README.md

metadata

language:
  - en
license: mit
tags:
  - chatml
  - mistral
  - instruct
  - openhermes
  - economics
datasets:
  - rxavier/economicus
base_model: teknium/OpenHermes-2.5-Mistral-7B
model-index:
  - name: Taurus-7B-1.0
    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: 63.57
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=rxavier/Taurus-7B-1.0
          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: 83.64
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=rxavier/Taurus-7B-1.0
          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: 63.5
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=rxavier/Taurus-7B-1.0
          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: 50.21
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=rxavier/Taurus-7B-1.0
          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: 78.14
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=rxavier/Taurus-7B-1.0
          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: 59.36
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=rxavier/Taurus-7B-1.0
          name: Open LLM Leaderboard
library_name: transformers

Taurus 7B 1.0

Description

Taurus is an OpenHermes 2.5 finetune using the Economicus dataset, an instruct dataset synthetically generated from Economics PhD textbooks.

The model was trained for 2 epochs (QLoRA) using axolotl. The exact config I used can be found here.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	66.40
AI2 Reasoning Challenge (25-Shot)	63.57
HellaSwag (10-Shot)	83.64
MMLU (5-Shot)	63.50
TruthfulQA (0-shot)	50.21
Winogrande (5-shot)	78.14
GSM8k (5-shot)	59.36

Prompt format

Taurus uses ChatML.

<|im_start|>system
System message
<|im_start|>user
User message<|im_end|>
<|im_start|>assistant

Usage

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig


model_id = "rxavier/Taurus-7B-1.0"
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16, #torch.float16 for older GPUs
    device_map="auto", # Requires having accelerate installed, useful in places like Colab with limited VRAM
)
tokenizer = AutoTokenizer.from_pretrained(model_id)
generation_config = GenerationConfig(
                bos_token_id=tokenizer.bos_token_id,
                eos_token_id=tokenizer.eos_token_id,
                pad_token_id=tokenizer.pad_token_id,
            )

system_message = "You are an expert in economics with PhD level knowledge. You are helpful, give thorough and clear explanations, and use equations and formulas where needed."
prompt = "Give me latex formulas for extended euler equations"

messages = [{"role": "system",
             "content": system_message},
            {"role": "user",
             "content": prompt}]
tokens = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt").to("cuda")

with torch.no_grad():
    outputs = model.generate(inputs=tokens, generation_config=generation_config, max_length=512)
print(tokenizer.decode(outputs.cpu().tolist()[0]))

GGUF quants

You can find GGUF quants for llama.cpp here.