README.md

metadata

language:
  - en
license: other
tags:
  - code
datasets:
  - ajibawa-2023/Code-290k-ShareGPT
model-index:
  - name: Code-290k-6.7B-Instruct
    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: 34.9
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ajibawa-2023/Code-290k-6.7B-Instruct
          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: 51.99
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ajibawa-2023/Code-290k-6.7B-Instruct
          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: 34.89
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ajibawa-2023/Code-290k-6.7B-Instruct
          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: 41.95
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ajibawa-2023/Code-290k-6.7B-Instruct
          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: 52.64
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ajibawa-2023/Code-290k-6.7B-Instruct
          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: 3.49
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ajibawa-2023/Code-290k-6.7B-Instruct
          name: Open LLM Leaderboard

Code-290k-6.7B-Instruct

This model is trained on DeepSeek-Coder-6.7B-Instruct. I have used my existing dataset Code-290k-ShareGPT for training purpose. It is trained on around 290000 set of codes. Along with Python, Java, JavaScript, GO, C++, Rust, Ruby, Sql, MySql, R, Julia, Haskell, etc. code with detailed explanation is used for training purpose. This model utilises Alpaca format. Besides code generation it will also give you explanation.

Training:

Entire dataset was trained on 4 x A100 80GB. For 3 epoch, training took 85 hours. DeepSeek-Coder codebase and DeepSpeed was used for training purpose.

This is a full fine tuned model.

Links for quantized models are given below.

Exllama

Exllama v2:Link

Extremely thankful to Bartowski for making Quantized version of the model.

Example Prompt:

This is a conversation with your helpful AI assistant. AI assistant can generate Code in various Programming Languages along with necessary explanation.

### Instruction:
{instruction}

### Response:

You can modify above Prompt as per your requirement. I have used Alpaca format.

I want to say special Thanks to the Open Source community for helping & guiding me to better understand the AI/Model development.

Thank you for your love & support.

Examples

1. Bayes Theorem - Python

2. Fermat's little theorem

3. The Arrhenius equation using R

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	36.64
AI2 Reasoning Challenge (25-Shot)	34.90
HellaSwag (10-Shot)	51.99
MMLU (5-Shot)	34.89
TruthfulQA (0-shot)	41.95
Winogrande (5-shot)	52.64
GSM8k (5-shot)	3.49