joshuasundance/phi3-mini-4k-qlora-python-code-20k-mypo-4k-rfc-pipe

This is a pipeline version of joshuasundance/phi3-mini-4k-qlora-python-code-20k-mypo-4k-rfc

Model Card for Model ID

• Base Model: https://huggingface.co/edumunozsala/phi3-mini-4k-qlora-python-code-20k
• Preference Dataset: https://huggingface.co/datasets/joshuasundance/mypo-4k-rfc
• Training Code: https://gist.github.com/joshuasundance-swca/a94672960733782865932a645587ccdc
• Training Metrics: trainer_state.json

This is an experimental model made by using joshuasundance/mypo-4k-rfc for DPO training of edumunozsala/phi3-mini-4k-qlora-python-code-20k.

The goal is to learn about model training and potentially get the base model to reliably produce Python with type hints. I chose edumunozsala/phi3-mini-4k-qlora-python-code-20k because I was able to train this model in one hour on my laptop.

Model Details

Model Description

This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.

• Developed by: Joshua Sundance Bailey
• Model type: phi 3 qlora DPO
• Language(s) (NLP): English
• License: MIT
• Finetuned from model [optional]: edumunozsala/phi3-mini-4k-qlora-python-code-20k

Model Sources [optional]

• Training Code: https://gist.github.com/joshuasundance-swca/a94672960733782865932a645587ccdc

Uses

For evaluation and testing only. Do not expect great results, and do not use this model for anything important. It has not been evaluated in any way after training.

Direct Use

from transformers import pipeline


pipe = pipeline(
    "text-generation",
    model="joshuasundance/phi3-mini-4k-qlora-python-code-20k-mypo-4k-rfc-pipe",
    trust_remote_code=True,
)


prompt_template = """### Instruction:
Below is an instruction that describes a task. Write a response that appropriately completes the request.

ALWAYS use Python type hints for mypy.

### Instruction:
{instruction}

### Input:
{input}

### Output:
"""


def invoke(user_instruction: str, user_input: str = "") -> str:
    prompt_str = prompt_template.format(instruction=user_instruction, input=user_input)
    prompt = pipe.tokenizer.apply_chat_template(
        [{"role": "user", "content": prompt_str}],
        tokenize=False,
        add_generation_prompt=True,
    )
    outputs = pipe(
        prompt,
        max_new_tokens=256,
        do_sample=True,
        num_beams=1,
        temperature=0.3,
        top_k=50,
        top_p=0.95,
        max_time=180,
    )  # , eos_token_id=eos_token)
    return outputs[0]["generated_text"][len(prompt) :].strip()


user_instruction = (
    "Write a Python function that takes 3 ints, x, y, and z, and returns (x*z)//y."
)
user_input = ""

invoke(user_instruction, user_input)

[More Information Needed]

Downstream Use [optional]

[More Information Needed]

Out-of-Scope Use

[More Information Needed]

Bias, Risks, and Limitations

[More Information Needed]

Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

How to Get Started with the Model

Use the code below to get started with the model.

[More Information Needed]

Training Details

Training Data

• Original qlora: iamtarun/python_code_instructions_18k_alpaca
• DPO: joshuasundance/mypo-4k-rfc

Training Procedure

See training code using peft, transformers, and trl

Preprocessing [optional]

See training code using peft, transformers, and trl

Training Hyperparameters

See training code using peft, transformers, and trl

Speeds, Sizes, Times [optional]

See trainer_state.json in this repo

[More Information Needed]

Evaluation

See trainer_state.json in this repo

Testing Data, Factors & Metrics

Testing Data

20% of DPO dataset (see training code)

[More Information Needed]

Factors

[More Information Needed]

Metrics

[More Information Needed]

Results

[More Information Needed]

Summary

Model Examination [optional]

[More Information Needed]

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

• Hardware Type: [More Information Needed]
• Hours used: [More Information Needed]
• Cloud Provider: [More Information Needed]
• Compute Region: [More Information Needed]
• Carbon Emitted: [More Information Needed]

Technical Specifications [optional]

Model Architecture and Objective

[More Information Needed]

Compute Infrastructure

[More Information Needed]

Hardware

[More Information Needed]

Software

[More Information Needed]

Citation [optional]

BibTeX:

[More Information Needed]

APA:

[More Information Needed]

Glossary [optional]

[More Information Needed]

More Information [optional]

[More Information Needed]

Model Card Authors [optional]

Joshua Sundance Bailey

Model Card Contact

Joshua Sundance Bailey