sharmaarush/pytorch_QA_model

Model Card for Model ID

This is a LoRA fine-tuned causal language model trained on a PyTorch Q&A dataset. The base model was adapted using PEFT (Parameter-Efficient Fine-Tuning) with low-rank adapters. It is designed to answer questions related to PyTorch concepts, APIs, and usage examples.

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: Arush Sharma
• Task: Causal Language Modeling (Q&A style)
• Finetuned from model: Qwen/Qwen2.5-3B

Model Sources [optional]

• Repository: https://github.com/Arush04/ML_Clutter/blob/main/PyTorch_Model.ipynb

Uses

• Educational purposes for learning PyTorch
• Assisting developers with PyTorch-related queries
• Small-scale research and experimentation

Training Details

Training Hyperparameters

• Training regime:

• LoRA rank (r): 32

• LoRA alpha: 32

• Dropout: 0.05

• Batch size: 2 (gradient accumulation: 8)

• Epochs: 2

• Optimizer: paged_adamw_8bit

• Precision: FP16

• Dataset: PyTorch Q&A dataset (custom curated) [https://huggingface.co/datasets/sharmaarush/Pytorch_QA]

Evaluation

How to Use

from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from peft import PeftModel
tokenizer = AutoTokenizer.from_pretrained("sharmaarush/pytorch_QA_model")
base_model = AutoModelForCausalLM.from_pretrained(
    model,
    device_map="auto"
)
model = PeftModel.from_pretrained(base_model, "sharmaarush/pytorch_QA_model")
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, device_map="auto")

# Run inference
prompt = "Explain FSDP2 in easier terms"
outputs = pipe(prompt, do_sample=True, temperature=0.7)
print(outputs[0]["generated_text"])