()
+        if padding not in ["center", "same"]:
+            raise ValueError("Padding must be 'center' or 'same'.")
+        self.padding = padding
+        self.mel_spec = torchaudio.transforms.MelSpectrogram(
+            sample_rate=sample_rate,
+            n_fft=n_fft,
+            hop_length=hop_length,
+            n_mels=n_mels,
+            center=padding == "center",
+            power=1,
+        )
+
+    def __call__(self, audio: torch.Tensor) -> torch.Tensor:
+        """
+        audio: Tensor([num_channels, num_samples])
+        """
+        return super().__call__(audio)
+
+    def forward(self, audio: torch.Tensor) -> torch.Tensor:
+        """
+        audio: Tensor([num_channels, num_samples])
+        """
+        mel: torch.Tensor = self.mel_spec(audio)
+        features = torch.log(torch.clip(mel, min=1e-5))
+        return features
+
+
+class ChatTTSProcessor:
+    def __init__(self, text_tokenizer):
+        self.audio_processor = MelSpectrogramFeatures()
+        self.text_tokenizer = text_tokenizer
+
+    def __call__(self, text_list, audio_list):
+        assert len(text_list) == len(audio_list)
+        input_ids_varlen = []
+        for text in text_list:
+            input_ids_ = self.text_tokenizer.encode(text, return_tensors="pt", add_special_tokens=False)  # [1, seq_len]
+            input_ids_ = input_ids_.squeeze(0)  # [seq_len]
+            input_ids_varlen.append(input_ids_)
+
+        audio_features_varlen = []
+        for audio in audio_list:
+            assert audio.shape.__len__() == 1  # [seq_len]
+            try:
+                mel = self.audio_processor(audio)  # [100(num_mel_bins), seq_len_mel]
+            except Exception as e:
+                raise e
+            audio_features_varlen.append(mel)
+
+        return {
+            "tts_input_ids_varlen": input_ids_varlen,  # return List[Tensor]
+            "tts_input_features_varlen": audio_features_varlen,  # return List[Tensor]
+        }

quantize_config.json

@@ -0,0 +1,14 @@
+{
+  "bits": 4,
+  "group_size": -1,
+  "damp_percent": 0.01,
+  "desc_act": true,
+  "static_groups": false,
+  "sym": true,
+  "true_sequential": true,
+  "lm_head": false,
+  "model_name_or_path": null,
+  "model_file_base_name": null,
+  "quant_method": "gptq",
+  "checkpoint_format": "gptq"
+}

resampler.py

@@ -0,0 +1,864 @@
+# coding=utf-8
+# Copyright 2025 The OpenBMB Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import warnings
+from functools import partial
+from typing import Optional
+from typing import Tuple
+
+import numpy as np
+import torch
+import torch.nn.functional as F
+from torch import nn
+from torch import Tensor
+from torch.nn.functional import *
+from torch.nn.init import trunc_normal_
+from torch.nn.modules.activation import *
+from transformers.integrations import is_deepspeed_zero3_enabled
+
+
+def get_2d_sincos_pos_embed(embed_dim, image_size):
+    """
+    image_size: image_size or (image_height, image_width)
+    return:
+    pos_embed: [image_height, image_width, embed_dim]
+    """
+    if isinstance(image_size, int):
+        grid_h_size, grid_w_size = image_size, image_size
+    else:
+        grid_h_size, grid_w_size = image_size[0], image_size[1]
+
+    grid_h = np.arange(grid_h_size, dtype=np.float32)
+    grid_w = np.arange(grid_w_size, dtype=np.float32)
+    grid = np.meshgrid(grid_w, grid_h)  # here w goes first
+    grid = np.stack(grid, axis=0)
+
+    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)
+    return pos_embed
+
+
+def get_2d_sincos_pos_embed_from_grid(embed_dim, grid):
+    assert embed_dim % 2 == 0
+
+    # use half of dimensions to encode grid_h
+    emb_h = get_1d_sincos_pos_embed_from_grid_new(embed_dim // 2, grid[0])  # (H, W, D/2)
+    emb_w = get_1d_sincos_pos_embed_from_grid_new(embed_dim // 2, grid[1])  # (H, W, D/2)
+
+    emb = np.concatenate([emb_h, emb_w], axis=-1)  # (H, W, D)
+    return emb
+
+
+def get_1d_sincos_pos_embed_from_grid_new(embed_dim, pos):
+    """
+    embed_dim: output dimension for each position
+    pos: a list of positions to be encoded: size (H, W)
+    out: (H, W, D)
+    """
+    assert embed_dim % 2 == 0
+    omega = np.arange(embed_dim // 2, dtype=np.float32)
+    omega /= embed_dim / 2.0
+    omega = 1.0 / 10000**omega  # (D/2,)
+
+    out = np.einsum("hw,d->hwd", pos, omega)  # (H, W, D/2), outer product
+
+    emb_sin = np.sin(out)  # (H, W, D/2)
+    emb_cos = np.cos(out)  # (H, W, D/2)
+
+    emb = np.concatenate([emb_sin, emb_cos], axis=-1)  # (H, W, D)
+    return emb
+
+
+class Resampler(nn.Module):
+    """
+    A 2D perceiver-resampler network with one cross attention layers by
+       given learnable queries and 2d sincos pos_emb
+    Outputs:
+        A tensor with the shape of (batch_size, num_queries, embed_dim)
+    """
+
+    def __init__(
+        self,
+        num_queries,
+        embed_dim,
+        num_heads,
+        kv_dim=None,
+        norm_layer=partial(nn.LayerNorm, eps=1e-6),
+        adaptive=False,
+        max_size=(70, 70),
+    ):
+        super().__init__()
+        self.num_queries = num_queries
+        self.embed_dim = embed_dim
+        self.num_heads = num_heads
+        self.adaptive = adaptive
+        self.max_size = max_size
+
+        self.query = nn.Parameter(torch.zeros(self.num_queries, embed_dim))
+
+        if kv_dim is not None and kv_dim != embed_dim:
+            self.kv_proj = nn.Linear(kv_dim, embed_dim, bias=False)
+        else:
+            self.kv_proj = nn.Identity()
+
+        self.attn = MultiheadAttention(embed_dim, num_heads)
+        self.ln_q = norm_layer(embed_dim)
+        self.ln_kv = norm_layer(embed_dim)
+
+        self.ln_post = norm_layer(embed_dim)
+        self.proj = nn.Parameter((embed_dim**-0.5) * torch.randn(embed_dim, embed_dim))
+
+        self._set_2d_pos_cache(self.max_size)
+
+    def _set_2d_pos_cache(self, max_size, device="cpu"):
+        if is_deepspeed_zero3_enabled():
+            device = "cuda"
+        pos_embed = torch.from_numpy(get_2d_sincos_pos_embed(self.embed_dim, max_size)).float().to(device)
+        self.register_buffer("pos_embed", pos_embed, persistent=False)
+
+    def _adjust_pos_cache(self, tgt_sizes, device):
+        max_h = torch.max(tgt_sizes[:, 0])
+        max_w = torch.max(tgt_sizes[:, 1])
+        if max_h > self.max_size[0] or max_w > self.max_size[1]:
+            self.max_size = [max(max_h, self.max_size[0]), max(max_w, self.max_size[1])]
+            self._set_2d_pos_cache(self.max_size, device)
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            trunc_normal_(m.weight, std=0.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+
+    def forward(self, x, tgt_sizes=None):
+        assert x.shape[0] == tgt_sizes.shape[0]
+        bs = x.shape[0]
+
+        device = x.device
+        dtype = x.dtype
+
+        patch_len = tgt_sizes[:, 0] * tgt_sizes[:, 1]
+
+        self._adjust_pos_cache(tgt_sizes, device=device)
+
+        max_patch_len = torch.max(patch_len)
+        key_padding_mask = torch.zeros((bs, max_patch_len), dtype=torch.bool, device=device)
+
+        pos_embed = []
+        for i in range(bs):
+            tgt_h, tgt_w = tgt_sizes[i]
+            pos_embed.append(self.pos_embed[:tgt_h, :tgt_w, :].reshape((tgt_h * tgt_w, -1)).to(dtype))  # patches * D
+            key_padding_mask[i, patch_len[i] :] = True
+
+        pos_embed = torch.nn.utils.rnn.pad_sequence(pos_embed, batch_first=True, padding_value=0.0).permute(
+            1, 0, 2
+        )  # BLD => L * B * D
+
+        x = self.kv_proj(x)  # B * L * D
+        x = self.ln_kv(x).permute(1, 0, 2)  # L * B * D
+
+        q = self.ln_q(self.query)  # Q * D
+
+        out = self.attn(
+            self._repeat(q, bs),  # Q * B * D
+            x + pos_embed,  # L * B * D +  L * B * D
+            x,
+            key_padding_mask=key_padding_mask,
+        )[0]
+        #  out: Q * B * D
+        x = out.permute(1, 0, 2)  # B * Q * D
+
+        x = self.ln_post(x)
+        x = x @ self.proj
+        return x
+
+    def _repeat(self, query, N: int):
+        return query.unsqueeze(1).repeat(1, N, 1)
+
+
+class MultiheadAttention(nn.MultiheadAttention):
+    def __init__(
+        self,
+        embed_dim,
+        num_heads,
+        dropout=0.0,
+        bias=True,
+        add_bias_kv=False,
+        add_zero_attn=False,
+        kdim=None,
+        vdim=None,
+        batch_first=False,
+        device=None,
+        dtype=None,
+    ):
+        super().__init__(
+            embed_dim, num_heads, dropout, bias, add_bias_kv, add_zero_attn, kdim, vdim, batch_first, device, dtype
+        )
+
+        # rewrite out_proj layer，with nn.Linear
+        self.out_proj = nn.Linear(embed_dim, embed_dim, bias=bias, device=device, dtype=dtype)
+
+    def forward(
+        self,
+        query: Tensor,
+        key: Tensor,
+        value: Tensor,
+        key_padding_mask: Optional[Tensor] = None,
+        need_weights: bool = True,
+        attn_mask: Optional[Tensor] = None,
+        average_attn_weights: bool = True,
+        is_causal: bool = False,
+    ) -> Tuple[Tensor, Optional[Tensor]]:
+        why_not_fast_path = ""
+        if (
+            (attn_mask is not None and torch.is_floating_point(attn_mask))
+            or (key_padding_mask is not None)
+            and torch.is_floating_point(key_padding_mask)
+        ):
+            why_not_fast_path = "floating-point masks are not supported for fast path."
+
+        is_batched = query.dim() == 3
+
+        key_padding_mask = _canonical_mask(
+            mask=key_padding_mask,
+            mask_name="key_padding_mask",
+            other_type=F._none_or_dtype(attn_mask),
+            other_name="attn_mask",
+            target_type=query.dtype,
+        )
+
+        attn_mask = _canonical_mask(
+            mask=attn_mask,
+            mask_name="attn_mask",
+            other_type=None,
+            other_name="",
+            target_type=query.dtype,
+            check_other=False,
+        )
+
+        if not is_batched:
+            why_not_fast_path = f"input not batched; expected query.dim() of 3 but got {query.dim()}"
+        elif query is not key or key is not value:
+            # When lifting this restriction, don't forget to either
+            # enforce that the dtypes all match or test cases where
+            # they don't!
+            why_not_fast_path = "non-self attention was used (query, key, and value are not the same Tensor)"
+        elif self.in_proj_bias is not None and query.dtype != self.in_proj_bias.dtype:
+            why_not_fast_path = (
+                f"dtypes of query ({query.dtype}) and self.in_proj_bias ({self.in_proj_bias.dtype}) don't match"
+            )
+        elif self.in_proj_weight is None:
+            why_not_fast_path = "in_proj_weight was None"
+        elif query.dtype != self.in_proj_weight.dtype:
+            # this case will fail anyway, but at least they'll get a useful error message.
+            why_not_fast_path = (
+                f"dtypes of query ({query.dtype}) and self.in_proj_weight ({self.in_proj_weight.dtype}) don't match"
+            )
+        elif self.training:
+            why_not_fast_path = "training is enabled"
+        elif (self.num_heads % 2) != 0:
+            why_not_fast_path = "self.num_heads is not even"
+        elif not self.batch_first:
+            why_not_fast_path = "batch_first was not True"
+        elif self.bias_k is not None:
+            why_not_fast_path = "self.bias_k was not None"
+        elif self.bias_v is not None:
+            why_not_fast_path = "self.bias_v was not None"
+        elif self.add_zero_attn:
+            why_not_fast_path = "add_zero_attn was enabled"
+        elif not self._qkv_same_embed_dim:
+            why_not_fast_path = "_qkv_same_embed_dim was not True"
+        elif query.is_nested and (key_padding_mask is not None or attn_mask is not None):
+            why_not_fast_path = "supplying both src_key_padding_mask and src_mask at the same time \
+                                 is not supported with NestedTensor input"
+        elif torch.is_autocast_enabled():
+            why_not_fast_path = "autocast is enabled"
+
+        if not why_not_fast_path:
+            tensor_args = (
+                query,
+                key,
+                value,
+                self.in_proj_weight,
+                self.in_proj_bias,
+                self.out_proj.weight,
+                self.out_proj.bias,
+            )
+            # We have to use list comprehensions below because TorchScript does not support
+            # generator expressions.
+            if torch.overrides.has_torch_function(tensor_args):
+                why_not_fast_path = "some Tensor argument has_torch_function"
+            elif _is_make_fx_tracing():
+                why_not_fast_path = "we are running make_fx tracing"
+            elif not all(_check_arg_device(x) for x in tensor_args):
+                why_not_fast_path = (
+                    "some Tensor argument's device is neither one of "
+                    f"cpu, cuda or {torch.utils.backend_registration._privateuse1_backend_name}"
+                )
+            elif torch.is_grad_enabled() and any(_arg_requires_grad(x) for x in tensor_args):
+                why_not_fast_path = (
+                    "grad is enabled and at least one of query or the "
+                    "input/output projection weights or biases requires_grad"
+                )
+            if not why_not_fast_path:
+                merged_mask, mask_type = self.merge_masks(attn_mask, key_padding_mask, query)
+
+                if self.in_proj_bias is not None and self.in_proj_weight is not None:
+                    return torch._native_multi_head_attention(
+                        query,
+                        key,
+                        value,
+                        self.embed_dim,
+                        self.num_heads,
+                        self.in_proj_weight,
+                        self.in_proj_bias,
+                        self.out_proj.weight,
+                        self.out_proj.bias,
+                        merged_mask,
+                        need_weights,
+                        average_attn_weights,
+                        mask_type,
+                    )
+
+        any_nested = query.is_nested or key.is_nested or value.is_nested
+        assert not any_nested, (
+            "MultiheadAttention does not support NestedTensor outside of its fast path. "
+            + f"The fast path was not hit because {why_not_fast_path}"
+        )
+
+        if self.batch_first and is_batched:
+            # make sure that the transpose op does not affect the "is" property
+            if key is value:
+                if query is key:
+                    query = key = value = query.transpose(1, 0)
+                else:
+                    query, key = (x.transpose(1, 0) for x in (query, key))
+                    value = key
+            else:
+                query, key, value = (x.transpose(1, 0) for x in (query, key, value))
+
+        if not self._qkv_same_embed_dim:
+            attn_output, attn_output_weights = self.multi_head_attention_forward(
+                query,
+                key,
+                value,
+                self.embed_dim,
+                self.num_heads,
+                self.in_proj_weight,
+                self.in_proj_bias,
+                self.bias_k,
+                self.bias_v,
+                self.add_zero_attn,
+                self.dropout,
+                self.out_proj.weight,
+                self.out_proj.bias,
+                training=self.training,
+                key_padding_mask=key_padding_mask,
+                need_weights=need_weights,
+                attn_mask=attn_mask,
+                use_separate_proj_weight=True,
+                q_proj_weight=self.q_proj_weight,
+                k_proj_weight=self.k_proj_weight,
+                v_proj_weight=self.v_proj_weight,
+                average_attn_weights=average_attn_weights,
+                is_causal=is_causal,
+            )
+        else:
+            attn_output, attn_output_weights = self.multi_head_attention_forward(
+                query,
+                key,
+                value,
+                self.embed_dim,
+                self.num_heads,
+                self.in_proj_weight,
+                self.in_proj_bias,
+                self.bias_k,
+                self.bias_v,
+                self.add_zero_attn,
+                self.dropout,
+                self.out_proj.weight,
+                self.out_proj.bias,
+                training=self.training,
+                key_padding_mask=key_padding_mask,
+                need_weights=need_weights,
+                attn_mask=attn_mask,
+                average_attn_weights=average_attn_weights,
+                is_causal=is_causal,
+            )
+        if self.batch_first and is_batched:
+            return attn_output.transpose(1, 0), attn_output_weights
+        else:
+            return attn_output, attn_output_weights
+
+    def multi_head_attention_forward(
+        self,
+        query: Tensor,
+        key: Tensor,
+        value: Tensor,
+        embed_dim_to_check: int,
+        num_heads: int,
+        in_proj_weight: Optional[Tensor],
+        in_proj_bias: Optional[Tensor],
+        bias_k: Optional[Tensor],
+        bias_v: Optional[Tensor],
+        add_zero_attn: bool,
+        dropout_p: float,
+        out_proj_weight: Tensor,
+        out_proj_bias: Optional[Tensor],
+        training: bool = True,
+        key_padding_mask: Optional[Tensor] = None,
+        need_weights: bool = True,
+        attn_mask: Optional[Tensor] = None,
+        use_separate_proj_weight: bool = False,
+        q_proj_weight: Optional[Tensor] = None,
+        k_proj_weight: Optional[Tensor] = None,
+        v_proj_weight: Optional[Tensor] = None,
+        static_k: Optional[Tensor] = None,
+        static_v: Optional[Tensor] = None,
+        average_attn_weights: bool = True,
+        is_causal: bool = False,
+    ) -> Tuple[Tensor, Optional[Tensor]]:
+        tens_ops = (query, key, value, in_proj_weight, in_proj_bias, bias_k, bias_v, out_proj_weight, out_proj_bias)
+
+        is_batched = _mha_shape_check(query, key, value, key_padding_mask, attn_mask, num_heads)
+
+        # For unbatched input, we unsqueeze at the expected batch-dim to pretend that the input
+        # is batched, run the computation and before returning squeeze the
+        # batch dimension so that the output doesn't carry this temporary batch dimension.
+        if not is_batched:
+            # unsqueeze if the input is unbatched
+            query = query.unsqueeze(1)
+            key = key.unsqueeze(1)
+            value = value.unsqueeze(1)
+            if key_padding_mask is not None:
+                key_padding_mask = key_padding_mask.unsqueeze(0)
+
+        # set up shape vars
+        tgt_len, bsz, embed_dim = query.shape
+        src_len, _, _ = key.shape
+
+        key_padding_mask = _canonical_mask(
+            mask=key_padding_mask,
+            mask_name="key_padding_mask",
+            other_type=F._none_or_dtype(attn_mask),
+            other_name="attn_mask",
+            target_type=query.dtype,
+        )
+
+        if is_causal and attn_mask is None:
+            raise RuntimeError(
+                "Need attn_mask if specifying the is_causal hint. "
+                "You may use the Transformer module method "
+                "`generate_square_subsequent_mask` to create this mask."
+            )
+
+        if is_causal and key_padding_mask is None and not need_weights:
+            # when we have a kpm or need weights, we need attn_mask
+            # Otherwise, we use the is_causal hint go as is_causal
+            # indicator to SDPA.
+            attn_mask = None
+        else:
+            attn_mask = _canonical_mask(
+                mask=attn_mask,
+                mask_name="attn_mask",
+                other_type=None,
+                other_name="",
+                target_type=query.dtype,
+                check_other=False,
+            )
+
+            if key_padding_mask is not None:
+                # We have the attn_mask, and use that to merge kpm into it.
+                # Turn off use of is_causal hint, as the merged mask is no
+                # longer causal.
+                is_causal = False
+
+        assert (
+            embed_dim == embed_dim_to_check
+        ), f"was expecting embedding dimension of {embed_dim_to_check}, but got {embed_dim}"
+        if isinstance(embed_dim, torch.Tensor):
+            # embed_dim can be a tensor when JIT tracing
+            head_dim = embed_dim.div(num_heads, rounding_mode="trunc")
+        else:
+            head_dim = embed_dim // num_heads
+        assert head_dim * num_heads == embed_dim, f"embed_dim {embed_dim} not divisible by num_heads {num_heads}"
+        if use_separate_proj_weight:
+            # allow MHA to have different embedding dimensions when separate projection weights are used
+            assert (
+                key.shape[:2] == value.shape[:2]
+            ), f"key's sequence and batch dims {key.shape[:2]} do not match value's {value.shape[:2]}"
+        else:
+            assert key.shape == value.shape, f"key shape {key.shape} does not match value shape {value.shape}"
+
+        #
+        # compute in-projection
+        #
+        if not use_separate_proj_weight:
+            assert in_proj_weight is not None, "use_separate_proj_weight is False but in_proj_weight is None"
+            q, k, v = _in_projection_packed(query, key, value, in_proj_weight, in_proj_bias)
+        else:
+            assert q_proj_weight is not None, "use_separate_proj_weight is True but q_proj_weight is None"
+            assert k_proj_weight is not None, "use_separate_proj_weight is True but k_proj_weight is None"
+            assert v_proj_weight is not None, "use_separate_proj_weight is True but v_proj_weight is None"
+            if in_proj_bias is None:
+                b_q = b_k = b_v = None
+            else:
+                b_q, b_k, b_v = in_proj_bias.chunk(3)
+            q, k, v = _in_projection(query, key, value, q_proj_weight, k_proj_weight, v_proj_weight, b_q, b_k, b_v)
+
+        # prep attention mask
+
+        if attn_mask is not None:
+            # ensure attn_mask's dim is 3
+            if attn_mask.dim() == 2:
+                correct_2d_size = (tgt_len, src_len)
+                if attn_mask.shape != correct_2d_size:
+                    raise RuntimeError(
+                        f"The shape of the 2D attn_mask is {attn_mask.shape}, but should be {correct_2d_size}."
+                    )
+                attn_mask = attn_mask.unsqueeze(0)
+            elif attn_mask.dim() == 3:
+                correct_3d_size = (bsz * num_heads, tgt_len, src_len)
+                if attn_mask.shape != correct_3d_size:
+                    raise RuntimeError(
+                        f"The shape of the 3D attn_mask is {attn_mask.shape}, but should be {correct_3d_size}."
+                    )
+            else:
+                raise RuntimeError(f"attn_mask's dimension {attn_mask.dim()} is not supported")
+
+        # add bias along batch dimension (currently second)
+        if bias_k is not None and bias_v is not None:
+            assert static_k is None, "bias cannot be added to static key."
+            assert static_v is None, "bias cannot be added to static value."
+            k = torch.cat([k, bias_k.repeat(1, bsz, 1)])
+            v = torch.cat([v, bias_v.repeat(1, bsz, 1)])
+            if attn_mask is not None:
+                attn_mask = pad(attn_mask, (0, 1))
+            if key_padding_mask is not None:
+                key_padding_mask = pad(key_padding_mask, (0, 1))
+        else:
+            assert bias_k is None
+            assert bias_v is None
+
+        #
+        # reshape q, k, v for multihead attention and make em batch first
+        #
+        q = q.view(tgt_len, bsz * num_heads, head_dim).transpose(0, 1)
+        if static_k is None:
+            k = k.view(k.shape[0], bsz * num_heads, head_dim).transpose(0, 1)
+        else:
+            # TODO finish disentangling control flow so we don't do in-projections when statics are passed
+            assert (
+                static_k.size(0) == bsz * num_heads
+            ), f"expecting static_k.size(0) of {bsz * num_heads}, but got {static_k.size(0)}"
+            assert static_k.size(2) == head_dim, f"expecting static_k.size(2) of {head_dim}, but got {static_k.size(2)}"
+            k = static_k
+        if static_v is None:
+            v = v.view(v.shape[0], bsz * num_heads, head_dim).transpose(0, 1)
+        else:
+            # TODO finish disentangling control flow so we don't do in-projections when statics are passed
+            assert (
+                static_v.size(0) == bsz * num_heads
+            ), f"expecting static_v.size(0) of {bsz * num_heads}, but got {static_v.size(0)}"
+            assert static_v.size(2) == head_dim, f"expecting static_v.size(2) of {head_dim}, but got {static_v.size(2)}"
+            v = static_v
+
+        # add zero attention along batch dimension (now first)
+        if add_zero_attn:
+            zero_attn_shape = (bsz * num_heads, 1, head_dim)
+            k = torch.cat([k, torch.zeros(zero_attn_shape, dtype=k.dtype, device=k.device)], dim=1)
+            v = torch.cat([v, torch.zeros(zero_attn_shape, dtype=v.dtype, device=v.device)], dim=1)
+            if attn_mask is not None:
+                attn_mask = pad(attn_mask, (0, 1))
+            if key_padding_mask is not None:
+                key_padding_mask = pad(key_padding_mask, (0, 1))
+
+        # update source sequence length after adjustments
+        src_len = k.size(1)
+
+        # merge key padding and attention masks
+        if key_padding_mask is not None:
+            assert key_padding_mask.shape == (
+                bsz,
+                src_len,
+            ), f"expecting key_padding_mask shape of {(bsz, src_len)}, but got {key_padding_mask.shape}"
+            key_padding_mask = (
+                key_padding_mask.view(bsz, 1, 1, src_len)
+                .expand(-1, num_heads, -1, -1)
+                .reshape(bsz * num_heads, 1, src_len)
+            )
+            if attn_mask is None:
+                attn_mask = key_padding_mask
+            else:
+                attn_mask = attn_mask + key_padding_mask
+
+        # adjust dropout probability
+        if not training:
+            dropout_p = 0.0
+
+        #
+        # (deep breath) calculate attention and out projection
+        #
+
+        if need_weights:
+            B, Nt, E = q.shape
+            q_scaled = q / math.sqrt(E)
+
+            assert not (is_causal and attn_mask is None), "FIXME: is_causal not implemented for need_weights"
+
+            if attn_mask is not None:
+                attn_output_weights = torch.baddbmm(attn_mask, q_scaled, k.transpose(-2, -1))
+            else:
+                attn_output_weights = torch.bmm(q_scaled, k.transpose(-2, -1))
+            attn_output_weights = softmax(attn_output_weights, dim=-1)
+            if dropout_p > 0.0:
+                attn_output_weights = dropout(attn_output_weights, p=dropout_p)
+
+            attn_output = torch.bmm(attn_output_weights, v)
+
+            attn_output = attn_output.transpose(0, 1).contiguous().view(tgt_len * bsz, embed_dim)
+            attn_output = self.out_proj(attn_output)
+            attn_output = attn_output.view(tgt_len, bsz, attn_output.size(1))
+
+            # optionally average attention weights over heads
+            attn_output_weights = attn_output_weights.view(bsz, num_heads, tgt_len, src_len)
+            if average_attn_weights:
+                attn_output_weights = attn_output_weights.mean(dim=1)
+
+            if not is_batched:
+                # squeeze the output if input was unbatched
+                attn_output = attn_output.squeeze(1)
+                attn_output_weights = attn_output_weights.squeeze(0)
+            return attn_output, attn_output_weights
+        else:
+            # attn_mask can be either (L,S) or (N*num_heads, L, S)
+            # if attn_mask's shape is (1, L, S) we need to unsqueeze to (1, 1, L, S)
+            # in order to match the input for SDPA of (N, num_heads, L, S)
+            if attn_mask is not None:
+                if attn_mask.size(0) == 1 and attn_mask.dim() == 3:
+                    attn_mask = attn_mask.unsqueeze(0)
+                else:
+                    attn_mask = attn_mask.view(bsz, num_heads, -1, src_len)
+
+            q = q.view(bsz, num_heads, tgt_len, head_dim)
+            k = k.view(bsz, num_heads, src_len, head_dim)
+            v = v.view(bsz, num_heads, src_len, head_dim)
+
+            attn_output = F.scaled_dot_product_attention(q, k, v, attn_mask, dropout_p, is_causal)
+            attn_output = attn_output.permute(2, 0, 1, 3).contiguous().view(bsz * tgt_len, embed_dim)
+
+            attn_output = self.out_proj(attn_output)
+            attn_output = attn_output.view(tgt_len, bsz, attn_output.size(1))
+            if not is_batched:
+                # squeeze the output if input was unbatched
+                attn_output = attn_output.squeeze(1)
+            return attn_output, None
+
+
+def _mha_shape_check(
+    query: Tensor,
+    key: Tensor,
+    value: Tensor,
+    key_padding_mask: Optional[Tensor],
+    attn_mask: Optional[Tensor],
+    num_heads: int,
+):
+    # Verifies the expected shape for `query, `key`, `value`, `key_padding_mask` and `attn_mask`
+    # and returns if the input is batched or not.
+    # Raises an error if `query` is not 2-D (unbatched) or 3-D (batched) tensor.
+
+    # Shape check.
+    if query.dim() == 3:
+        # Batched Inputs
+        is_batched = True
+        assert key.dim() == 3 and value.dim() == 3, (
+            "For batched (3-D) `query`, expected `key` and `value` to be 3-D"
+            f" but found {key.dim()}-D and {value.dim()}-D tensors respectively"
+        )
+        if key_padding_mask is not None:
+            assert key_padding_mask.dim() == 2, (
+                "For batched (3-D) `query`, expected `key_padding_mask` to be `None` or 2-D"
+                f" but found {key_padding_mask.dim()}-D tensor instead"
+            )
+        if attn_mask is not None:
+            assert attn_mask.dim() in (2, 3), (
+                "For batched (3-D) `query`, expected `attn_mask` to be `None`, 2-D or 3-D"
+                f" but found {attn_mask.dim()}-D tensor instead"
+            )
+    elif query.dim() == 2:
+        # Unbatched Inputs
+        is_batched = False
+        assert key.dim() == 2 and value.dim() == 2, (
+            "For unbatched (2-D) `query`, expected `key` and `value` to be 2-D"
+            f" but found {key.dim()}-D and {value.dim()}-D tensors respectively"
+        )
+
+        if key_padding_mask is not None:
+            assert key_padding_mask.dim() == 1, (
+                "For unbatched (2-D) `query`, expected `key_padding_mask` to be `None` or 1-D"
+                f" but found {key_padding_mask.dim()}-D tensor instead"
+            )
+
+        if attn_mask is not None:
+            assert attn_mask.dim() in (2, 3), (
+                "For unbatched (2-D) `query`, expected `attn_mask` to be `None`, 2-D or 3-D"
+                f" but found {attn_mask.dim()}-D tensor instead"
+            )
+            if attn_mask.dim() == 3:
+                expected_shape = (num_heads, query.shape[0], key.shape[0])
+                assert (
+                    attn_mask.shape == expected_shape
+                ), f"Expected `attn_mask` shape to be {expected_shape} but got {attn_mask.shape}"
+    else:
+        raise AssertionError(
+            f"query should be unbatched 2D or batched 3D tensor but received {query.dim()}-D query tensor"
+        )
+
+    return is_batched
+
+
+def _canonical_mask(
+    mask: Optional[Tensor],
+    mask_name: str,
+    other_type: Optional[DType],
+    other_name: str,
+    target_type: DType,
+    check_other: bool = True,
+) -> Optional[Tensor]:
+
+    if mask is not None:
+        _mask_dtype = mask.dtype
+        _mask_is_float = torch.is_floating_point(mask)
+        if _mask_dtype != torch.bool and not _mask_is_float:
+            raise AssertionError(f"only bool and floating types of {mask_name} are supported")
+        if check_other and other_type is not None:
+            if _mask_dtype != other_type:
+                warnings.warn(
+                    f"Support for mismatched {mask_name} and {other_name} "
+                    "is deprecated. Use same type for both instead."
+                )
+        if not _mask_is_float:
+            mask = torch.zeros_like(mask, dtype=target_type).masked_fill_(mask, float("-inf"))
+    return mask
+
+
+def _in_projection_packed(
+    q: Tensor,
+    k: Tensor,
+    v: Tensor,
+    w: Tensor,
+    b: Optional[Tensor] = None,
+) -> List[Tensor]:
+    r"""
+    Performs the in-projection step of the attention operation, using packed weights.
+    Output is a triple containing projection tensors for query, key and value.
+    Args:
+        q, k, v: query, key and value tensors to be projected. For self-attention,
+            these are typically the same tensor; for encoder-decoder attention,
+            k and v are typically the same tensor. (We take advantage of these
+            identities for performance if they are present.) Regardless, q, k and v
+            must share a common embedding dimension; otherwise their shapes may vary.
+        w: projection weights for q, k and v, packed into a single tensor. Weights
+            are packed along dimension 0, in q, k, v order.
+        b: optional projection biases for q, k and v, packed into a single tensor
+            in q, k, v order.
+    Shape:
+        Inputs:
+        - q: :math:`(..., E)` where E is the embedding dimension
+        - k: :math:`(..., E)` where E is the embedding dimension
+        - v: :math:`(..., E)` where E is the embedding dimension
+        - w: :math:`(E * 3, E)` where E is the embedding dimension
+        - b: :math:`E * 3` where E is the embedding dimension
+        Output:
+        - in output list :math:`[q', k', v']`, each output tensor will have the
+            same shape as the corresponding input tensor.
+    """
+    E = q.size(-1)
+    if k is v:
+        if q is k:
+            # self-attention
+            proj = linear(q, w, b)
+            # reshape to 3, E and not E, 3 is deliberate for better memory coalescing and keeping same order as chunk()
+            proj = proj.unflatten(-1, (3, E)).unsqueeze(0).transpose(0, -2).squeeze(-2).contiguous()
+            return proj[0], proj[1], proj[2]
+        else:
+            # encoder-decoder attention
+            w_q, w_kv = w.split([E, E * 2])
+            if b is None:
+                b_q = b_kv = None
+            else:
+                b_q, b_kv = b.split([E, E * 2])
+            q_proj = linear(q, w_q, b_q)
+            kv_proj = linear(k, w_kv, b_kv)
+            # reshape to 2, E and not E, 2 is deliberate for better memory coalescing and keeping same order as chunk()
+            kv_proj = kv_proj.unflatten(-1, (2, E)).unsqueeze(0).transpose(0, -2).squeeze(-2).contiguous()
+            return (q_proj, kv_proj[0], kv_proj[1])
+    else:
+        w_q, w_k, w_v = w.chunk(3)
+        if b is None:
+            b_q = b_k = b_v = None
+        else:
+            b_q, b_k, b_v = b.chunk(3)
+        return linear(q, w_q, b_q), linear(k, w_k, b_k), linear(v, w_v, b_v)
+
+
+def _in_projection(
+    q: Tensor,
+    k: Tensor,
+    v: Tensor,
+    w_q: Tensor,
+    w_k: Tensor,
+    w_v: Tensor,
+    b_q: Optional[Tensor] = None,
+    b_k: Optional[Tensor] = None,
+    b_v: Optional[Tensor] = None,
+) -> Tuple[Tensor, Tensor, Tensor]:
+    r"""
+    Performs the in-projection step of the attention operation. This is simply
+    a triple of linear projections, with shape constraints on the weights which
+    ensure embedding dimension uniformity in the projected outputs.
+    Output is a triple containing projection tensors for query, key and value.
+    Args:
+        q, k, v: query, key and value tensors to be projected.
+        w_q, w_k, w_v: weights for q, k and v, respectively.
+        b_q, b_k, b_v: optional biases for q, k and v, respectively.
+    Shape:
+        Inputs:
+        - q: :math:`(Qdims..., Eq)` where Eq is the query embedding dimension and Qdims are any
+            number of leading dimensions.
+        - k: :math:`(Kdims..., Ek)` where Ek is the key embedding dimension and Kdims are any
+            number of leading dimensions.
+        - v: :math:`(Vdims..., Ev)` where Ev is the value embedding dimension and Vdims are any
+            number of leading dimensions.
+        - w_q: :math:`(Eq, Eq)`
+        - w_k: :math:`(Eq, Ek)`
+        - w_v: :math:`(Eq, Ev)`
+        - b_q: :math:`(Eq)`
+        - b_k: :math:`(Eq)`
+        - b_v: :math:`(Eq)`
+        Output: in output triple :math:`(q', k', v')`,
+         - q': :math:`[Qdims..., Eq]`
+         - k': :math:`[Kdims..., Eq]`
+         - v': :math:`[Vdims..., Eq]`
+    """
+    Eq, Ek, Ev = q.size(-1), k.size(-1), v.size(-1)
+    assert w_q.shape == (Eq, Eq), f"expecting query weights shape of {(Eq, Eq)}, but got {w_q.shape}"
+    assert w_k.shape == (Eq, Ek), f"expecting key weights shape of {(Eq, Ek)}, but got {w_k.shape}"
+    assert w_v.shape == (Eq, Ev), f"expecting value weights shape of {(Eq, Ev)}, but got {w_v.shape}"
+    assert b_q is None or b_q.shape == (Eq,), f"expecting query bias shape of {(Eq,)}, but got {b_q.shape}"
+    assert b_k is None or b_k.shape == (Eq,), f"expecting key bias shape of {(Eq,)}, but got {b_k.shape}"
+    assert b_v is None or b_v.shape == (Eq,), f"expecting value bias shape of {(Eq,)}, but got {b_v.shape}"
+    return linear(q, w_q, b_q), linear(k, w_k, b_k), linear(v, w_v, b_v)

special_tokens_map.json

@@ -0,0 +1,264 @@
+{
+  "additional_special_tokens": [
+    {
+      "content": "<image>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "</image>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<ref>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "</ref>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<box>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "</box>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<quad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "</quad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<point>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "</point>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<slice>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "</slice>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<image_id>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "</image_id>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<unit>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "</unit>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<asr>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "</asr>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<query>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "</query>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|audio_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|audio|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|audio_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|spk_bos|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|spk|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|spk_eos|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|tts_bos|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|tts_eos|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|listen|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|speak|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|interrupt|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|vad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<|vad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<reserved_43>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    },
+    {
+      "content": "<reserved_53>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false
+    }
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": "<unk>"
+}

tokenization_minicpmo_fast.py

@@ -0,0 +1,110 @@
+# coding=utf-8
+# Copyright 2025 The OpenBMB Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from transformers import Qwen2TokenizerFast
+
+
+class MiniCPMOTokenizerFast(Qwen2TokenizerFast):
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+        # image
+        self.im_start = "<image>"
+        self.im_end = "</image>"
+        self.ref_start = "<ref>"
+        self.ref_end = "</ref>"
+        self.box_start = "<box>"
+        self.box_end = "</box>"
+        self.quad_start = "<quad>"
+        self.quad_end = "</quad>"
+        self.slice_start = "<slice>"
+        self.slice_end = "</slice>"
+        self.im_id_start = "<image_id>"
+        self.im_id_end = "</image_id>"
+
+        # audio
+        self.audio_start = "<|audio_start|>"
+        self.audio_end = "<|audio_end|>"
+        self.spk_start = "<|spk_bos|>"
+        self.spk_end = "<|spk_eos|>"
+        self.tts_start = "<|tts_bos|>"
+        self.tts_end = "<|tts_eos|>"
+
+    @property
+    def eos_id(self):
+        return self.eos_token_id
+
+    @property
+    def bos_id(self):
+        return self.bos_token_id
+
+    @property
+    def unk_id(self):
+        return self.unk_token_id
+
+    @property
+    def im_start_id(self):
+        return self.convert_tokens_to_ids(self.im_start)
+
+    @property
+    def im_end_id(self):
+        return self.convert_tokens_to_ids(self.im_end)
+
+    @property
+    def slice_start_id(self):
+        return self.convert_tokens_to_ids(self.slice_start)
+
+    @property
+    def slice_end_id(self):
+        return self.convert_tokens_to_ids(self.slice_end)
+
+    @property
+    def im_id_start_id(self):
+        return self.convert_tokens_to_ids(self.im_id_start)
+
+    @property
+    def im_id_end_id(self):
+        return self.convert_tokens_to_ids(self.im_id_end)
+
+    @property
+    def audio_start_id(self):
+        return self.convert_tokens_to_ids(self.audio_start)
+
+    @property
+    def audio_end_id(self):
+        return self.convert_tokens_to_ids(self.audio_end)
+
+    @property
+    def spk_start_id(self):
+        return self.convert_tokens_to_ids(self.spk_start)
+
+    @property
+    def spk_end_id(self):
+        return self.convert_tokens_to_ids(self.spk_end)
+
+    @property
+    def tts_start_id(self):
+        return self.convert_tokens_to_ids(self.tts_start)
+
+    @property
+    def tts_end_id(self):
+        return self.convert_tokens_to_ids(self.tts_end)
+
+    @staticmethod
+    def escape(text: str) -> str:
+        return text
+
+    @staticmethod
+    def unescape(text: str) -> str:
+        return text

tokenizer_config.json

@@ -0,0 +1,523 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "128244": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<image>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151666": {
+      "content": "</image>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151667": {
+      "content": "<ref>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151668": {
+      "content": "</ref>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151669": {
+      "content": "<box>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151670": {
+      "content": "</box>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151671": {
+      "content": "<quad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151672": {
+      "content": "</quad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151673": {
+      "content": "<point>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151674": {
+      "content": "</point>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151675": {
+      "content": "<slice>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151676": {
+      "content": "</slice>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151677": {
+      "content": "<image_id>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151678": {
+      "content": "</image_id>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151679": {
+      "content": "<unit>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151680": {
+      "content": "</unit>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151681": {
+      "content": "<asr>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151682": {
+      "content": "</asr>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151683": {
+      "content": "<query>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151684": {
+      "content": "</query>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151685": {
+      "content": "<|audio_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151686": {
+      "content": "<|audio|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151687": {
+      "content": "<|audio_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151688": {
+      "content": "<|spk_bos|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151689": {
+      "content": "<|spk|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151690": {
+      "content": "<|spk_eos|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151691": {
+      "content": "<|tts_bos|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151692": {
+      "content": "<|tts_eos|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151693": {
+      "content": "<|listen|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151694": {
+      "content": "<|speak|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151695": {
+      "content": "<|interrupt|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151696": {
+      "content": "<|vad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151697": {
+      "content": "<|vad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151698": {
+      "content": "<reserved_43>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151699": {
+      "content": "<reserved_53>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<image>",
+    "</image>",
+    "<ref>",
+    "</ref>",
+    "<box>",
+    "</box>",
+    "<quad>",
+    "</quad>",
+    "<point>",
+    "</point>",
+    "<slice>",
+    "</slice>",
+    "<image_id>",
+    "</image_id>",
+    "<unit>",
+    "</unit>",
+    "<asr>",
+    "</asr>",
+    "<query>",
+    "</query>",
+    "<|audio_start|>",
+    "<|audio|>",
+    "<|audio_end|>",
+    "<|spk_bos|>",
+    "<|spk|>",
+    "<|spk_eos|>",
+    "<|tts_bos|>",
+    "<|tts_eos|>",
+    "<|listen|>",
+    "<|speak|>",
+    "<|interrupt|>",
+    "<|vad_start|>",
+    "<|vad_end|>",
+    "<reserved_43>",
+    "<reserved_53>"
+  ],
+  "bos_token": "<|im_start|>",
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "auto_map": {
+    "AutoTokenizer": [
+      "tokenization_minicpmo_fast.MiniCPMOTokenizerFast",
+      null
+    ]
+  },
+  "tokenizer_class": "MiniCPMOTokenizerFast",
+  "unk_token": "<unk>"
+}

utils.py

@@ -0,0 +1,203 @@
+# coding=utf-8
+# Copyright 2025 The OpenBMB Team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import logging
+import re
+
+import librosa
+import numpy as np
+
+logger = logging.getLogger(__name__)
+
+
+def is_silent(data):
+    if np.abs(data).max() < 3e-3:
+        return True
+    else:
+        return False
+
+
+def sentence_end(txt):
+    for c in [".", "。", "!", "?", "！", "？"]:
+        if c in txt:
+            if c == ".":  # check not number before it like 1.
+                idx = txt.find(c)
+                if idx > 0:
+                    if txt[idx - 1].isdigit():
+                        continue
+            return c
+    return ""
+
+
+class NumberToTextConverter:
+    r"""
+    A helper class to ensure text-to-speech (TTS) systems read numeric digits
+    in the desired language (Chinese or English) digit-by-digit. It forcibly
+    replaces all numeric substrings in text with their language-specific
+    textual representations, thereby reducing the likelihood of TTS mistakes
+    on numbers.
+    Note: MiniCPM-o 2.6 only use this in streaming mode.
+
+    Attributes:
+        num_to_chinese (dict):
+            Mapping from digit (str) to its Chinese textual form (str).
+        num_to_english (dict):
+            Mapping from digit (str) to its English textual form (str).
+
+    Example:
+        >>> converter = NumberToTextConverter()
+        >>> converter.replace_numbers_with_text("我有2个苹果", language="chinese")
+        '我有两个苹果'
+        >>> converter.replace_numbers_with_text("I have 23 books", language="english")
+        'I have two three books'
+    """
+
+    def __init__(self):
+        self.num_to_chinese = {
+            "0": "零",
+            "1": "一",
+            "2": "二",
+            "3": "三",
+            "4": "四",
+            "5": "五",
+            "6": "六",
+            "7": "七",
+            "8": "八",
+            "9": "九",
+        }
+        self.num_to_english = {
+            "0": "zero",
+            "1": "one",
+            "2": "two",
+            "3": "three",
+            "4": "four",
+            "5": "five",
+            "6": "six",
+            "7": "seven",
+            "8": "eight",
+            "9": "nine",
+        }
+
+    def number_to_chinese_digit_by_digit(self, num_str):
+        result = ""
+        for char in num_str:
+            if char in self.num_to_chinese:
+                result += self.num_to_chinese[char]
+        return result
+
+    def number_to_english_digit_by_digit(self, num_str):
+        result = []
+        for char in num_str:
+            if char in self.num_to_english:
+                result.append(self.num_to_english[char])
+        return " ".join(result)
+
+    def detect_language(self, text):
+        chinese_count = len(re.findall(r"[\u4e00-\u9fff]", text))
+        english_count = len(re.findall(r"[a-zA-Z]", text))
+        return "chinese" if chinese_count >= english_count else "english"
+
+    def replace_numbers_with_text(self, text, language=None):
+        if language is None:
+            language = self.detect_language(text)
+        numbers = re.findall(r"\d+", text)
+
+        for num in numbers:
+            if language == "chinese":
+                replacement = self.number_to_chinese_digit_by_digit(num)
+            else:
+                replacement = self.number_to_english_digit_by_digit(num)
+            text = text.replace(num, replacement, 1)
+
+        return text
+
+
+class VoiceChecker:
+    r"""
+    A simple utility class to detect silence or low variation in consecutive audio chunks by comparing
+    the mel-spectrogram distances. It keeps track of consecutive zero-distance and low-distance chunks
+    to decide if the audio is considered "bad" (e.g., overly silent or not changing enough).
+
+    Attributes:
+        previous_mel (`np.ndarray` or `None`):
+            Holds the previously observed mel-spectrogram in decibel scale. Used to compute
+            the next distance; reset via :meth:`reset`.
+        consecutive_zeros (`int`):
+            The number of consecutive chunks that were detected as silent (distance = 0).
+        consecutive_low_distance (`int`):
+            The number of consecutive chunks whose distance was below the threshold.
+
+    Example:
+        >>> checker = VoiceChecker()
+        >>> # Suppose we have audio_wav (list or np.ndarray) and mel_spec (np.ndarray)
+        >>> # We split them into chunks and call checker.is_bad(...)
+        >>> is_audio_bad = checker.is_bad(audio_wav, mel_spec, chunk_size=2560, thresh=100.0)
+        >>> if is_audio_bad:
+        ...     print("Audio deemed bad!")
+        >>> # Reset states if needed
+        >>> checker.reset()
+    """
+
+    def __init__(self):
+        self.previous_mel = None
+        self.consecutive_zeros = 0
+        self.consecutive_low_distance = 0
+
+    def compute_distance(self, audio_chunk, mel_spec):
+        if is_silent(audio_chunk):
+            return 0.0  # 检查是否为空白片段
+
+        mel_db = librosa.power_to_db(mel_spec)
+        if self.previous_mel is None:
+            self.previous_mel = mel_db
+            return -1.0
+
+        distance = np.linalg.norm(np.mean(mel_db, axis=1) - np.mean(self.previous_mel, axis=1))
+        self.previous_mel = mel_db
+        return distance
+
+    def is_bad(self, audio_wav, mel_spec, chunk_size=2560, thresh=100.0):
+        num_chunks = len(audio_wav) // chunk_size
+        mel_chunk_size = mel_spec.shape[-1] // num_chunks
+        for i in range(num_chunks):
+            audio_chunk = audio_wav[i * chunk_size : (i + 1) * chunk_size]
+            mel_spec_chunk = mel_spec[:, i * mel_chunk_size : (i + 1) * mel_chunk_size]
+
+            distance = self.compute_distance(audio_chunk, mel_spec_chunk)
+            logger.warning(
+                f"mel dist: {distance:.1f}, zero: {self.consecutive_zeros}, low: {self.consecutive_low_distance}"
+            )
+            if distance == 0:
+                self.consecutive_low_distance = 0  # reset
+                self.consecutive_zeros += 1
+                if self.consecutive_zeros >= 12:
+                    logger.warning("VoiceChecker detected 1.2 s silent. Marking as failed.")
+                    return True
+            elif distance < thresh:
+                self.consecutive_zeros = 0
+                self.consecutive_low_distance += 1
+                if self.consecutive_low_distance >= 5:
+                    logger.warning("VoiceChecker detected 5 consecutive low distance chunks. Marking as failed.")
+                    return True
+            else:
+                self.consecutive_low_distance = 0
+                self.consecutive_zeros = 0
+
+        return False
+
+    def reset(self):
+        self.previous_mel = None
+        self.consecutive_zeros = 0
+        self.consecutive_low_distance = 0