README.md

metadata

language:
  - fr
  - en
license: other
library_name: transformers
tags:
  - chat
  - qwen
  - qwen2.5
  - finetune
  - french
  - english
base_model: Qwen/Qwen2.5-3B
datasets:
  - MaziyarPanahi/french_instruct_sharegpt
  - arcee-ai/EvolKit-20k
model_name: calme-3.2-instruct-3b
license_name: qwen-research
license_link: https://huggingface.co./Qwen/Qwen2.5-3B/blob/main/LICENSE
pipeline_tag: text-generation
inference: false
model_creator: MaziyarPanahi
quantized_by: MaziyarPanahi
model-index:
  - name: calme-3.2-instruct-3b
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: IFEval (0-Shot)
          type: HuggingFaceH4/ifeval
          args:
            num_few_shot: 0
        metrics:
          - type: inst_level_strict_acc and prompt_level_strict_acc
            value: 55.33
            name: strict accuracy
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/calme-3.2-instruct-3b
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: BBH (3-Shot)
          type: BBH
          args:
            num_few_shot: 3
        metrics:
          - type: acc_norm
            value: 27.98
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/calme-3.2-instruct-3b
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MATH Lvl 5 (4-Shot)
          type: hendrycks/competition_math
          args:
            num_few_shot: 4
        metrics:
          - type: exact_match
            value: 9.89
            name: exact match
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/calme-3.2-instruct-3b
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: GPQA (0-shot)
          type: Idavidrein/gpqa
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 4.47
            name: acc_norm
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/calme-3.2-instruct-3b
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MuSR (0-shot)
          type: TAUR-Lab/MuSR
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 8.79
            name: acc_norm
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/calme-3.2-instruct-3b
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU-PRO (5-shot)
          type: TIGER-Lab/MMLU-Pro
          config: main
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 29.48
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/open-llm-leaderboard/open_llm_leaderboard?query=MaziyarPanahi/calme-3.2-instruct-3b
          name: Open LLM Leaderboard

This is avery small model, so it might not perform well for some prompts and may be sensitive to hyper parameters. I would appreciate any feedback to see if I can fix any issues in the next iteration. ❤️

MaziyarPanahi/calme-3.2-instruct-3b

This model is an advanced iteration of the powerful Qwen/Qwen2.5-3B, specifically fine-tuned to enhance its capabilities in generic domains.

⚡ Quantized GGUF

All GGUF models are available here: MaziyarPanahi/calme-3.2-instruct-3b-GGUF

🏆 Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	22.66
IFEval (0-Shot)	55.33
BBH (3-Shot)	27.98
MATH Lvl 5 (4-Shot)	9.89
GPQA (0-shot)	4.47
MuSR (0-shot)	8.79
MMLU-PRO (5-shot)	29.48

Prompt Template

This model uses ChatML prompt template:

<|im_start|>system
{System}
<|im_end|>
<|im_start|>user
{User}
<|im_end|>
<|im_start|>assistant
{Assistant}

How to use

# Use a pipeline as a high-level helper

from transformers import pipeline

messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe = pipeline("text-generation", model="MaziyarPanahi/calme-3.2-instruct-3b")
pipe(messages)


# Load model directly

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("MaziyarPanahi/calme-3.2-instruct-3b")
model = AutoModelForCausalLM.from_pretrained("MaziyarPanahi/calme-3.2-instruct-3b")

Ethical Considerations

As with any large language model, users should be aware of potential biases and limitations. We recommend implementing appropriate safeguards and human oversight when deploying this model in production environments.