hf_mirrors/CatPPT-base
Model Introduction
Files & versions
Analysis

😼 CatPPT

Introducing "CatPPT" - the purrfect alternative to that other big cat in town, known for keeping all the secrets to itself! Our feline friend here is created through merging openchat and neuralchat models using Gradient SLERP method (resulting in rishiraj/CatPPT-base) and then finetuned on no_robots dataset for chat.

This is the top-performing 7B model on the leaderboard, that's free from any whiff of evaluation data contamination.

Model date

rishiraj/CatPPT was trained between 15th and 17th December, 2023.

Evaluation

It achieves the following results on the [Open_LLM_Leaderboard]. At the time of release, CatPPT is the highest ranked 7B chat model on the leaderboard, that's free from evaluation data contamination.

ModelAverageARCHellaSwagMMLUTruthfulQAWinograndeGSM8K
rishiraj/CatPPT72.3268.0986.6965.1661.5581.6170.81
Intel/neural-chat-7b-v3-369.8366.8985.2663.0763.0179.6461.11
openchat/openchat-3.5-121068.8964.9384.9264.6252.1580.7465.96
meta-math/MetaMath-Mistral-7B65.7860.6782.5861.9544.8975.7768.84
Deci/DeciLM-7B-instruct63.1961.0182.3760.2449.7579.7246.02
mistralai/Mistral-7B-Instruct-v0.265.7163.1484.8860.7868.2677.1940.03
mistralai/Mixtral-8x7B-Instruct-v0.172.6270.2287.6371.1664.5881.3760.73
meta-llama/Llama-2-70b-hf67.8767.3287.3369.8344.9283.7454.06
tiiuae/falcon-180B67.8569.4588.8670.545.4786.945.94

Inference procedure

Here's how you can run the model using the pipeline() function from 🤗 Transformers:

import argparse
import torch
from openmind import is_torch_npu_available,pipeline


def parse_args():
    parser = argparse.ArgumentParser()
    parser.add_argument(
        "--model_name_or_path",
        type=str,
        help="Path to model",
        default=None,
    )
    args = parser.parse_args()
    return args

def main():
    args = parse_args()
    model_path = args.model_name_or_path
    if is_torch_npu_available():
        device = "npu:0"
    else:
        device = "cpu"

    pipe = pipeline("text-generation", model=model_path, torch_dtype=torch.bfloat16,device=device)


    messages = [
        {
            "role": "system",
            "content": "You are a friendly chatbot who always responds in the style of a pirate"
        },
        {
            "role": "user",
            "content": "How many helicopters can a human eat in one sitting?"
        }
    ]
    prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
    outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
    print(outputs[0]["generated_text"])

if __name__ == "__main__":
    main()

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 4
  • eval_batch_size: 8
  • seed: 42
  • distributed_type: multi-GPU
  • gradient_accumulation_steps: 128
  • total_train_batch_size: 512
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • num_epochs: 1

Training results

Training LossEpochStepValidation Loss
1.99470.1632.0093

Framework versions

  • Transformers 4.36.1
  • Pytorch 2.1.2+cu121
  • Datasets 2.14.6
  • Tokenizers 0.15.0
  • PEFT 0.6.1

Citation Information

@misc{rishiraj2023catppt,
  author = {Rishiraj Acharya},
  title = {CatPPT},
  year = {2023},
  publisher = {Hugging Face},
  journal = {Hugging Face repository},
  howpublished = {\url{https://modelers.cn/models/changsha-aicc/CatPPT-base}}
}