README.md

metadata

language:
  - en
license: other
library_name: peft
tags:
  - text-generation-inference
  - sft
base_model: Qwen/Qwen1.5-1.8B-Chat
model-index:
  - name: finetune_test_qwen15-1-8b-sft-lora
    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: 36.18
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/finetune_test_qwen15-1-8b-sft-lora
          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: 57.77
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/finetune_test_qwen15-1-8b-sft-lora
          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: 44.96
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/finetune_test_qwen15-1-8b-sft-lora
          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: 38
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/finetune_test_qwen15-1-8b-sft-lora
          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: 61.17
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/finetune_test_qwen15-1-8b-sft-lora
          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: 21.53
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=eren23/finetune_test_qwen15-1-8b-sft-lora
          name: Open LLM Leaderboard

Lora sft finetuned version of Qwen/Qwen1.5-1.8B-Chat

from peft import PeftModel, PeftConfig
from transformers import AutoModelForCausalLM

config = PeftConfig.from_pretrained("eren23/finetune_test_qwen15-1-8b-sft")
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen1.5-1.8B-Chat")
model = PeftModel.from_pretrained(model, "eren23/finetune_test_qwen15-1-8b-sft")
model = model.to("cuda")

from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto

# make prediction
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen1.5-1.8B-Chat")

prompt = "Give me a short introduction to large language model."
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(device)

generated_ids = model.generate(
    model_inputs.input_ids,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

Framework versions

• PEFT 0.8.2

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	43.27
AI2 Reasoning Challenge (25-Shot)	36.18
HellaSwag (10-Shot)	57.77
MMLU (5-Shot)	44.96
TruthfulQA (0-shot)	38.00
Winogrande (5-shot)	61.17
GSM8k (5-shot)	21.53