NeuralTrix-bf16 / README.md
leaderboard-pr-bot's picture
Adding Evaluation Results
e6c8f95 verified
|
raw
history blame
5.12 kB
metadata
license: apache-2.0
tags:
  - merge
  - mergekit
  - lazymergekit
  - bardsai/jaskier-7b-dpo-v3.3
  - CultriX/NeuralTrix-v4-bf16
  - CultriX/NeuralTrix-7B-dpo
base_model:
  - bardsai/jaskier-7b-dpo-v3.3
  - CultriX/NeuralTrix-v4-bf16
  - CultriX/NeuralTrix-7B-dpo
model-index:
  - name: NeuralTrix-bf16
    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: 72.87
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CultriX/NeuralTrix-bf16
          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: 89.12
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CultriX/NeuralTrix-bf16
          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: 64.27
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CultriX/NeuralTrix-bf16
          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: 79.54
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CultriX/NeuralTrix-bf16
          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: 84.61
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CultriX/NeuralTrix-bf16
          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: 67.17
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CultriX/NeuralTrix-bf16
          name: Open LLM Leaderboard

NeuralTrix-bf16

NeuralTrix-bf16 is a merge of the following models using LazyMergekit:

🧩 Configuration

models:
  - model: eren23/dpo-binarized-NeuralTrix-7B
    # no parameters necessary for base model
  - model: bardsai/jaskier-7b-dpo-v3.3
    parameters:
      density: 0.65
      weight: 0.4
  - model: CultriX/NeuralTrix-v4-bf16
    parameters:
      density: 0.6
      weight: 0.35
  - model: CultriX/NeuralTrix-7B-dpo
    parameters:
      density: 0.6
      weight: 0.35
merge_method: dare_ties
base_model: eren23/dpo-binarized-NeuralTrix-7B
parameters:
  int8_mask: true
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "CultriX/"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 76.26
AI2 Reasoning Challenge (25-Shot) 72.87
HellaSwag (10-Shot) 89.12
MMLU (5-Shot) 64.27
TruthfulQA (0-shot) 79.54
Winogrande (5-shot) 84.61
GSM8k (5-shot) 67.17