Training in progress, step 400

README.md

@@ -0,0 +1,70 @@
+---
+license: other
+base_model: microsoft/Phi-3-mini-128k-instruct
+tags:
+- llama-factory
+- full
+- generated_from_trainer
+model-index:
+- name: sft
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/dongfu/huggingface/runs/4r5mtf1q)
+# sft
+
+This model is a fine-tuned version of [microsoft/Phi-3-mini-128k-instruct](https://huggingface.co/microsoft/Phi-3-mini-128k-instruct) on the pairrm-data dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.4639
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 1e-05
+- train_batch_size: 1
+- eval_batch_size: 1
+- seed: 42
+- distributed_type: multi-GPU
+- num_devices: 8
+- gradient_accumulation_steps: 16
+- total_train_batch_size: 128
+- total_eval_batch_size: 8
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: cosine
+- lr_scheduler_warmup_ratio: 0.05
+- num_epochs: 2
+
+### Training results
+
+| Training Loss | Epoch  | Step | Validation Loss |
+|:-------------:|:------:|:----:|:---------------:|
+| 0.4483        | 0.3469 | 500  | 0.4420          |
+| 0.453         | 0.6938 | 1000 | 0.4246          |
+| 0.2827        | 1.0407 | 1500 | 0.4557          |
+| 0.3098        | 1.3876 | 2000 | 0.4560          |
+| 0.2522        | 1.7345 | 2500 | 0.4652          |
+
+
+### Framework versions
+
+- Transformers 4.43.1
+- Pytorch 2.3.0+cu121
+- Datasets 2.20.0
+- Tokenizers 0.19.1

adapter_config.json

@@ -0,0 +1,31 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "microsoft/Phi-3-mini-128k-instruct",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_up_proj",
+    "qkv_proj",
+    "o_proj",
+    "down_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}

adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54f5f419c1266309c1e7823b7e7fde2909dd8355fec825493980b66b14bd687d
+size 50365768

added_tokens.json

@@ -0,0 +1,13 @@
+{
+  "<|assistant|>": 32001,
+  "<|endoftext|>": 32000,
+  "<|end|>": 32007,
+  "<|placeholder1|>": 32002,
+  "<|placeholder2|>": 32003,
+  "<|placeholder3|>": 32004,
+  "<|placeholder4|>": 32005,
+  "<|placeholder5|>": 32008,
+  "<|placeholder6|>": 32009,
+  "<|system|>": 32006,
+  "<|user|>": 32010
+}

all_results.json

@@ -0,0 +1,12 @@
+{
+    "epoch": 1.9994796635157401,
+    "eval_loss": 0.4639037847518921,
+    "eval_runtime": 767.9829,
+    "eval_samples_per_second": 12.644,
+    "eval_steps_per_second": 1.581,
+    "total_flos": 420396754255872.0,
+    "train_loss": 0.3626429720482505,
+    "train_runtime": 106139.7498,
+    "train_samples_per_second": 3.476,
+    "train_steps_per_second": 0.027
+}

config.json

@@ -0,0 +1,139 @@
+{
+    "_name_or_path": "microsoft/Phi-3-mini-128k-instruct",
+    "architectures": [
+        "Phi3ForCausalLM"
+    ],
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "auto_map": {
+        "AutoConfig": "configuration_phi3.Phi3Config",
+        "AutoModel": "modeling_phi3.Phi3ForCausalLM",
+        "AutoModelForCausalLM": "modeling_phi3.Phi3ForCausalLM"
+    },
+    "bos_token_id": 1,
+    "embd_pdrop": 0.0,
+    "eos_token_id": 32000,
+    "hidden_act": "silu",
+    "hidden_size": 3072,
+    "initializer_range": 0.02,
+    "intermediate_size": 8192,
+    "max_position_embeddings": 131072,
+    "model_type": "phi3",
+    "num_attention_heads": 32,
+    "num_hidden_layers": 32,
+    "num_key_value_heads": 32,
+    "original_max_position_embeddings": 4096,
+    "pad_token_id": 32000,
+    "resid_pdrop": 0.0,
+    "rms_norm_eps": 1e-05,
+    "rope_scaling": {
+        "long_factor": [
+            1.0700000524520874,
+            1.1200000047683716,
+            1.149999976158142,
+            1.4199999570846558,
+            1.5699999332427979,
+            1.7999999523162842,
+            2.129999876022339,
+            2.129999876022339,
+            3.009999990463257,
+            5.910000324249268,
+            6.950000286102295,
+            9.070000648498535,
+            9.930000305175781,
+            10.710000038146973,
+            11.130000114440918,
+            14.609999656677246,
+            15.409998893737793,
+            19.809999465942383,
+            37.279998779296875,
+            38.279998779296875,
+            38.599998474121094,
+            40.12000274658203,
+            46.20000457763672,
+            50.940006256103516,
+            53.66000747680664,
+            54.9373893737793,
+            56.89738845825195,
+            57.28738784790039,
+            59.98738479614258,
+            60.86738586425781,
+            60.887386322021484,
+            61.71739196777344,
+            62.91739273071289,
+            62.957393646240234,
+            63.41739273071289,
+            63.8173942565918,
+            63.83739471435547,
+            63.897396087646484,
+            63.93739700317383,
+            64.06739807128906,
+            64.11434936523438,
+            64.12435150146484,
+            64.15435028076172,
+            64.19435119628906,
+            64.24435424804688,
+            64.57435607910156,
+            64.69000244140625,
+            64.76000213623047
+        ],
+        "short_factor": [
+            1.1,
+            1.1,
+            1.1,
+            1.3000000000000003,
+            1.3500000000000003,
+            1.3500000000000003,
+            1.4000000000000004,
+            1.5500000000000005,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.000000000000001,
+            2.0500000000000007,
+            2.0500000000000007,
+            2.0500000000000007,
+            2.0500000000000007,
+            2.0500000000000007,
+            2.0500000000000007,
+            2.1000000000000005,
+            2.1000000000000005,
+            2.1500000000000004,
+            2.25,
+            2.25,
+            2.25,
+            2.25,
+            2.25,
+            2.3999999999999995,
+            2.4499999999999993,
+            2.499999999999999,
+            2.6999999999999984,
+            2.6999999999999984,
+            2.7499999999999982,
+            2.799999999999998,
+            2.8999999999999977,
+            3.049999999999997
+        ],
+        "type": "longrope"
+    },
+    "rope_theta": 10000.0,
+    "sliding_window": 262144,
+    "tie_word_embeddings": false,
+    "torch_dtype": "bfloat16",
+    "transformers_version": "4.43.1",
+    "use_cache": false,
+    "vocab_size": 32064
+}

configuration_phi3.py

@@ -0,0 +1,227 @@
+# coding=utf-8
+# Copyright 2024 Microsoft and the HuggingFace Inc. 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.
+
+""" Phi-3 model configuration"""
+
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+PHI3_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "microsoft/Phi-3-mini-4k-instruct": "https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/resolve/main/config.json",
+    "microsoft/Phi-3-mini-128k-instruct": "https://huggingface.co/microsoft/Phi-3-mini-128k-instruct/resolve/main/config.json",
+}
+
+
+class Phi3Config(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`Phi3Model`]. It is used to instantiate a Phi-3
+    model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
+    defaults will yield a similar configuration to that of the
+    [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct).
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 32064):
+            Vocabulary size of the Phi-3 model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`Phi3Model`].
+        hidden_size (`int`, *optional*, defaults to 3072):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 8192):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer decoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer decoder.
+        num_key_value_heads (`int`, *optional*):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details checkout [this
+            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to
+            `num_attention_heads`.
+        resid_pdrop (`float`, *optional*, defaults to 0.0):
+            Dropout probability for mlp outputs.
+        embd_pdrop (`int`, *optional*, defaults to 0.0):
+            The dropout ratio for the embeddings.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio after computing the attention scores.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to 4096):
+            The maximum sequence length that this model might ever be used with.
+        original_max_position_embeddings (`int`, *optional*, defaults to 4096):
+            The maximum sequence length that this model was trained with. This is used to determine the size of the
+            original RoPE embeddings when using long scaling.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-05):
+            The epsilon value used for the RMSNorm.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`. Whether to tie weight embeddings or not.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings
+        rope_theta (`float`, *optional*, defaults to 10000.0):
+            The base period of the RoPE embeddings.
+        rope_scaling (`dict`, *optional*):
+            The scaling strategy for the RoPE embeddings. If `None`, no scaling is applied. If a dictionary, it must
+            contain the following keys: `type`, `short_factor` and `long_factor`. The `type` must be `longrope` and 
+            the `short_factor` and `long_factor` must be lists of numbers with the same length as the hidden size 
+            divided by the number of attention heads divided by 2.
+        bos_token_id (`int`, *optional*, defaults to 1):
+            The id of the "beginning-of-sequence" token.
+        eos_token_id (`int`, *optional*, defaults to 32000):
+            The id of the "end-of-sequence" token.
+        pad_token_id (`int`, *optional*, defaults to 32000):
+            The id of the padding token.
+        sliding_window (`int`, *optional*):
+            Sliding window attention window size. If `None`, no sliding window is applied.
+
+    Example:
+
+    ```python
+    >>> from transformers import Phi3Model, Phi3Config
+
+    >>> # Initializing a Phi-3 style configuration
+    >>> configuration = Phi3Config.from_pretrained("microsoft/Phi-3-mini-4k-instruct")
+
+    >>> # Initializing a model from the configuration
+    >>> model = Phi3Model(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "phi3"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=32064,
+        hidden_size=3072,
+        intermediate_size=8192,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=None,
+        resid_pdrop=0.0,
+        embd_pdrop=0.0,
+        attention_dropout=0.0,
+        hidden_act="silu",
+        max_position_embeddings=4096,
+        original_max_position_embeddings=4096,
+        initializer_range=0.02,
+        rms_norm_eps=1e-5,
+        use_cache=True,
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        rope_scaling=None,
+        bos_token_id=1,
+        eos_token_id=32000,
+        pad_token_id=32000,
+        sliding_window=None,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.resid_pdrop = resid_pdrop
+        self.embd_pdrop = embd_pdrop
+        self.attention_dropout = attention_dropout
+        self.hidden_act = hidden_act
+        self.max_position_embeddings = max_position_embeddings
+        self.original_max_position_embeddings = original_max_position_embeddings
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self._rope_scaling_adjustment()
+        self._rope_scaling_validation()
+        self.sliding_window = sliding_window
+
+        super().__init__(
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            pad_token_id=pad_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+
+    def _rope_scaling_adjustment(self):
+        """
+        Adjust the `type` of the `rope_scaling` configuration for backward compatibility.
+        """
+        if self.rope_scaling is None:
+            return
+
+        rope_scaling_type = self.rope_scaling.get("type", None)
+
+        # For backward compatibility if previous version used "su" or "yarn"
+        if rope_scaling_type is not None and rope_scaling_type in ["su", "yarn"]:
+            self.rope_scaling["type"] = "longrope"
+
+    def _rope_scaling_validation(self):
+        """
+        Validate the `rope_scaling` configuration.
+        """
+        if self.rope_scaling is None:
+            return
+
+        if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) != 3:
+            raise ValueError(
+                "`rope_scaling` must be a dictionary with three fields, `type`, `short_factor` and `long_factor`, "
+                f"got {self.rope_scaling}"
+            )
+        rope_scaling_type = self.rope_scaling.get("type", None)
+        rope_scaling_short_factor = self.rope_scaling.get("short_factor", None)
+        rope_scaling_long_factor = self.rope_scaling.get("long_factor", None)
+        if rope_scaling_type is None or rope_scaling_type not in ["longrope"]:
+            raise ValueError(f"`rope_scaling`'s type field must be one of ['longrope'], got {rope_scaling_type}")
+        if not (
+            isinstance(rope_scaling_short_factor, list)
+            and all(isinstance(x, (int, float)) for x in rope_scaling_short_factor)
+        ):
+            raise ValueError(
+                f"`rope_scaling`'s short_factor field must be a list of numbers, got {rope_scaling_short_factor}"
+            )
+        if not len(rope_scaling_short_factor) == self.hidden_size // self.num_attention_heads // 2:
+            raise ValueError(
+                f"`rope_scaling`'s short_factor field must have length {self.hidden_size // self.num_attention_heads // 2}, got {len(rope_scaling_short_factor)}"
+            )
+        if not (
+            isinstance(rope_scaling_long_factor, list)
+            and all(isinstance(x, (int, float)) for x in rope_scaling_long_factor)
+        ):
+            raise ValueError(
+                f"`rope_scaling`'s long_factor field must be a list of numbers, got {rope_scaling_long_factor}"
+            )
+        if not len(rope_scaling_long_factor) == self.hidden_size // self.num_attention_heads // 2:
+            raise ValueError(
+                f"`rope_scaling`'s long_factor field must have length {self.hidden_size // self.num_attention_heads // 2}, got {len(rope_scaling_long_factor)}"
+            )

eval_results.json

@@ -0,0 +1,7 @@
+{
+    "epoch": 1.9994796635157401,
+    "eval_loss": 0.4639037847518921,
+    "eval_runtime": 767.9829,
+    "eval_samples_per_second": 12.644,
+    "eval_steps_per_second": 1.581
+}

generation_config.json

@@ -0,0 +1,11 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": [
+    32000,
+    32001,
+    32007
+  ],
+  "pad_token_id": 32000,
+  "transformers_version": "4.43.1"
+}

model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e51913b5d4b2c5b9f556692dde82a887ccf41ed386680e92d00e1589ed61b0b
+size 4972489328

model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff6b8aa0650a11fcfab7d2f27cacc25b029ff1dd10a83ff74e8dd2e04476c9c7
+size 2669692552

model.safetensors.index.json

@@ -0,0 +1,202 @@
+{
+  "metadata": {
+    "total_size": 7642159104
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00002-of-00002.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.21.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.mlp.gate_up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.qkv_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.gate_up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.qkv_proj.weight": "model-00001-of-00002.safetensors",
+    "model.norm.weight": "model-00002-of-00002.safetensors"
+  }
+}

modeling_phi3.py

@@ -0,0 +1,1563 @@
+# coding=utf-8
+# Copyright 2024 Microsoft and the HuggingFace Inc. 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 Phi-3 model."""
+
+import inspect
+import math
+import warnings
+from typing import List, Optional, Tuple, Union
+
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.modeling_attn_mask_utils import _prepare_4d_causal_attention_mask
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPast,
+    CausalLMOutputWithPast,
+    SequenceClassifierOutputWithPast,
+    TokenClassifierOutput,
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.utils import (
+    add_code_sample_docstrings,
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    is_flash_attn_2_available,
+    is_flash_attn_greater_or_equal_2_10,
+    logging,
+    replace_return_docstrings,
+)
+from .configuration_phi3 import Phi3Config
+
+
+logger = logging.get_logger(__name__)
+
+# Transformers scans dependencies in the modeling file, causing issues on conditional loading. The regex only ignores try/catch blocks, but not if statements
+# if is_flash_attn_2_available():
+_flash_supports_window_size = False
+try:
+    from flash_attn import flash_attn_func, flash_attn_varlen_func
+    from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input  # noqa
+
+    _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters)
+except ImportError as error:
+    logger.warning(
+        f"`flash-attention` package not found, consider installing for better performance: {error}."
+    )
+    if not _flash_supports_window_size:
+        logger.warning(
+            "Current `flash-attention` does not support `window_size`. Either upgrade or use `attn_implementation='eager'`."
+        )
+
+_CHECKPOINT_FOR_DOC = "microsoft/Phi-3-mini-4k-instruct"
+_CONFIG_FOR_DOC = "Phi3Config"
+
+PHI3_PRETRAINED_MODEL_ARCHIVE_LIST = [
+    "microsoft/Phi-3-mini-4k-instruct",
+    "microsoft/Phi-3-mini-128k-instruct",
+    # See all Phi-3 models at https://huggingface.co/models?filter=Phi-3
+]
+
+
+# Copied from transformers.models.llama.modeling_llama.LlamaRMSNorm with Llama->Phi3
+class Phi3RMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+        Phi3RMSNorm is equivalent to T5LayerNorm
+        """
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+        return self.weight * hidden_states.to(input_dtype)
+
+
+# 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.int32), (1, 0))
+    return (
+        indices,
+        cu_seqlens,
+        max_seqlen_in_batch,
+    )
+
+
+# Copied from transformers.models.gemma.modeling_gemma.GemmaRotaryEmbedding with gemma->phi3, Gemma->Phi3
+class Phi3RotaryEmbedding(nn.Module):
+    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
+        super().__init__()
+
+        self.dim = dim
+        self.max_position_embeddings = max_position_embeddings
+        self.base = base
+        self.register_buffer("inv_freq", None, persistent=False)
+
+    @torch.no_grad()
+    def forward(self, x, position_ids, seq_len=None):
+        # x: [bs, num_attention_heads, seq_len, head_size]
+        if self.inv_freq is None:
+            self.inv_freq = 1.0 / (
+                self.base ** (torch.arange(0, self.dim, 2, dtype=torch.int64, device=x.device).float() / self.dim)
+            )
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+        # Force float32 since bfloat16 loses precision on long contexts
+        # See https://github.com/huggingface/transformers/pull/29285
+        device_type = x.device.type
+        device_type = device_type if isinstance(device_type, str) and device_type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+class Phi3LongRoPEScaledRotaryEmbedding(Phi3RotaryEmbedding):
+    def __init__(self, dim, config, device=None):
+        super().__init__(dim, config.max_position_embeddings, config.rope_theta, device)
+
+        self.short_factor = config.rope_scaling["short_factor"]
+        self.long_factor = config.rope_scaling["long_factor"]
+        self.original_max_position_embeddings = config.original_max_position_embeddings
+
+    @torch.no_grad()
+    def forward(self, x, position_ids, seq_len=None):
+        seq_len = torch.max(position_ids) + 1
+        if seq_len > self.original_max_position_embeddings:
+            ext_factors = torch.tensor(self.long_factor, dtype=torch.float32, device=x.device)
+        else:
+            ext_factors = torch.tensor(self.short_factor, dtype=torch.float32, device=x.device)
+
+        inv_freq_shape = torch.arange(0, self.dim, 2, dtype=torch.int64, device=x.device).float() / self.dim
+        self.inv_freq = 1.0 / (ext_factors * self.base**inv_freq_shape)
+
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+
+        # Force float32 since bfloat16 loses precision on long contexts
+        # See https://github.com/huggingface/transformers/pull/29285
+        device_type = x.device.type
+        device_type = device_type if isinstance(device_type, str) and device_type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+
+            scale = self.max_position_embeddings / self.original_max_position_embeddings
+            if scale <= 1.0:
+                scaling_factor = 1.0
+            else:
+                scaling_factor = math.sqrt(1 + math.log(scale) / math.log(self.original_max_position_embeddings))
+
+            cos = emb.cos() * scaling_factor
+            sin = emb.sin() * scaling_factor
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+# Copied from transformers.models.llama.modeling_llama.rotate_half
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+# Copied from transformers.models.llama.modeling_llama.apply_rotary_pos_emb
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`, *optional*):
+            Deprecated and unused.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+class Phi3MLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+
+        self.config = config
+        self.gate_up_proj = nn.Linear(config.hidden_size, 2 * config.intermediate_size, bias=False)
+        self.down_proj = nn.Linear(config.intermediate_size, config.hidden_size, bias=False)
+
+        self.activation_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, hidden_states: torch.FloatTensor) -> torch.FloatTensor:
+        up_states = self.gate_up_proj(hidden_states)
+
+        gate, up_states = up_states.chunk(2, dim=-1)
+        up_states = up_states * self.activation_fn(gate)
+
+        return self.down_proj(up_states)
+
+
+# Copied from transformers.models.llama.modeling_llama.repeat_kv with llama->phi
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
+
+
+class Phi3Attention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: Phi3Config, layer_idx: Optional[int] = None):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        if layer_idx is None:
+            logger.warning_once(
+                f"Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will "
+                "lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` "
+                "when creating this class."
+            )
+
+        self.attention_dropout = config.attention_dropout
+        self.hidden_size = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.hidden_size // self.num_heads
+        self.num_key_value_heads = config.num_key_value_heads
+        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
+        self.max_position_embeddings = config.max_position_embeddings
+        self.original_max_position_embeddings = config.original_max_position_embeddings
+        self.rope_theta = config.rope_theta
+        self.rope_scaling = config.rope_scaling
+        self.is_causal = True
+
+        if (self.head_dim * self.num_heads) != self.hidden_size:
+            raise ValueError(
+                f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
+                f" and `num_heads`: {self.num_heads})."
+            )
+
+        op_size = self.num_heads * self.head_dim + 2 * (self.num_key_value_heads * self.head_dim)
+        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)
+        self.qkv_proj = nn.Linear(self.hidden_size, op_size, bias=False)
+        self._init_rope()
+
+    def _init_rope(self):
+        if self.rope_scaling is None:
+            self.rotary_emb = Phi3RotaryEmbedding(
+                self.head_dim,
+                max_position_embeddings=self.max_position_embeddings,
+                base=self.rope_theta,
+            )
+        else:
+            scaling_type = self.config.rope_scaling["type"]
+            if scaling_type == "longrope":
+                self.rotary_emb = Phi3LongRoPEScaledRotaryEmbedding(self.head_dim, self.config)
+            else:
+                raise ValueError(f"Unknown RoPE scaling type {scaling_type}")
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        logger.warning_once("You are not running the flash-attention implementation, expect numerical differences.")
+
+        bsz, q_len, _ = hidden_states.size()
+
+        qkv = self.qkv_proj(hidden_states)
+        query_pos = self.num_heads * self.head_dim
+        query_states = qkv[..., :query_pos]
+        key_states = qkv[..., query_pos : query_pos + self.num_key_value_heads * self.head_dim]
+        value_states = qkv[..., query_pos + self.num_key_value_heads * self.head_dim :]
+
+        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_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            if self.layer_idx is None:
+                raise ValueError(
+                    f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
+                    "for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
+                    "with a layer index."
+                )
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        cos, sin = self.rotary_emb(value_states, position_ids, 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)
+
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+
+        if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
+            raise ValueError(
+                f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is"
+                f" {attn_weights.size()}"
+            )
+
+        if attention_mask is not None:
+            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+                raise ValueError(
+                    f"Attention mask should be of size {(bsz, 1, q_len, kv_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(value_states.dtype)
+        attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training)
+
+        attn_output = torch.matmul(attn_weights, value_states)
+
+        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
+            raise ValueError(
+                f"`attn_output` should be of size {(bsz, 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(bsz, q_len, self.hidden_size)
+
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+
+class Phi3FlashAttention2(Phi3Attention):
+    """
+    Phi-3 flash attention module. This module inherits from `Phi3Attention` 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.
+    """
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaFlashAttention2.__init__
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+        # TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1.
+        # flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0.
+        # Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left).
+        self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        **kwargs,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        # Phi3FlashAttention2 attention does not support output_attentions
+
+        if not _flash_supports_window_size:
+            logger.warning_once(
+                "The current flash attention version does not support sliding window attention. Please use `attn_implementation='eager'` or upgrade flash-attn library."
+            )
+            raise ValueError("The current flash attention version does not support sliding window attention.")
+
+        output_attentions = False
+
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+
+            # overwrite attention_mask with padding_mask
+            attention_mask = kwargs.pop("padding_mask")
+
+        bsz, q_len, _ = hidden_states.size()
+
+        qkv = self.qkv_proj(hidden_states)
+        query_pos = self.num_heads * self.head_dim
+        query_states = qkv[..., :query_pos]
+        key_states = qkv[..., query_pos : query_pos + self.num_key_value_heads * self.head_dim]
+        value_states = qkv[..., query_pos + self.num_key_value_heads * self.head_dim :]
+
+        # 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_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            if self.layer_idx is None:
+                raise ValueError(
+                    f"The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} "
+                    "for auto-regressive decoding with k/v caching, please make sure to initialize the attention class "
+                    "with a layer index."
+                )
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+
+        # Because the input can be padded, the absolute sequence length depends on the max position id.
+        rotary_seq_len = max(kv_seq_len, position_ids[:, -1].max().item()) + 1
+        cos, sin = self.rotary_emb(value_states, position_ids, seq_len=rotary_seq_len)
+
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+
+        use_sliding_windows = (
+            _flash_supports_window_size
+            and getattr(self.config, "sliding_window", None) is not None
+            and kv_seq_len > self.config.sliding_window
+        )
+
+        if past_key_value is not None:
+            # Activate slicing cache only if the config has a value `sliding_windows` attribute
+            cache_has_contents = past_key_value.get_seq_length(self.layer_idx) > 0
+            if (
+                getattr(self.config, "sliding_window", None) is not None
+                and kv_seq_len > self.config.sliding_window
+                and cache_has_contents
+            ):
+                slicing_tokens = 1 - self.config.sliding_window
+
+                past_key = past_key_value[self.layer_idx][0]
+                past_value = past_key_value[self.layer_idx][1]
+
+                past_key = past_key[:, :, slicing_tokens:, :].contiguous()
+                past_value = past_value[:, :, slicing_tokens:, :].contiguous()
+
+                if past_key.shape[-2] != self.config.sliding_window - 1:
+                    raise ValueError(
+                        f"past key must have a shape of (`batch_size, num_heads, self.config.sliding_window-1, head_dim`), got"
+                        f" {past_key.shape}"
+                    )
+
+                if attention_mask is not None:
+                    attention_mask = attention_mask[:, slicing_tokens:]
+                    attention_mask = torch.cat([attention_mask, torch.ones_like(attention_mask[:, -1:])], dim=-1)
+
+            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)
+
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        attn_dropout = self.attention_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.
+
+        if query_states.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.qkv_proj.weight.dtype
+
+            logger.warning_once(
+                f"The input hidden states seems to be silently casted in float32, this might be related to"
+                f" 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)
+
+        # Reashape to the expected shape for Flash Attention
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+
+        attn_output = self._flash_attention_forward(
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            q_len,
+            dropout=attn_dropout,
+            use_sliding_windows=use_sliding_windows,
+        )
+
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+    # Copied from transformers.models.mistral.modeling_mistral.MistralFlashAttention2._flash_attention_forward
+    def _flash_attention_forward(
+        self,
+        query_states,
+        key_states,
+        value_states,
+        attention_mask,
+        query_length,
+        dropout=0.0,
+        softmax_scale=None,
+        use_sliding_windows=False,
+    ):
+        """
+        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 (`float`):
+                Attention dropout
+            softmax_scale (`float`, *optional*):
+                The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
+            use_sliding_windows (`bool`, *optional*):
+                Whether to activate sliding window attention.
+        """
+        if not self._flash_attn_uses_top_left_mask:
+            causal = self.is_causal
+        else:
+            # 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
+
+            if not use_sliding_windows:
+                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,
+                )
+            else:
+                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,
+                    window_size=(self.config.sliding_window, self.config.sliding_window),
+                )
+
+            attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
+        else:
+            if not use_sliding_windows:
+                attn_output = flash_attn_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                )
+            else:
+                attn_output = flash_attn_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                    window_size=(self.config.sliding_window, self.config.sliding_window),
+                )
+
+        return attn_output
+
+    # Copied from transformers.models.mistral.modeling_mistral.MistralFlashAttention2._upad_input
+    def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
+        batch_size, kv_seq_len, num_heads, head_dim = key_layer.shape
+
+        # On the first iteration we need to properly re-create the padding mask
+        # by slicing it on the proper place
+        if kv_seq_len != attention_mask.shape[-1]:
+            attention_mask_num_tokens = attention_mask.shape[-1]
+            attention_mask = attention_mask[:, attention_mask_num_tokens - kv_seq_len :]
+
+        indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
+
+        key_layer = index_first_axis(key_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
+        value_layer = index_first_axis(value_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
+
+        if query_length == kv_seq_len:
+            query_layer = index_first_axis(
+                query_layer.reshape(batch_size * kv_seq_len, 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.llama.modeling_llama.LlamaSdpaAttention with Llama->Phi3
+# TODO @Arthur no longer copied from LLama after static cache
+class Phi3SdpaAttention(Phi3Attention):
+    """
+    Phi3 attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from
+    `Phi3Attention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to
+    SDPA API.
+    """
+
+    # Adapted from Phi3Attention.forward
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        if output_attentions:
+            # TODO: Improve this warning with e.g. `model.config.attn_implementation = "manual"` once this is implemented.
+            logger.warning_once(
+                "Phi3Model is using Phi3SdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, "
+                'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
+            )
+            return super().forward(
+                hidden_states=hidden_states,
+                attention_mask=attention_mask,
+                position_ids=position_ids,
+                past_key_value=past_key_value,
+                output_attentions=output_attentions,
+                use_cache=use_cache,
+            )
+
+        bsz, q_len, _ = hidden_states.size()
+
+        qkv = self.qkv_proj(hidden_states)
+        query_pos = self.num_heads * self.head_dim
+        query_states = qkv[..., :query_pos]
+        key_states = qkv[..., query_pos : query_pos + self.num_key_value_heads * self.head_dim]
+        value_states = qkv[..., query_pos + self.num_key_value_heads * self.head_dim :]
+
+        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_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_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, position_ids, 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)
+
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        if attention_mask is not None:
+            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+                raise ValueError(
+                    f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
+                )
+
+        # SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask,
+        # Reference: https://github.com/pytorch/pytorch/issues/112577.
+        if query_states.device.type == "cuda" and attention_mask is not None:
+            query_states = query_states.contiguous()
+            key_states = key_states.contiguous()
+            value_states = value_states.contiguous()
+
+        attn_output = torch.nn.functional.scaled_dot_product_attention(
+            query_states,
+            key_states,
+            value_states,
+            attn_mask=attention_mask,
+            dropout_p=self.attention_dropout if self.training else 0.0,
+            # The q_len > 1 is necessary to match with AttentionMaskConverter.to_causal_4d that does not create a causal mask in case q_len == 1.
+            is_causal=self.is_causal and attention_mask is None and q_len > 1,
+        )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.view(bsz, q_len, self.hidden_size)
+
+        attn_output = self.o_proj(attn_output)
+
+        return attn_output, None, past_key_value
+
+
+PHI3_ATTENTION_CLASSES = {
+    "eager": Phi3Attention,
+    "flash_attention_2": Phi3FlashAttention2,
+    "sdpa": Phi3SdpaAttention,
+}
+
+
+class Phi3DecoderLayer(nn.Module):
+    def __init__(self, config: Phi3Config, layer_idx: int):
+        super().__init__()
+
+        self.config = config
+        self.self_attn = PHI3_ATTENTION_CLASSES[config._attn_implementation](config, layer_idx=layer_idx)
+
+        self.mlp = Phi3MLP(config)
+        self.input_layernorm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+        self.resid_attn_dropout = nn.Dropout(config.resid_pdrop)
+        self.resid_mlp_dropout = nn.Dropout(config.resid_pdrop)
+        self.post_attention_layernorm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        **kwargs,
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`):
+                input to the layer of shape `(batch, seq_len, embed_dim)`
+            attention_mask (`torch.FloatTensor`, *optional*): attention mask of size
+                `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
+            position_ids (`torch.LongTensor` of shape `({0})`, *optional*):
+                Indices of positions of each input sequence tokens in the position embeddings. Selected in the range
+                `[0, config.n_positions - 1]`. [What are position IDs?](../glossary#position-ids)
+            output_attentions (`bool`, *optional*):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            use_cache (`bool`, *optional*):
+                If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
+                (see `past_key_values`).
+            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
+        """
+
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        attn_outputs, self_attn_weights, present_key_value = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+        )
+
+        hidden_states = residual + self.resid_attn_dropout(attn_outputs)
+
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + self.resid_mlp_dropout(hidden_states)
+
+        outputs = (hidden_states,)
+
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        if use_cache:
+            outputs += (present_key_value,)
+
+        return outputs
+
+
+PHI3_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 ([`Phi3Config`]):
+            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.
+"""
+
+
+@add_start_docstrings(
+    "The bare Phi-3 model outputting raw hidden-states without any specific head on top.",
+    PHI3_START_DOCSTRING,
+)
+class Phi3PreTrainedModel(PreTrainedModel):
+    config_class = Phi3Config
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["Phi3DecoderLayer"]
+    _skip_keys_device_placement = "past_key_values"
+    _supports_flash_attn_2 = True
+    _supports_sdpa = False
+    _supports_cache_class = True
+
+    _version = "0.0.5"
+
+    def _init_weights(self, module):
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+
+
+PHI3_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        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)
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            If `past_key_values` is used, optionally only the last `input_ids` have to be input (see
+            `past_key_values`).
+
+            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
+            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
+            information on the default strategy.
+
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.n_positions - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        past_key_values (`Cache` or `tuple(tuple(torch.FloatTensor))`, *optional*):
+            Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
+            blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values`
+            returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`.
+
+            Two formats are allowed:
+            - a [`~cache_utils.Cache`] instance;
+            - Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of
+            shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`). This is also known as the legacy
+            cache format.
+
+            The model will output the same cache format that is fed as input. If no `past_key_values` are passed, the
+            legacy cache format will be returned.
+
+            If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't
+            have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `input_ids`
+            of shape `(batch_size, sequence_length)`.
+        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            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.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        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.
+"""
+
+
+@add_start_docstrings(
+    "The bare Phi-3 model outputting raw hidden-states without any specific head on top.",
+    PHI3_START_DOCSTRING,
+)
+class Phi3Model(Phi3PreTrainedModel):
+    """
+    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`Phi3DecoderLayer`]
+
+    Args:
+        config: Phi3Config
+    """
+
+    def __init__(self, config: Phi3Config):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
+        self.embed_dropout = nn.Dropout(config.embd_pdrop)
+        self.layers = nn.ModuleList(
+            [Phi3DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self._attn_implementation = config._attn_implementation
+        self.norm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+        self.gradient_checkpointing = False
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.embed_tokens = value
+
+    @add_start_docstrings_to_model_forward(PHI3_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, BaseModelOutputWithPast]:
+        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
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # retrieve input_ids and inputs_embeds
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
+        elif input_ids is not None:
+            batch_size, seq_length = input_ids.shape[:2]
+        elif inputs_embeds is not None:
+            batch_size, seq_length = inputs_embeds.shape[:2]
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        past_key_values_length = 0
+
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+                )
+                use_cache = False
+
+        if use_cache:
+            use_legacy_cache = not isinstance(past_key_values, Cache)
+            if use_legacy_cache:
+                past_key_values = DynamicCache.from_legacy_cache(past_key_values)
+            past_key_values_length = past_key_values.get_usable_length(seq_length)
+
+        if position_ids is None:
+            device = input_ids.device if input_ids is not None else inputs_embeds.device
+            position_ids = torch.arange(
+                past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
+            )
+            position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
+        else:
+            position_ids = position_ids.view(-1, seq_length).long()
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if attention_mask is not None and self._attn_implementation == "flash_attention_2" and use_cache:
+            is_padding_right = attention_mask[:, -1].sum().item() != batch_size
+            if is_padding_right:
+                raise ValueError(
+                    "You are attempting to perform batched generation with padding_side='right'"
+                    " this may lead to unexpected behaviour for Flash Attention version of Phi3. Make sure to "
+                    " call `tokenizer.padding_side  = 'left'` before tokenizing the input. "
+                )
+
+        if self._attn_implementation == "flash_attention_2":
+            # 2d mask is passed through the layers
+            attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None
+        else:
+            # 4d mask is passed through the layers
+            attention_mask = _prepare_4d_causal_attention_mask(
+                attention_mask,
+                (batch_size, seq_length),
+                inputs_embeds,
+                past_key_values_length,
+                sliding_window=self.config.sliding_window,
+            )
+
+        hidden_states = inputs_embeds
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        next_decoder_cache = None
+
+        for decoder_layer in self.layers:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    decoder_layer.__call__,
+                    hidden_states,
+                    attention_mask,
+                    position_ids,
+                    past_key_values,
+                    output_attentions,
+                    use_cache,
+                )
+            else:
+                layer_outputs = decoder_layer(
+                    hidden_states,
+                    attention_mask=attention_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_values,
+                    output_attentions=output_attentions,
+                    use_cache=use_cache,
+                )
+
+            hidden_states = layer_outputs[0]
+
+            if use_cache:
+                next_decoder_cache = layer_outputs[2 if output_attentions else 1]
+
+            if output_attentions:
+                all_self_attns += (layer_outputs[1],)
+
+        hidden_states = self.norm(hidden_states)
+
+        # add hidden states from the last decoder layer
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+
+        next_cache = None
+        if use_cache:
+            next_cache = next_decoder_cache.to_legacy_cache() if use_legacy_cache else next_decoder_cache
+        if not return_dict:
+            return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=next_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+
+
+class Phi3ForCausalLM(Phi3PreTrainedModel):
+    _tied_weights_keys = ["lm_head.weight"]
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.__init__ with Llama->Phi3
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = Phi3Model(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.get_input_embeddings
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.set_input_embeddings
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.get_output_embeddings
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.set_output_embeddings
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.set_decoder
+    def set_decoder(self, decoder):
+        self.model = decoder
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.get_decoder
+    def get_decoder(self):
+        return self.model
+
+    # Ignore copy
+    @add_start_docstrings_to_model_forward(PHI3_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, CausalLMOutputWithPast]:
+        r"""
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+
+        Returns:
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, Phi3ForCausalLM
+
+        >>> model = Phi3ForCausalLM.from_pretrained("microsoft/phi-3-mini-4k-instruct")
+        >>> tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-3-mini-4k-instruct")
+
+        >>> prompt = "This is an example script ."
+        >>> inputs = tokenizer(prompt, return_tensors="pt")
+
+        >>> # Generate
+        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
+        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        'This is an example script .\n Certainly! Below is a sample script that demonstrates a simple task, such as calculating the sum'
+        ```"""
+
+        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
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = outputs[0]
+        logits = self.lm_head(hidden_states)
+        logits = logits.float()
+
+        loss = None
+        if labels is not None:
+            # Shift so that tokens < n predict n
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            # Flatten the tokens
+            loss_fct = CrossEntropyLoss()
+            shift_logits = shift_logits.view(-1, self.config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            # Enable model parallelism
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss = loss_fct(shift_logits, shift_labels)
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            return (loss,) + output if loss is not None else output
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    # Copied from transformers.models.persimmon.modeling_persimmon.PersimmonForCausalLM.prepare_inputs_for_generation
+    def prepare_inputs_for_generation(
+        self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
+    ):
+        if past_key_values is not None:
+            if isinstance(past_key_values, Cache):
+                cache_length = past_key_values.get_seq_length()
+                past_length = past_key_values.seen_tokens
+                max_cache_length = past_key_values.get_max_length()
+            else:
+                cache_length = past_length = past_key_values[0][0].shape[2]
+                max_cache_length = None
+
+            # Keep only the unprocessed tokens:
+            # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
+            # some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
+            # input)
+            if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
+                input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :]
+            # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
+            # input_ids based on the past_length.
+            elif past_length < input_ids.shape[1]:
+                input_ids = input_ids[:, past_length:]
+            # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.
+
+            # If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
+            if (
+                max_cache_length is not None
+                and attention_mask is not None
+                and cache_length + input_ids.shape[1] > max_cache_length
+            ):
+                attention_mask = attention_mask[:, -max_cache_length:]
+
+        position_ids = kwargs.get("position_ids", None)
+        if attention_mask is not None and position_ids is None:
+            # create position_ids on the fly for batch generation
+            position_ids = attention_mask.long().cumsum(-1) - 1
+            position_ids.masked_fill_(attention_mask == 0, 1)
+            if past_key_values:
+                position_ids = position_ids[:, -input_ids.shape[1] :]
+
+        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
+        if inputs_embeds is not None and past_key_values is None:
+            model_inputs = {"inputs_embeds": inputs_embeds}
+        else:
+            model_inputs = {"input_ids": input_ids}
+
+        model_inputs.update(
+            {
+                "position_ids": position_ids,
+                "past_key_values": past_key_values,
+                "use_cache": kwargs.get("use_cache"),
+                "attention_mask": attention_mask,
+            }
+        )
+        return model_inputs
+
+    @staticmethod
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM._reorder_cache
+    def _reorder_cache(past_key_values, beam_idx):
+        reordered_past = ()
+        for layer_past in past_key_values:
+            reordered_past += (
+                tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past),
+            )
+        return reordered_past
+
+
+@add_start_docstrings(
+    """
+    The [`Phi3Model`] with a sequence classification head on top (linear layer).
+
+    [`Phi3ForSequenceClassification`] uses the last token in order to do the classification, as other causal models
+    (e.g. GPT-2) do.
+
+    Since it does classification on the last token, it requires to know the position of the last token. If a
+    `pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each row. If
+    no `pad_token_id` is defined, it simply takes the last value in each row of the batch. Since it cannot guess the
+    padding tokens when `inputs_embeds` are passed instead of `input_ids`, it does the same (take the last value in
+    each row of the batch).
+    """,
+    PHI3_START_DOCSTRING,
+)
+# Copied from transformers.models.llama.modeling_llama.LlamaForSequenceClassification with Llama->Phi3, LLAMA->PHI3, self.transformer->self.model, transformer_outputs->model_outputs
+class Phi3ForSequenceClassification(Phi3PreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = Phi3Model(config)
+        self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    @add_start_docstrings_to_model_forward(PHI3_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        model_outputs = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        hidden_states = model_outputs[0]
+        logits = self.score(hidden_states)
+
+        if input_ids is not None:
+            batch_size = input_ids.shape[0]
+        else:
+            batch_size = inputs_embeds.shape[0]
+
+        if self.config.pad_token_id is None and batch_size != 1:
+            raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.")
+        if self.config.pad_token_id is None:
+            sequence_lengths = -1
+        else:
+            if input_ids is not None:
+                # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
+                sequence_lengths = torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
+                sequence_lengths = sequence_lengths % input_ids.shape[-1]
+                sequence_lengths = sequence_lengths.to(logits.device)
+            else:
+                sequence_lengths = -1
+
+        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = "regression"
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = "single_label_classification"
+                else:
+                    self.config.problem_type = "multi_label_classification"
+
+            if self.config.problem_type == "regression":
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(pooled_logits, labels)
+            elif self.config.problem_type == "single_label_classification":
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1))
+            elif self.config.problem_type == "multi_label_classification":
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(pooled_logits, labels)
+        if not return_dict:
+            output = (pooled_logits,) + model_outputs[1:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutputWithPast(
+            loss=loss,
+            logits=pooled_logits,
+            past_key_values=model_outputs.past_key_values,
+            hidden_states=model_outputs.hidden_states,
+            attentions=model_outputs.attentions,
+        )
+
+
+@add_start_docstrings(
+    """
+    [`Phi3Model`] with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for
+    Named-Entity-Recognition (NER) tasks.
+    """,
+    PHI3_START_DOCSTRING,
+)
+# Copied from transformers.models.mpt.modeling_mpt.MptForTokenClassification with Mpt->Phi3,MPT->PHI3,self.transformer->self.model,transformer_outputs->model_outputs
+class Phi3ForTokenClassification(Phi3PreTrainedModel):
+    def __init__(self, config: Phi3Config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.model = Phi3Model(config)
+        if hasattr(config, "classifier_dropout") and config.classifier_dropout is not None:
+            classifier_dropout = config.classifier_dropout
+        elif hasattr(config, "hidden_dropout") and config.hidden_dropout is not None:
+            classifier_dropout = config.hidden_dropout
+        else:
+            classifier_dropout = 0.1
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @add_start_docstrings_to_model_forward(PHI3_INPUTS_DOCSTRING)
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=TokenClassifierOutput,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **deprecated_arguments,
+    ) -> Union[Tuple[torch.Tensor], TokenClassifierOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        model_outputs = self.model(
+            input_ids,
+            past_key_values=past_key_values,
+            attention_mask=attention_mask,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = model_outputs[0]
+        hidden_states = self.dropout(hidden_states)
+        logits = self.classifier(hidden_states)
+
+        loss = None
+        if labels is not None:
+            # move labels to correct device to enable model parallelism
+            labels = labels.to(logits.device)
+            batch_size, seq_length = labels.shape
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(
+                logits.view(batch_size * seq_length, self.num_labels), labels.view(batch_size * seq_length)
+            )
+
+        if not return_dict:
+            output = (logits,) + model_outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return TokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=model_outputs.hidden_states,
+            attentions=model_outputs.attentions,
+        )

special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|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": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

tokenizer_config.json

@@ -0,0 +1,132 @@
+{
+  "add_bos_token": false,
+  "add_eos_token": false,
+  "add_prefix_space": null,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": false
+    },
+    "32000": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32001": {
+      "content": "<|assistant|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32002": {
+      "content": "<|placeholder1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32003": {
+      "content": "<|placeholder2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32004": {
+      "content": "<|placeholder3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32005": {
+      "content": "<|placeholder4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32006": {
+      "content": "<|system|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32007": {
+      "content": "<|end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32008": {
+      "content": "<|placeholder5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32009": {
+      "content": "<|placeholder6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    },
+    "32010": {
+      "content": "<|user|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "chat_template": "{{ '<s>' }}{% if messages[0]['role'] == 'system' %}{% set system_message = messages[0]['content'] %}{% endif %}{% if system_message is defined %}{{ '<|system|>\n' + system_message + '<|end|>\n' }}{% endif %}{% for message in messages %}{% set content = message['content'] %}{% if message['role'] == 'user' %}{{ '<|user|>\n' + content + '<|end|>\n<|assistant|>\n' }}{% elif message['role'] == 'assistant' %}{{ content + '<|end|>' + '\n' }}{% endif %}{% endfor %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|end|>",
+  "legacy": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "split_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.9994796635157401,
+    "total_flos": 420396754255872.0,
+    "train_loss": 0.3626429720482505,
+    "train_runtime": 106139.7498,
+    "train_samples_per_second": 3.476,
+    "train_steps_per_second": 0.027
+}

trainer_log.jsonl

@@ -0,0 +1,205 @@
+{"current_steps": 2, "total_steps": 780, "loss": 0.7593, "learning_rate": 5.128205128205128e-07, "epoch": 0.0025622547842101048, "percentage": 0.26, "elapsed_time": "0:00:23", "remaining_time": "2:34:45", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 4, "total_steps": 780, "loss": 0.8076, "learning_rate": 1.0256410256410257e-06, "epoch": 0.0051245095684202095, "percentage": 0.51, "elapsed_time": "0:00:48", "remaining_time": "2:35:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 6, "total_steps": 780, "loss": 0.7059, "learning_rate": 1.5384615384615387e-06, "epoch": 0.007686764352630314, "percentage": 0.77, "elapsed_time": "0:01:15", "remaining_time": "2:43:15", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 8, "total_steps": 780, "loss": 0.715, "learning_rate": 2.0512820512820513e-06, "epoch": 0.010249019136840419, "percentage": 1.03, "elapsed_time": "0:01:47", "remaining_time": "2:52:15", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 10, "total_steps": 780, "loss": 0.7175, "learning_rate": 2.564102564102564e-06, "epoch": 0.012811273921050524, "percentage": 1.28, "elapsed_time": "0:02:12", "remaining_time": "2:50:07", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 12, "total_steps": 780, "loss": 0.6976, "learning_rate": 3.0769230769230774e-06, "epoch": 0.015373528705260629, "percentage": 1.54, "elapsed_time": "0:02:43", "remaining_time": "2:54:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 14, "total_steps": 780, "loss": 0.7615, "learning_rate": 3.58974358974359e-06, "epoch": 0.017935783489470735, "percentage": 1.79, "elapsed_time": "0:03:07", "remaining_time": "2:51:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 16, "total_steps": 780, "loss": 0.6735, "learning_rate": 4.102564102564103e-06, "epoch": 0.020498038273680838, "percentage": 2.05, "elapsed_time": "0:03:34", "remaining_time": "2:50:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 18, "total_steps": 780, "loss": 0.6885, "learning_rate": 4.615384615384616e-06, "epoch": 0.023060293057890945, "percentage": 2.31, "elapsed_time": "0:03:58", "remaining_time": "2:48:30", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 20, "total_steps": 780, "loss": 0.768, "learning_rate": 5.128205128205128e-06, "epoch": 0.025622547842101048, "percentage": 2.56, "elapsed_time": "0:04:22", "remaining_time": "2:46:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 22, "total_steps": 780, "loss": 0.7092, "learning_rate": 5.641025641025641e-06, "epoch": 0.028184802626311154, "percentage": 2.82, "elapsed_time": "0:04:48", "remaining_time": "2:45:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 24, "total_steps": 780, "loss": 0.6764, "learning_rate": 6.153846153846155e-06, "epoch": 0.030747057410521257, "percentage": 3.08, "elapsed_time": "0:05:13", "remaining_time": "2:44:50", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 26, "total_steps": 780, "loss": 0.6386, "learning_rate": 6.666666666666667e-06, "epoch": 0.033309312194731364, "percentage": 3.33, "elapsed_time": "0:05:41", "remaining_time": "2:44:55", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 28, "total_steps": 780, "loss": 0.6218, "learning_rate": 7.17948717948718e-06, "epoch": 0.03587156697894147, "percentage": 3.59, "elapsed_time": "0:06:05", "remaining_time": "2:43:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 30, "total_steps": 780, "loss": 0.5853, "learning_rate": 7.692307692307694e-06, "epoch": 0.03843382176315158, "percentage": 3.85, "elapsed_time": "0:06:34", "remaining_time": "2:44:14", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 32, "total_steps": 780, "loss": 0.5441, "learning_rate": 8.205128205128205e-06, "epoch": 0.040996076547361676, "percentage": 4.1, "elapsed_time": "0:07:01", "remaining_time": "2:44:10", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 34, "total_steps": 780, "loss": 0.5458, "learning_rate": 8.717948717948719e-06, "epoch": 0.04355833133157178, "percentage": 4.36, "elapsed_time": "0:07:26", "remaining_time": "2:43:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 36, "total_steps": 780, "loss": 0.4589, "learning_rate": 9.230769230769232e-06, "epoch": 0.04612058611578189, "percentage": 4.62, "elapsed_time": "0:07:50", "remaining_time": "2:41:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 38, "total_steps": 780, "loss": 0.4386, "learning_rate": 9.743589743589744e-06, "epoch": 0.048682840899991996, "percentage": 4.87, "elapsed_time": "0:08:13", "remaining_time": "2:40:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 40, "total_steps": 780, "loss": 0.4215, "learning_rate": 9.99995506314361e-06, "epoch": 0.051245095684202095, "percentage": 5.13, "elapsed_time": "0:08:38", "remaining_time": "2:39:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 42, "total_steps": 780, "loss": 0.3888, "learning_rate": 9.999595573138845e-06, "epoch": 0.0538073504684122, "percentage": 5.38, "elapsed_time": "0:09:06", "remaining_time": "2:39:55", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 44, "total_steps": 780, "loss": 0.3749, "learning_rate": 9.99887661897616e-06, "epoch": 0.05636960525262231, "percentage": 5.64, "elapsed_time": "0:09:31", "remaining_time": "2:39:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 46, "total_steps": 780, "loss": 0.3543, "learning_rate": 9.997798252347382e-06, "epoch": 0.058931860036832415, "percentage": 5.9, "elapsed_time": "0:09:59", "remaining_time": "2:39:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 48, "total_steps": 780, "loss": 0.3565, "learning_rate": 9.996360550785619e-06, "epoch": 0.061494114821042514, "percentage": 6.15, "elapsed_time": "0:10:27", "remaining_time": "2:39:23", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 50, "total_steps": 780, "loss": 0.3242, "learning_rate": 9.994563617659665e-06, "epoch": 0.06405636960525263, "percentage": 6.41, "elapsed_time": "0:10:54", "remaining_time": "2:39:14", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 52, "total_steps": 780, "loss": 0.3334, "learning_rate": 9.992407582166582e-06, "epoch": 0.06661862438946273, "percentage": 6.67, "elapsed_time": "0:11:23", "remaining_time": "2:39:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 54, "total_steps": 780, "loss": 0.3741, "learning_rate": 9.989892599322404e-06, "epoch": 0.06918087917367283, "percentage": 6.92, "elapsed_time": "0:11:52", "remaining_time": "2:39:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 56, "total_steps": 780, "loss": 0.3331, "learning_rate": 9.987018849950996e-06, "epoch": 0.07174313395788294, "percentage": 7.18, "elapsed_time": "0:12:17", "remaining_time": "2:38:59", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 58, "total_steps": 780, "loss": 0.3335, "learning_rate": 9.983786540671052e-06, "epoch": 0.07430538874209304, "percentage": 7.44, "elapsed_time": "0:12:44", "remaining_time": "2:38:39", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 60, "total_steps": 780, "loss": 0.3344, "learning_rate": 9.980195903881231e-06, "epoch": 0.07686764352630315, "percentage": 7.69, "elapsed_time": "0:13:09", "remaining_time": "2:37:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 62, "total_steps": 780, "loss": 0.3055, "learning_rate": 9.976247197743465e-06, "epoch": 0.07942989831051325, "percentage": 7.95, "elapsed_time": "0:13:31", "remaining_time": "2:36:38", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 64, "total_steps": 780, "loss": 0.3187, "learning_rate": 9.97194070616438e-06, "epoch": 0.08199215309472335, "percentage": 8.21, "elapsed_time": "0:13:55", "remaining_time": "2:35:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 66, "total_steps": 780, "loss": 0.2998, "learning_rate": 9.967276738774897e-06, "epoch": 0.08455440787893347, "percentage": 8.46, "elapsed_time": "0:14:23", "remaining_time": "2:35:39", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 68, "total_steps": 780, "loss": 0.3251, "learning_rate": 9.962255630907964e-06, "epoch": 0.08711666266314357, "percentage": 8.72, "elapsed_time": "0:14:48", "remaining_time": "2:35:02", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 70, "total_steps": 780, "loss": 0.317, "learning_rate": 9.956877743574437e-06, "epoch": 0.08967891744735366, "percentage": 8.97, "elapsed_time": "0:15:15", "remaining_time": "2:34:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 72, "total_steps": 780, "loss": 0.31, "learning_rate": 9.951143463437145e-06, "epoch": 0.09224117223156378, "percentage": 9.23, "elapsed_time": "0:15:41", "remaining_time": "2:34:18", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 74, "total_steps": 780, "loss": 0.3158, "learning_rate": 9.94505320278307e-06, "epoch": 0.09480342701577388, "percentage": 9.49, "elapsed_time": "0:16:05", "remaining_time": "2:33:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 76, "total_steps": 780, "loss": 0.3163, "learning_rate": 9.938607399493714e-06, "epoch": 0.09736568179998399, "percentage": 9.74, "elapsed_time": "0:16:30", "remaining_time": "2:32:59", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 78, "total_steps": 780, "loss": 0.3132, "learning_rate": 9.931806517013612e-06, "epoch": 0.09992793658419409, "percentage": 10.0, "elapsed_time": "0:16:56", "remaining_time": "2:32:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 80, "total_steps": 780, "loss": 0.315, "learning_rate": 9.924651044317017e-06, "epoch": 0.10249019136840419, "percentage": 10.26, "elapsed_time": "0:17:20", "remaining_time": "2:31:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 82, "total_steps": 780, "loss": 0.2851, "learning_rate": 9.917141495872733e-06, "epoch": 0.1050524461526143, "percentage": 10.51, "elapsed_time": "0:17:45", "remaining_time": "2:31:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 84, "total_steps": 780, "loss": 0.3036, "learning_rate": 9.909278411607134e-06, "epoch": 0.1076147009368244, "percentage": 10.77, "elapsed_time": "0:18:10", "remaining_time": "2:30:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 86, "total_steps": 780, "loss": 0.3374, "learning_rate": 9.90106235686534e-06, "epoch": 0.11017695572103452, "percentage": 11.03, "elapsed_time": "0:18:38", "remaining_time": "2:30:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 88, "total_steps": 780, "loss": 0.316, "learning_rate": 9.892493922370575e-06, "epoch": 0.11273921050524462, "percentage": 11.28, "elapsed_time": "0:19:02", "remaining_time": "2:29:45", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 90, "total_steps": 780, "loss": 0.3284, "learning_rate": 9.883573724181683e-06, "epoch": 0.11530146528945472, "percentage": 11.54, "elapsed_time": "0:19:25", "remaining_time": "2:28:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 92, "total_steps": 780, "loss": 0.3266, "learning_rate": 9.87430240364885e-06, "epoch": 0.11786372007366483, "percentage": 11.79, "elapsed_time": "0:19:51", "remaining_time": "2:28:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 94, "total_steps": 780, "loss": 0.3104, "learning_rate": 9.864680627367476e-06, "epoch": 0.12042597485787493, "percentage": 12.05, "elapsed_time": "0:20:16", "remaining_time": "2:28:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 96, "total_steps": 780, "loss": 0.3221, "learning_rate": 9.854709087130261e-06, "epoch": 0.12298822964208503, "percentage": 12.31, "elapsed_time": "0:20:43", "remaining_time": "2:27:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 98, "total_steps": 780, "loss": 0.2913, "learning_rate": 9.844388499877457e-06, "epoch": 0.12555048442629513, "percentage": 12.56, "elapsed_time": "0:21:11", "remaining_time": "2:27:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 100, "total_steps": 780, "loss": 0.2974, "learning_rate": 9.833719607645325e-06, "epoch": 0.12811273921050526, "percentage": 12.82, "elapsed_time": "0:21:37", "remaining_time": "2:27:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 102, "total_steps": 780, "loss": 0.3698, "learning_rate": 9.822703177512783e-06, "epoch": 0.13067499399471535, "percentage": 13.08, "elapsed_time": "0:22:06", "remaining_time": "2:26:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 104, "total_steps": 780, "loss": 0.2995, "learning_rate": 9.811340001546252e-06, "epoch": 0.13323724877892545, "percentage": 13.33, "elapsed_time": "0:22:30", "remaining_time": "2:26:20", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 106, "total_steps": 780, "loss": 0.304, "learning_rate": 9.799630896742716e-06, "epoch": 0.13579950356313555, "percentage": 13.59, "elapsed_time": "0:22:57", "remaining_time": "2:25:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 108, "total_steps": 780, "loss": 0.3032, "learning_rate": 9.787576704970965e-06, "epoch": 0.13836175834734565, "percentage": 13.85, "elapsed_time": "0:23:20", "remaining_time": "2:25:13", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 110, "total_steps": 780, "loss": 0.2915, "learning_rate": 9.77517829291108e-06, "epoch": 0.14092401313155578, "percentage": 14.1, "elapsed_time": "0:23:44", "remaining_time": "2:24:38", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 112, "total_steps": 780, "loss": 0.3192, "learning_rate": 9.762436551992117e-06, "epoch": 0.14348626791576588, "percentage": 14.36, "elapsed_time": "0:24:09", "remaining_time": "2:24:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 114, "total_steps": 780, "loss": 0.2999, "learning_rate": 9.74935239832801e-06, "epoch": 0.14604852269997598, "percentage": 14.62, "elapsed_time": "0:24:33", "remaining_time": "2:23:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 116, "total_steps": 780, "loss": 0.31, "learning_rate": 9.735926772651703e-06, "epoch": 0.14861077748418608, "percentage": 14.87, "elapsed_time": "0:24:59", "remaining_time": "2:23:03", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 118, "total_steps": 780, "loss": 0.292, "learning_rate": 9.722160640247523e-06, "epoch": 0.15117303226839618, "percentage": 15.13, "elapsed_time": "0:25:25", "remaining_time": "2:22:35", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 120, "total_steps": 780, "loss": 0.3077, "learning_rate": 9.708054990881763e-06, "epoch": 0.1537352870526063, "percentage": 15.38, "elapsed_time": "0:25:51", "remaining_time": "2:22:12", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 122, "total_steps": 780, "loss": 0.3132, "learning_rate": 9.693610838731532e-06, "epoch": 0.1562975418368164, "percentage": 15.64, "elapsed_time": "0:26:16", "remaining_time": "2:21:41", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 124, "total_steps": 780, "loss": 0.3139, "learning_rate": 9.678829222311827e-06, "epoch": 0.1588597966210265, "percentage": 15.9, "elapsed_time": "0:26:40", "remaining_time": "2:21:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 126, "total_steps": 780, "loss": 0.2992, "learning_rate": 9.663711204400872e-06, "epoch": 0.1614220514052366, "percentage": 16.15, "elapsed_time": "0:27:08", "remaining_time": "2:20:50", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 128, "total_steps": 780, "loss": 0.3066, "learning_rate": 9.6482578719637e-06, "epoch": 0.1639843061894467, "percentage": 16.41, "elapsed_time": "0:27:36", "remaining_time": "2:20:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 130, "total_steps": 780, "loss": 0.3121, "learning_rate": 9.632470336074009e-06, "epoch": 0.1665465609736568, "percentage": 16.67, "elapsed_time": "0:27:59", "remaining_time": "2:19:59", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 132, "total_steps": 780, "loss": 0.2991, "learning_rate": 9.616349731834271e-06, "epoch": 0.16910881575786693, "percentage": 16.92, "elapsed_time": "0:28:24", "remaining_time": "2:19:26", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 134, "total_steps": 780, "loss": 0.3227, "learning_rate": 9.599897218294122e-06, "epoch": 0.17167107054207703, "percentage": 17.18, "elapsed_time": "0:28:48", "remaining_time": "2:18:53", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 136, "total_steps": 780, "loss": 0.3025, "learning_rate": 9.583113978367026e-06, "epoch": 0.17423332532628713, "percentage": 17.44, "elapsed_time": "0:29:16", "remaining_time": "2:18:37", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 138, "total_steps": 780, "loss": 0.3179, "learning_rate": 9.56600121874523e-06, "epoch": 0.17679558011049723, "percentage": 17.69, "elapsed_time": "0:29:46", "remaining_time": "2:18:33", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 140, "total_steps": 780, "loss": 0.2906, "learning_rate": 9.548560169812997e-06, "epoch": 0.17935783489470733, "percentage": 17.95, "elapsed_time": "0:30:14", "remaining_time": "2:18:15", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 142, "total_steps": 780, "loss": 0.2968, "learning_rate": 9.530792085558151e-06, "epoch": 0.18192008967891746, "percentage": 18.21, "elapsed_time": "0:30:38", "remaining_time": "2:17:40", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 144, "total_steps": 780, "loss": 0.3076, "learning_rate": 9.512698243481914e-06, "epoch": 0.18448234446312756, "percentage": 18.46, "elapsed_time": "0:31:06", "remaining_time": "2:17:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 146, "total_steps": 780, "loss": 0.292, "learning_rate": 9.49427994450705e-06, "epoch": 0.18704459924733766, "percentage": 18.72, "elapsed_time": "0:31:32", "remaining_time": "2:16:57", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 148, "total_steps": 780, "loss": 0.3337, "learning_rate": 9.47553851288434e-06, "epoch": 0.18960685403154776, "percentage": 18.97, "elapsed_time": "0:32:00", "remaining_time": "2:16:39", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 150, "total_steps": 780, "loss": 0.2854, "learning_rate": 9.45647529609736e-06, "epoch": 0.19216910881575786, "percentage": 19.23, "elapsed_time": "0:32:23", "remaining_time": "2:16:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 152, "total_steps": 780, "loss": 0.328, "learning_rate": 9.437091664765611e-06, "epoch": 0.19473136359996798, "percentage": 19.49, "elapsed_time": "0:32:46", "remaining_time": "2:15:23", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 154, "total_steps": 780, "loss": 0.2831, "learning_rate": 9.41738901254596e-06, "epoch": 0.19729361838417808, "percentage": 19.74, "elapsed_time": "0:33:14", "remaining_time": "2:15:09", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 156, "total_steps": 780, "loss": 0.2899, "learning_rate": 9.397368756032445e-06, "epoch": 0.19985587316838818, "percentage": 20.0, "elapsed_time": "0:33:39", "remaining_time": "2:14:39", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 158, "total_steps": 780, "loss": 0.2796, "learning_rate": 9.37703233465443e-06, "epoch": 0.20241812795259828, "percentage": 20.26, "elapsed_time": "0:34:03", "remaining_time": "2:14:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 160, "total_steps": 780, "loss": 0.2965, "learning_rate": 9.356381210573092e-06, "epoch": 0.20498038273680838, "percentage": 20.51, "elapsed_time": "0:34:27", "remaining_time": "2:13:32", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 162, "total_steps": 780, "loss": 0.2884, "learning_rate": 9.33541686857632e-06, "epoch": 0.2075426375210185, "percentage": 20.77, "elapsed_time": "0:34:55", "remaining_time": "2:13:14", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 164, "total_steps": 780, "loss": 0.297, "learning_rate": 9.31414081597194e-06, "epoch": 0.2101048923052286, "percentage": 21.03, "elapsed_time": "0:35:22", "remaining_time": "2:12:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 166, "total_steps": 780, "loss": 0.2862, "learning_rate": 9.292554582479349e-06, "epoch": 0.2126671470894387, "percentage": 21.28, "elapsed_time": "0:35:46", "remaining_time": "2:12:18", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 168, "total_steps": 780, "loss": 0.2958, "learning_rate": 9.270659720119533e-06, "epoch": 0.2152294018736488, "percentage": 21.54, "elapsed_time": "0:36:15", "remaining_time": "2:12:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 170, "total_steps": 780, "loss": 0.2988, "learning_rate": 9.248457803103476e-06, "epoch": 0.2177916566578589, "percentage": 21.79, "elapsed_time": "0:36:42", "remaining_time": "2:11:43", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 172, "total_steps": 780, "loss": 0.2803, "learning_rate": 9.225950427718974e-06, "epoch": 0.22035391144206903, "percentage": 22.05, "elapsed_time": "0:37:10", "remaining_time": "2:11:22", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 174, "total_steps": 780, "loss": 0.2957, "learning_rate": 9.203139212215868e-06, "epoch": 0.22291616622627913, "percentage": 22.31, "elapsed_time": "0:37:35", "remaining_time": "2:10:53", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 176, "total_steps": 780, "loss": 0.2933, "learning_rate": 9.180025796689692e-06, "epoch": 0.22547842101048923, "percentage": 22.56, "elapsed_time": "0:38:01", "remaining_time": "2:10:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 178, "total_steps": 780, "loss": 0.2926, "learning_rate": 9.156611842963753e-06, "epoch": 0.22804067579469933, "percentage": 22.82, "elapsed_time": "0:38:24", "remaining_time": "2:09:52", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 180, "total_steps": 780, "loss": 0.3111, "learning_rate": 9.132899034469648e-06, "epoch": 0.23060293057890943, "percentage": 23.08, "elapsed_time": "0:38:47", "remaining_time": "2:09:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 182, "total_steps": 780, "loss": 0.2948, "learning_rate": 9.108889076126226e-06, "epoch": 0.23316518536311953, "percentage": 23.33, "elapsed_time": "0:39:11", "remaining_time": "2:08:46", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 184, "total_steps": 780, "loss": 0.2872, "learning_rate": 9.084583694217012e-06, "epoch": 0.23572744014732966, "percentage": 23.59, "elapsed_time": "0:39:36", "remaining_time": "2:08:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 186, "total_steps": 780, "loss": 0.3073, "learning_rate": 9.059984636266082e-06, "epoch": 0.23828969493153976, "percentage": 23.85, "elapsed_time": "0:40:04", "remaining_time": "2:08:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 188, "total_steps": 780, "loss": 0.2866, "learning_rate": 9.035093670912424e-06, "epoch": 0.24085194971574986, "percentage": 24.1, "elapsed_time": "0:40:30", "remaining_time": "2:07:33", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 190, "total_steps": 780, "loss": 0.298, "learning_rate": 9.009912587782772e-06, "epoch": 0.24341420449995996, "percentage": 24.36, "elapsed_time": "0:40:54", "remaining_time": "2:07:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 192, "total_steps": 780, "loss": 0.2644, "learning_rate": 8.984443197362938e-06, "epoch": 0.24597645928417006, "percentage": 24.62, "elapsed_time": "0:41:18", "remaining_time": "2:06:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 194, "total_steps": 780, "loss": 0.2986, "learning_rate": 8.958687330867634e-06, "epoch": 0.24853871406838018, "percentage": 24.87, "elapsed_time": "0:41:44", "remaining_time": "2:06:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 196, "total_steps": 780, "loss": 0.2826, "learning_rate": 8.932646840108818e-06, "epoch": 0.25110096885259026, "percentage": 25.13, "elapsed_time": "0:42:09", "remaining_time": "2:05:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 198, "total_steps": 780, "loss": 0.2824, "learning_rate": 8.906323597362547e-06, "epoch": 0.2536632236368004, "percentage": 25.38, "elapsed_time": "0:42:37", "remaining_time": "2:05:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 200, "total_steps": 780, "loss": 0.2836, "learning_rate": 8.879719495234363e-06, "epoch": 0.2562254784210105, "percentage": 25.64, "elapsed_time": "0:43:02", "remaining_time": "2:04:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 202, "total_steps": 780, "loss": 0.2799, "learning_rate": 8.852836446523213e-06, "epoch": 0.2587877332052206, "percentage": 25.9, "elapsed_time": "0:43:30", "remaining_time": "2:04:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 204, "total_steps": 780, "loss": 0.3027, "learning_rate": 8.825676384083936e-06, "epoch": 0.2613499879894307, "percentage": 26.15, "elapsed_time": "0:43:55", "remaining_time": "2:04:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 206, "total_steps": 780, "loss": 0.3032, "learning_rate": 8.798241260688273e-06, "epoch": 0.2639122427736408, "percentage": 26.41, "elapsed_time": "0:44:20", "remaining_time": "2:03:33", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 208, "total_steps": 780, "loss": 0.3044, "learning_rate": 8.770533048884483e-06, "epoch": 0.2664744975578509, "percentage": 26.67, "elapsed_time": "0:44:45", "remaining_time": "2:03:06", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 210, "total_steps": 780, "loss": 0.2784, "learning_rate": 8.742553740855507e-06, "epoch": 0.26903675234206104, "percentage": 26.92, "elapsed_time": "0:45:12", "remaining_time": "2:02:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 212, "total_steps": 780, "loss": 0.3142, "learning_rate": 8.71430534827574e-06, "epoch": 0.2715990071262711, "percentage": 27.18, "elapsed_time": "0:45:42", "remaining_time": "2:02:26", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 214, "total_steps": 780, "loss": 0.2592, "learning_rate": 8.685789902166395e-06, "epoch": 0.27416126191048124, "percentage": 27.44, "elapsed_time": "0:46:08", "remaining_time": "2:02:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 216, "total_steps": 780, "loss": 0.2881, "learning_rate": 8.657009452749466e-06, "epoch": 0.2767235166946913, "percentage": 27.69, "elapsed_time": "0:46:32", "remaining_time": "2:01:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 218, "total_steps": 780, "loss": 0.3017, "learning_rate": 8.627966069300332e-06, "epoch": 0.27928577147890143, "percentage": 27.95, "elapsed_time": "0:46:58", "remaining_time": "2:01:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 220, "total_steps": 780, "loss": 0.2781, "learning_rate": 8.598661839998972e-06, "epoch": 0.28184802626311156, "percentage": 28.21, "elapsed_time": "0:47:22", "remaining_time": "2:00:35", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 222, "total_steps": 780, "loss": 0.296, "learning_rate": 8.569098871779828e-06, "epoch": 0.28441028104732163, "percentage": 28.46, "elapsed_time": "0:47:51", "remaining_time": "2:00:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 224, "total_steps": 780, "loss": 0.3161, "learning_rate": 8.539279290180315e-06, "epoch": 0.28697253583153176, "percentage": 28.72, "elapsed_time": "0:48:18", "remaining_time": "1:59:53", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 226, "total_steps": 780, "loss": 0.2948, "learning_rate": 8.509205239188017e-06, "epoch": 0.28953479061574183, "percentage": 28.97, "elapsed_time": "0:48:45", "remaining_time": "1:59:30", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 228, "total_steps": 780, "loss": 0.2736, "learning_rate": 8.478878881086505e-06, "epoch": 0.29209704539995196, "percentage": 29.23, "elapsed_time": "0:49:08", "remaining_time": "1:58:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 230, "total_steps": 780, "loss": 0.2954, "learning_rate": 8.448302396299906e-06, "epoch": 0.2946593001841621, "percentage": 29.49, "elapsed_time": "0:49:32", "remaining_time": "1:58:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 232, "total_steps": 780, "loss": 0.3134, "learning_rate": 8.417477983236107e-06, "epoch": 0.29722155496837216, "percentage": 29.74, "elapsed_time": "0:49:57", "remaining_time": "1:58:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 234, "total_steps": 780, "loss": 0.2767, "learning_rate": 8.386407858128707e-06, "epoch": 0.2997838097525823, "percentage": 30.0, "elapsed_time": "0:50:21", "remaining_time": "1:57:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 236, "total_steps": 780, "loss": 0.2783, "learning_rate": 8.355094254877665e-06, "epoch": 0.30234606453679236, "percentage": 30.26, "elapsed_time": "0:50:46", "remaining_time": "1:57:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 238, "total_steps": 780, "loss": 0.2871, "learning_rate": 8.323539424888695e-06, "epoch": 0.3049083193210025, "percentage": 30.51, "elapsed_time": "0:51:10", "remaining_time": "1:56:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 240, "total_steps": 780, "loss": 0.2747, "learning_rate": 8.291745636911382e-06, "epoch": 0.3074705741052126, "percentage": 30.77, "elapsed_time": "0:51:40", "remaining_time": "1:56:15", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 242, "total_steps": 780, "loss": 0.2737, "learning_rate": 8.259715176876069e-06, "epoch": 0.3100328288894227, "percentage": 31.03, "elapsed_time": "0:52:08", "remaining_time": "1:55:54", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 244, "total_steps": 780, "loss": 0.2889, "learning_rate": 8.2274503477295e-06, "epoch": 0.3125950836736328, "percentage": 31.28, "elapsed_time": "0:52:37", "remaining_time": "1:55:35", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 246, "total_steps": 780, "loss": 0.2822, "learning_rate": 8.19495346926924e-06, "epoch": 0.3151573384578429, "percentage": 31.54, "elapsed_time": "0:53:00", "remaining_time": "1:55:03", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 248, "total_steps": 780, "loss": 0.284, "learning_rate": 8.162226877976886e-06, "epoch": 0.317719593242053, "percentage": 31.79, "elapsed_time": "0:53:28", "remaining_time": "1:54:41", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 250, "total_steps": 780, "loss": 0.2915, "learning_rate": 8.129272926850079e-06, "epoch": 0.32028184802626314, "percentage": 32.05, "elapsed_time": "0:53:52", "remaining_time": "1:54:12", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 252, "total_steps": 780, "loss": 0.2842, "learning_rate": 8.096093985233323e-06, "epoch": 0.3228441028104732, "percentage": 32.31, "elapsed_time": "0:54:18", "remaining_time": "1:53:46", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 254, "total_steps": 780, "loss": 0.3203, "learning_rate": 8.062692438647628e-06, "epoch": 0.32540635759468334, "percentage": 32.56, "elapsed_time": "0:54:44", "remaining_time": "1:53:22", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 256, "total_steps": 780, "loss": 0.2828, "learning_rate": 8.029070688619013e-06, "epoch": 0.3279686123788934, "percentage": 32.82, "elapsed_time": "0:55:10", "remaining_time": "1:52:55", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 258, "total_steps": 780, "loss": 0.2672, "learning_rate": 7.995231152505815e-06, "epoch": 0.33053086716310354, "percentage": 33.08, "elapsed_time": "0:55:36", "remaining_time": "1:52:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 260, "total_steps": 780, "loss": 0.292, "learning_rate": 7.961176263324902e-06, "epoch": 0.3330931219473136, "percentage": 33.33, "elapsed_time": "0:56:04", "remaining_time": "1:52:09", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 262, "total_steps": 780, "loss": 0.2987, "learning_rate": 7.92690846957673e-06, "epoch": 0.33565537673152374, "percentage": 33.59, "elapsed_time": "0:56:29", "remaining_time": "1:51:41", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 264, "total_steps": 780, "loss": 0.2881, "learning_rate": 7.892430235069317e-06, "epoch": 0.33821763151573386, "percentage": 33.85, "elapsed_time": "0:56:58", "remaining_time": "1:51:22", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 266, "total_steps": 780, "loss": 0.2912, "learning_rate": 7.857744038741076e-06, "epoch": 0.34077988629994393, "percentage": 34.1, "elapsed_time": "0:57:22", "remaining_time": "1:50:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 268, "total_steps": 780, "loss": 0.2672, "learning_rate": 7.822852374482597e-06, "epoch": 0.34334214108415406, "percentage": 34.36, "elapsed_time": "0:57:51", "remaining_time": "1:50:32", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 270, "total_steps": 780, "loss": 0.2921, "learning_rate": 7.787757750957335e-06, "epoch": 0.34590439586836413, "percentage": 34.62, "elapsed_time": "0:58:17", "remaining_time": "1:50:06", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 272, "total_steps": 780, "loss": 0.2676, "learning_rate": 7.752462691421245e-06, "epoch": 0.34846665065257426, "percentage": 34.87, "elapsed_time": "0:58:42", "remaining_time": "1:49:38", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 274, "total_steps": 780, "loss": 0.2576, "learning_rate": 7.716969733541357e-06, "epoch": 0.3510289054367844, "percentage": 35.13, "elapsed_time": "0:59:05", "remaining_time": "1:49:06", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 276, "total_steps": 780, "loss": 0.2686, "learning_rate": 7.681281429213328e-06, "epoch": 0.35359116022099446, "percentage": 35.38, "elapsed_time": "0:59:35", "remaining_time": "1:48:49", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 278, "total_steps": 780, "loss": 0.2678, "learning_rate": 7.645400344377953e-06, "epoch": 0.3561534150052046, "percentage": 35.64, "elapsed_time": "1:00:01", "remaining_time": "1:48:24", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 280, "total_steps": 780, "loss": 0.2907, "learning_rate": 7.609329058836694e-06, "epoch": 0.35871566978941466, "percentage": 35.9, "elapsed_time": "1:00:28", "remaining_time": "1:47:59", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 282, "total_steps": 780, "loss": 0.298, "learning_rate": 7.5730701660661795e-06, "epoch": 0.3612779245736248, "percentage": 36.15, "elapsed_time": "1:00:52", "remaining_time": "1:47:30", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 284, "total_steps": 780, "loss": 0.263, "learning_rate": 7.536626273031747e-06, "epoch": 0.3638401793578349, "percentage": 36.41, "elapsed_time": "1:01:20", "remaining_time": "1:47:07", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 286, "total_steps": 780, "loss": 0.2733, "learning_rate": 7.500000000000001e-06, "epoch": 0.366402434142045, "percentage": 36.67, "elapsed_time": "1:01:46", "remaining_time": "1:46:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 288, "total_steps": 780, "loss": 0.3159, "learning_rate": 7.4631939803504215e-06, "epoch": 0.3689646889262551, "percentage": 36.92, "elapsed_time": "1:02:13", "remaining_time": "1:46:17", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 290, "total_steps": 780, "loss": 0.2878, "learning_rate": 7.426210860386032e-06, "epoch": 0.3715269437104652, "percentage": 37.18, "elapsed_time": "1:02:39", "remaining_time": "1:45:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 292, "total_steps": 780, "loss": 0.2829, "learning_rate": 7.3890532991431174e-06, "epoch": 0.3740891984946753, "percentage": 37.44, "elapsed_time": "1:03:04", "remaining_time": "1:45:25", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 294, "total_steps": 780, "loss": 0.2646, "learning_rate": 7.3517239682000675e-06, "epoch": 0.37665145327888544, "percentage": 37.69, "elapsed_time": "1:03:28", "remaining_time": "1:44:56", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 296, "total_steps": 780, "loss": 0.301, "learning_rate": 7.314225551485273e-06, "epoch": 0.3792137080630955, "percentage": 37.95, "elapsed_time": "1:03:53", "remaining_time": "1:44:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 298, "total_steps": 780, "loss": 0.2622, "learning_rate": 7.276560745084167e-06, "epoch": 0.38177596284730564, "percentage": 38.21, "elapsed_time": "1:04:18", "remaining_time": "1:44:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 300, "total_steps": 780, "loss": 0.2901, "learning_rate": 7.2387322570453724e-06, "epoch": 0.3843382176315157, "percentage": 38.46, "elapsed_time": "1:04:45", "remaining_time": "1:43:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 302, "total_steps": 780, "loss": 0.2576, "learning_rate": 7.2007428071860045e-06, "epoch": 0.38690047241572584, "percentage": 38.72, "elapsed_time": "1:05:10", "remaining_time": "1:43:10", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 304, "total_steps": 780, "loss": 0.2716, "learning_rate": 7.162595126896111e-06, "epoch": 0.38946272719993597, "percentage": 38.97, "elapsed_time": "1:05:35", "remaining_time": "1:42:42", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 306, "total_steps": 780, "loss": 0.2716, "learning_rate": 7.1242919589422974e-06, "epoch": 0.39202498198414604, "percentage": 39.23, "elapsed_time": "1:06:02", "remaining_time": "1:42:18", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 308, "total_steps": 780, "loss": 0.2978, "learning_rate": 7.085836057270521e-06, "epoch": 0.39458723676835616, "percentage": 39.49, "elapsed_time": "1:06:27", "remaining_time": "1:41:50", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 310, "total_steps": 780, "loss": 0.2499, "learning_rate": 7.047230186808085e-06, "epoch": 0.39714949155256624, "percentage": 39.74, "elapsed_time": "1:06:54", "remaining_time": "1:41:26", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 312, "total_steps": 780, "loss": 0.3018, "learning_rate": 7.008477123264849e-06, "epoch": 0.39971174633677636, "percentage": 40.0, "elapsed_time": "1:07:18", "remaining_time": "1:40:58", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 314, "total_steps": 780, "loss": 0.2834, "learning_rate": 6.96957965293365e-06, "epoch": 0.4022740011209865, "percentage": 40.26, "elapsed_time": "1:07:47", "remaining_time": "1:40:36", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 316, "total_steps": 780, "loss": 0.3008, "learning_rate": 6.9305405724899876e-06, "epoch": 0.40483625590519656, "percentage": 40.51, "elapsed_time": "1:08:14", "remaining_time": "1:40:12", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 318, "total_steps": 780, "loss": 0.2753, "learning_rate": 6.891362688790925e-06, "epoch": 0.4073985106894067, "percentage": 40.77, "elapsed_time": "1:08:39", "remaining_time": "1:39:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 320, "total_steps": 780, "loss": 0.2943, "learning_rate": 6.8520488186733e-06, "epoch": 0.40996076547361676, "percentage": 41.03, "elapsed_time": "1:09:03", "remaining_time": "1:39:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 322, "total_steps": 780, "loss": 0.2692, "learning_rate": 6.812601788751192e-06, "epoch": 0.4125230202578269, "percentage": 41.28, "elapsed_time": "1:09:29", "remaining_time": "1:38:50", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 324, "total_steps": 780, "loss": 0.2961, "learning_rate": 6.773024435212678e-06, "epoch": 0.415085275042037, "percentage": 41.54, "elapsed_time": "1:09:53", "remaining_time": "1:38:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 326, "total_steps": 780, "loss": 0.2898, "learning_rate": 6.733319603615941e-06, "epoch": 0.4176475298262471, "percentage": 41.79, "elapsed_time": "1:10:20", "remaining_time": "1:37:57", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 328, "total_steps": 780, "loss": 0.2555, "learning_rate": 6.693490148684654e-06, "epoch": 0.4202097846104572, "percentage": 42.05, "elapsed_time": "1:10:44", "remaining_time": "1:37:29", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 330, "total_steps": 780, "loss": 0.3043, "learning_rate": 6.653538934102743e-06, "epoch": 0.4227720393946673, "percentage": 42.31, "elapsed_time": "1:11:08", "remaining_time": "1:37:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 332, "total_steps": 780, "loss": 0.3098, "learning_rate": 6.6134688323084884e-06, "epoch": 0.4253342941788774, "percentage": 42.56, "elapsed_time": "1:11:39", "remaining_time": "1:36:41", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 334, "total_steps": 780, "loss": 0.276, "learning_rate": 6.573282724288001e-06, "epoch": 0.42789654896308754, "percentage": 42.82, "elapsed_time": "1:12:02", "remaining_time": "1:36:11", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 336, "total_steps": 780, "loss": 0.2893, "learning_rate": 6.532983499368078e-06, "epoch": 0.4304588037472976, "percentage": 43.08, "elapsed_time": "1:12:30", "remaining_time": "1:35:49", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 338, "total_steps": 780, "loss": 0.2522, "learning_rate": 6.492574055008474e-06, "epoch": 0.43302105853150774, "percentage": 43.33, "elapsed_time": "1:12:55", "remaining_time": "1:35:21", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 340, "total_steps": 780, "loss": 0.2556, "learning_rate": 6.452057296593568e-06, "epoch": 0.4355833133157178, "percentage": 43.59, "elapsed_time": "1:13:26", "remaining_time": "1:35:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 342, "total_steps": 780, "loss": 0.2795, "learning_rate": 6.411436137223479e-06, "epoch": 0.43814556809992794, "percentage": 43.85, "elapsed_time": "1:13:53", "remaining_time": "1:34:37", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 344, "total_steps": 780, "loss": 0.2619, "learning_rate": 6.370713497504607e-06, "epoch": 0.44070782288413807, "percentage": 44.1, "elapsed_time": "1:14:19", "remaining_time": "1:34:12", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 346, "total_steps": 780, "loss": 0.2748, "learning_rate": 6.329892305339659e-06, "epoch": 0.44327007766834814, "percentage": 44.36, "elapsed_time": "1:14:43", "remaining_time": "1:33:43", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 348, "total_steps": 780, "loss": 0.2731, "learning_rate": 6.288975495717124e-06, "epoch": 0.44583233245255827, "percentage": 44.62, "elapsed_time": "1:15:10", "remaining_time": "1:33:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 350, "total_steps": 780, "loss": 0.2797, "learning_rate": 6.247966010500258e-06, "epoch": 0.44839458723676834, "percentage": 44.87, "elapsed_time": "1:15:35", "remaining_time": "1:32:52", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 352, "total_steps": 780, "loss": 0.2724, "learning_rate": 6.206866798215571e-06, "epoch": 0.45095684202097847, "percentage": 45.13, "elapsed_time": "1:16:00", "remaining_time": "1:32:24", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 354, "total_steps": 780, "loss": 0.2728, "learning_rate": 6.165680813840822e-06, "epoch": 0.4535190968051886, "percentage": 45.38, "elapsed_time": "1:16:26", "remaining_time": "1:31:59", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 356, "total_steps": 780, "loss": 0.2733, "learning_rate": 6.124411018592568e-06, "epoch": 0.45608135158939866, "percentage": 45.64, "elapsed_time": "1:16:52", "remaining_time": "1:31:33", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 358, "total_steps": 780, "loss": 0.2688, "learning_rate": 6.0830603797132574e-06, "epoch": 0.4586436063736088, "percentage": 45.9, "elapsed_time": "1:17:16", "remaining_time": "1:31:05", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 360, "total_steps": 780, "loss": 0.2505, "learning_rate": 6.041631870257882e-06, "epoch": 0.46120586115781886, "percentage": 46.15, "elapsed_time": "1:17:42", "remaining_time": "1:30:39", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 362, "total_steps": 780, "loss": 0.2749, "learning_rate": 6.000128468880223e-06, "epoch": 0.463768115942029, "percentage": 46.41, "elapsed_time": "1:18:10", "remaining_time": "1:30:16", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 364, "total_steps": 780, "loss": 0.2541, "learning_rate": 5.958553159618693e-06, "epoch": 0.46633037072623906, "percentage": 46.67, "elapsed_time": "1:18:37", "remaining_time": "1:29:51", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 366, "total_steps": 780, "loss": 0.2721, "learning_rate": 5.916908931681781e-06, "epoch": 0.4688926255104492, "percentage": 46.92, "elapsed_time": "1:19:08", "remaining_time": "1:29:31", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 368, "total_steps": 780, "loss": 0.2774, "learning_rate": 5.8751987792331365e-06, "epoch": 0.4714548802946593, "percentage": 47.18, "elapsed_time": "1:19:33", "remaining_time": "1:29:04", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 370, "total_steps": 780, "loss": 0.2497, "learning_rate": 5.833425701176294e-06, "epoch": 0.4740171350788694, "percentage": 47.44, "elapsed_time": "1:19:59", "remaining_time": "1:28:38", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 372, "total_steps": 780, "loss": 0.2686, "learning_rate": 5.79159270093905e-06, "epoch": 0.4765793898630795, "percentage": 47.69, "elapsed_time": "1:20:25", "remaining_time": "1:28:12", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 374, "total_steps": 780, "loss": 0.2797, "learning_rate": 5.749702786257529e-06, "epoch": 0.4791416446472896, "percentage": 47.95, "elapsed_time": "1:20:52", "remaining_time": "1:27:47", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 376, "total_steps": 780, "loss": 0.2665, "learning_rate": 5.707758968959923e-06, "epoch": 0.4817038994314997, "percentage": 48.21, "elapsed_time": "1:21:17", "remaining_time": "1:27:20", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 378, "total_steps": 780, "loss": 0.2753, "learning_rate": 5.6657642647499545e-06, "epoch": 0.48426615421570984, "percentage": 48.46, "elapsed_time": "1:21:44", "remaining_time": "1:26:55", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 380, "total_steps": 780, "loss": 0.2445, "learning_rate": 5.62372169299004e-06, "epoch": 0.4868284089999199, "percentage": 48.72, "elapsed_time": "1:22:08", "remaining_time": "1:26:28", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 382, "total_steps": 780, "loss": 0.2933, "learning_rate": 5.581634276484211e-06, "epoch": 0.48939066378413004, "percentage": 48.97, "elapsed_time": "1:22:34", "remaining_time": "1:26:01", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 384, "total_steps": 780, "loss": 0.2502, "learning_rate": 5.539505041260779e-06, "epoch": 0.4919529185683401, "percentage": 49.23, "elapsed_time": "1:22:56", "remaining_time": "1:25:32", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 386, "total_steps": 780, "loss": 0.263, "learning_rate": 5.497337016354757e-06, "epoch": 0.49451517335255024, "percentage": 49.49, "elapsed_time": "1:23:24", "remaining_time": "1:25:08", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 388, "total_steps": 780, "loss": 0.2494, "learning_rate": 5.45513323359009e-06, "epoch": 0.49707742813676037, "percentage": 49.74, "elapsed_time": "1:23:49", "remaining_time": "1:24:41", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 390, "total_steps": 780, "loss": 0.2431, "learning_rate": 5.412896727361663e-06, "epoch": 0.49963968292097044, "percentage": 50.0, "elapsed_time": "1:24:11", "remaining_time": "1:24:11", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 392, "total_steps": 780, "loss": 0.248, "learning_rate": 5.370630534417133e-06, "epoch": 0.5022019377051805, "percentage": 50.26, "elapsed_time": "1:24:40", "remaining_time": "1:23:48", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 394, "total_steps": 780, "loss": 0.2522, "learning_rate": 5.328337693638591e-06, "epoch": 0.5047641924893906, "percentage": 50.51, "elapsed_time": "1:25:02", "remaining_time": "1:23:19", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 396, "total_steps": 780, "loss": 0.2856, "learning_rate": 5.286021245824075e-06, "epoch": 0.5073264472736008, "percentage": 50.77, "elapsed_time": "1:25:28", "remaining_time": "1:22:53", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 398, "total_steps": 780, "loss": 0.2626, "learning_rate": 5.243684233468933e-06, "epoch": 0.5098887020578109, "percentage": 51.03, "elapsed_time": "1:25:54", "remaining_time": "1:22:27", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 400, "total_steps": 780, "loss": 0.2738, "learning_rate": 5.201329700547077e-06, "epoch": 0.512450956842021, "percentage": 51.28, "elapsed_time": "1:26:19", "remaining_time": "1:22:00", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 402, "total_steps": 780, "loss": 0.2511, "learning_rate": 5.158960692292122e-06, "epoch": 0.515013211626231, "percentage": 51.54, "elapsed_time": "1:26:48", "remaining_time": "1:21:37", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 404, "total_steps": 780, "loss": 0.2957, "learning_rate": 5.116580254978447e-06, "epoch": 0.5175754664104412, "percentage": 51.79, "elapsed_time": "1:27:14", "remaining_time": "1:21:11", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 406, "total_steps": 780, "loss": 0.2704, "learning_rate": 5.074191435702155e-06, "epoch": 0.5201377211946513, "percentage": 52.05, "elapsed_time": "1:27:39", "remaining_time": "1:20:44", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 408, "total_steps": 780, "loss": 0.3206, "learning_rate": 5.031797282162007e-06, "epoch": 0.5226999759788614, "percentage": 52.31, "elapsed_time": "1:28:05", "remaining_time": "1:20:18", "throughput": "0.00", "total_tokens": 0}
+{"current_steps": 410, "total_steps": 780, "loss": 0.2536, "learning_rate": 4.98940084244029e-06, "epoch": 0.5252622307630715, "percentage": 52.56, "elapsed_time": "1:28:27", "remaining_time": "1:19:49", "throughput": "0.00", "total_tokens": 0}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc69a558b63bc4deebc71791d74e928bce77047a97984154b8bece89dccb197a
+size 5432