mwitiderrick_-_open_llama_3b_code_instruct_0.1-4bits

Quantization made by Richard Erkhov.

Github

Discord

Request more models

open_llama_3b_code_instruct_0.1 - bnb 4bits

• Model creator: https://huggingface.co/mwitiderrick/
• Original model: https://huggingface.co/mwitiderrick/open_llama_3b_code_instruct_0.1/

Original model description:

language:

• en license: apache-2.0 library_name: transformers tags:

• transformers datasets:

• mwitiderrick/AlpacaCode base_model: openlm-research/open_llama_3b inference: true model_type: llama prompt_template: '### Instruction:\n

  {prompt}

  Response:

  ' created_by: mwitiderrick pipeline_tag: text-generation model-index:

• name: mwitiderrick/open_llama_3b_instruct_v_0.2 results:

  • task: type: text-generation dataset: name: hellaswag type: hellaswag metrics:
    • type: hellaswag (0-Shot) value: 0.6581 name: hellaswag(0-Shot)
  • task: type: text-generation dataset: name: winogrande type: winogrande metrics:
    • type: winogrande (0-Shot) value: 0.6267 name: winogrande(0-Shot)
  • task: type: text-generation dataset: name: arc_challenge type: arc_challenge metrics:
    • type: arc_challenge (0-Shot) value: 0.3712 name: arc_challenge(0-Shot) source: url: https://huggingface.co/mwitiderrick/open_llama_3b_instruct_v_0.2 name: open_llama_3b_instruct_v_0.2 model card
  • 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: 41.21 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_code_instruct_0.1 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: 66.96 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_code_instruct_0.1 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: 27.82 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_code_instruct_0.1 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: 35.01 source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_code_instruct_0.1 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: 65.43 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_code_instruct_0.1 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: 1.9 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=mwitiderrick/open_llama_3b_code_instruct_0.1 name: Open LLM Leaderboard

---

OpenLLaMA Code Instruct: An Open Reproduction of LLaMA

This is an OpenLlama model that has been fine-tuned on 1 epoch of the AlpacaCode dataset (122K rows).

Prompt Template

### Instruction:

{query}

### Response:
<Leave new line for model to respond> 

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM,pipeline

tokenizer = AutoTokenizer.from_pretrained("mwitiderrick/open_llama_3b_code_instruct_0.1")
model = AutoModelForCausalLM.from_pretrained("mwitiderrick/open_llama_3b_code_instruct_0.1")
query = "Write a quick sort algorithm in Python"
text_gen = pipeline(task="text-generation", model=model, tokenizer=tokenizer, max_length=200)
output = text_gen(f"### Instruction:\n{query}\n### Response:\n")
print(output[0]['generated_text'])
"""
### Instruction:
write a quick sort algorithm in Python
### Response:
def quick_sort(arr):
    if len(arr) <= 1:
        return arr
    else:
        pivot = arr[len(arr) // 2]
        left = [x for x in arr if x < pivot]
        middle = [x for x in arr if x == pivot]
        right = [x for x in arr if x > pivot]
        return quick_sort(left) + middle + quick_sort(right)

arr = [5,2,4,3,1]
print(quick_sort(arr))
"""
[1, 2, 3, 4, 5]
"""

Metrics

Detailed metrics

|  Tasks   |Version|Filter|n-shot|Metric|Value |   |Stderr|
|----------|-------|------|-----:|------|-----:|---|-----:|
|winogrande|Yaml   |none  |     0|acc   |0.6267|±  |0.0136|
|hellaswag|Yaml   |none  |     0|acc     |0.4962|±  |0.0050|
|         |       |none  |     0|acc_norm|0.6581|±  |0.0047|
|arc_challenge|Yaml   |none  |     0|acc     |0.3481|±  |0.0139|
|             |       |none  |     0|acc_norm|0.3712|±  |0.0141|
|truthfulqa|N/A    |none  |     0|bleu_max   | 24.2580|±  |0.5985|
|          |       |none  |     0|bleu_acc   |  0.2876|±  |0.0003|
|          |       |none  |     0|bleu_diff  | -8.3685|±  |0.6065|
|          |       |none  |     0|rouge1_max | 49.3907|±  |0.7350|
|          |       |none  |     0|rouge1_acc |  0.2558|±  |0.0002|
|          |       |none  |     0|rouge1_diff|-10.6617|±  |0.6450|
|          |       |none  |     0|rouge2_max | 32.4189|±  |0.9587|
|          |       |none  |     0|rouge2_acc |  0.2142|±  |0.0002|
|          |       |none  |     0|rouge2_diff|-12.9903|±  |0.9539|
|          |       |none  |     0|rougeL_max | 46.2337|±  |0.7493|
|          |       |none  |     0|rougeL_acc |  0.2424|±  |0.0002|
|          |       |none  |     0|rougeL_diff|-11.0285|±  |0.6576|
|          |       |none  |     0|acc        |  0.3072|±  |0.0405|

Open LLM Leaderboard Evaluation Results

Detailed results can be found here