AlgorithmicResearchGroup/phi-physics

Table of Contents

1. 1.TL;DR
2. 2.Model Details
3. 3.Usage
4. 4.Uses
5. 5.Citation

TL;DR

This is a Phi-1_5 model trained on camel-ai/physics. This model is for research purposes only and *should not be used in production settings*.

Model Description

• —Model type: Language model
• —Language(s) (NLP): English
• —License: Apache 2.0
• —Related Models: Phi-1_5

Usage

Find below some example scripts on how to use the model in transformers:

Using the Pytorch model

from huggingface_hub import notebook_login
from datasets import load_dataset, Dataset
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer

model = "ArtifactAI/phi-physics"

model = AutoModelForCausalLM.from_pretrained(base_model, trust_remote_code= True)
tokenizer = AutoTokenizer.from_pretrained(base_model, trust_remote_code=True)

def generate(prompt):
  inputs = tokenizer(f'''Below is an instruction that describes a task. Write a response that appropriately completes the request If you are adding additional white spaces, stop writing".\n\n### Instruction:\n{prompt}.\n\n### Response:\n ''', return_tensors="pt", return_attention_mask=False)
  streamer = TextStreamer(tokenizer, skip_prompt= True)
  _ = model.generate(**inputs, streamer=streamer, max_new_tokens=500)
  
generate("What are the common techniques used in identifying a new species, and how can scientists accurately categorize it within the existing taxonomy system?")

Training Data

The model was trained on camel-ai/phi-physics, a dataset of question/answer pairs.

Training procedure

The following bitsandbytes quantization config was used during training:

• —quant_method: bitsandbytes
• —loadin8bit: False
• —loadin4bit: True
• —llmint8threshold: 6.0
• —llmint8skip_modules: None
• —llmint8enablefp32cpu_offload: False
• —llmint8hasfp16weight: False
• —bnb4bitquant_type: nf4
• —bnb4bitusedoublequant: True
• —bnb4bitcompute_dtype: float16

Framework versions

• —PEFT 0.6.2

Training procedure

The following bitsandbytes quantization config was used during training:

• —quant_method: bitsandbytes
• —loadin8bit: False
• —loadin4bit: True
• —llmint8threshold: 6.0
• —llmint8skip_modules: None
• —llmint8enablefp32cpu_offload: False
• —llmint8hasfp16weight: False
• —bnb4bitquant_type: nf4
• —bnb4bitusedoublequant: True
• —bnb4bitcompute_dtype: float16

Framework versions

• —PEFT 0.6.2

Citation

@misc{phi-math,
    title={phi-physics},
    author={Matthew Kenney},
    year={2023}
}