README.md

metadata

license: apache-2.0
tags:
  - merge
  - mergekit
  - lazymergekit
  - fblgit/UNA-TheBeagle-7b-v1
  - udkai/Turdus
model-index:
  - name: Marcoroni-7b-DPO-Merge
    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: 73.04
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=nfaheem/Marcoroni-7b-DPO-Merge
          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: 88.8
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=nfaheem/Marcoroni-7b-DPO-Merge
          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: 64.24
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=nfaheem/Marcoroni-7b-DPO-Merge
          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: 70.47
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=nfaheem/Marcoroni-7b-DPO-Merge
          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: 85.24
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=nfaheem/Marcoroni-7b-DPO-Merge
          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: 67.63
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=nfaheem/Marcoroni-7b-DPO-Merge
          name: Open LLM Leaderboard

Marcoroni-7b-DPO-Merge

Marcoroni-7b-DPO-Merge is a merge of the following models using mergekit and inspired by Maxime Labonne's work:

• fblgit/UNA-TheBeagle-7b-v1
• udkai/Turdus

🧩 Configuration

models:
  - model: madatnlp/marcoroni-7b-v3-safetensor
    # no parameters necessary for base model
  - model: fblgit/UNA-TheBeagle-7b-v1
    parameters:
      density: 0.3
      weight: 0.5
  - model: udkai/Turdus
    parameters:
      density: 0.7
      weight: 0.3
merge_method: ties
base_model: madatnlp/marcoroni-7b-v3-safetensor
parameters:
  normalize: true
dtype: float16

💻 Example Python Code

from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

model_name_or_path = "nfaheem/Marcoroni-7b-DPO-Merge"
model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
                                             device_map="auto",
                                             revision="main")

tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)

prompt = "Write a story about llamas"
system_message = "You are a story writing assistant"
prompt_template=f'''{prompt}
'''

print("\n\n*** Generate:")

input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()
output = model.generate(inputs=input_ids, temperature=0.7, do_sample=True, top_p=0.95, top_k=40, max_new_tokens=512)
print(tokenizer.decode(output[0]))

# Inference can also be done using transformers' pipeline

print("*** Pipeline:")
pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_new_tokens=512,
    do_sample=True,
    temperature=0.7,
    top_p=0.95,
    top_k=40,
    repetition_penalty=1.1
)

print(pipe(prompt_template)[0]['generated_text'])

📋 Summary Eval:

Average	ARC	HellaSwag	MMLU	TruthfulQA	Winogrande	GSM8K
74.9	73.04	88.8	64.24	70.47	85.24	67.63

📈 Huggingface Leaderboard

It's Ranked # 1 on HuggingFace Leaderboard among around 13B parameters (01/15/2024)

Model	Average	ARC	HellaSwag	MMLU	Truthful QA	Winogrande	GSM8K
nfaheem/Marcoroni-7b-DPO-Merge	74.9	73.04	88.8	64.24	70.47	85.24	67.63
mlabonne/Beagle14-7b	74.76	72.95	87.95	64.7	68.38	82.64	71.42
udkai/Turdus	74.66	73.38	88.56	64.52	67.11	86.66	67.7
CultriX/MergeTrix-7B	74.33	72.24	87.84	64.88	66.27	83.5	71.19
fblgit/UNA-TheBeagle-7b-v1	73.87	73.04	88	63.48	69.85	82.16	66.72

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	74.90
AI2 Reasoning Challenge (25-Shot)	73.04
HellaSwag (10-Shot)	88.80
MMLU (5-Shot)	64.24
TruthfulQA (0-shot)	70.47
Winogrande (5-shot)	85.24
GSM8k (5-shot)	67.63