README.md

metadata

language:
  - tr
  - en
license: apache-2.0
datasets:
  - malhajar/meditron-tr
model-index:
  - name: Mistral-7B-v0.2-meditron-turkish
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: AI2 Reasoning Challenge (25-Shot)
          type: ai2_arc
          config: ARC-Challenge
          split: test
          args:
            num_few_shot: 25
        metrics:
          - type: acc_norm
            value: 59.56
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=malhajar/Mistral-7B-v0.2-meditron-turkish
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: HellaSwag (10-Shot)
          type: hellaswag
          split: validation
          args:
            num_few_shot: 10
        metrics:
          - type: acc_norm
            value: 81.79
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=malhajar/Mistral-7B-v0.2-meditron-turkish
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU (5-Shot)
          type: cais/mmlu
          config: all
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 60.35
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=malhajar/Mistral-7B-v0.2-meditron-turkish
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: TruthfulQA (0-shot)
          type: truthful_qa
          config: multiple_choice
          split: validation
          args:
            num_few_shot: 0
        metrics:
          - type: mc2
            value: 66.19
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=malhajar/Mistral-7B-v0.2-meditron-turkish
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: Winogrande (5-shot)
          type: winogrande
          config: winogrande_xl
          split: validation
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 76.24
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=malhajar/Mistral-7B-v0.2-meditron-turkish
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: GSM8k (5-shot)
          type: gsm8k
          config: main
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 35.94
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=malhajar/Mistral-7B-v0.2-meditron-turkish
          name: Open LLM Leaderboard

Model Card for Model ID

Mistral-7B-v0.2-meditron-turkish is a finetuned Mistral Model version using Freeze technique on Turkish Meditron dataset of malhajar/meditron-7b-tr using SFT Training. This model can answer information about different excplicit ideas in medicine in Turkish and English

Model Description

• Finetuned by: Mohamad Alhajar
• Language(s) (NLP): Turkish,English
• Finetuned from model: mistralai/Mistral-7B-Instruct-v0.2

Prompt Template For Turkish Generation

### Kullancı:

Prompt Template For English Generation

### User:

How to Get Started with the Model

Use the code sample provided in the original post to interact with the model.

from transformers import AutoTokenizer,AutoModelForCausalLM
 
model_id = "malhajar/Mistral-7B-v0.2-meditron-turkish"
model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
                                             device_map="auto",
                                             torch_dtype=torch.float16,
                                             revision="main")

tokenizer = AutoTokenizer.from_pretrained(model_id)

question: "Akciğer kanseri nedir?"
# For generating a response
prompt = '''
### Kullancı:
{question} 
'''
input_ids = tokenizer(prompt, return_tensors="pt").input_ids
output = model.generate(inputs=input_ids,max_new_tokens=512,pad_token_id=tokenizer.eos_token_id,top_k=50, do_sample=True,
        top_p=0.95)
response = tokenizer.decode(output[0])

print(response)

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	63.34
AI2 Reasoning Challenge (25-Shot)	59.56
HellaSwag (10-Shot)	81.79
MMLU (5-Shot)	60.35
TruthfulQA (0-shot)	66.19
Winogrande (5-shot)	76.24
GSM8k (5-shot)	35.94