akp4u/amit-gc

LoRA Fine-tuned Pythia-410M

This model is a LoRA (Low-Rank Adaptation) fine-tuned version of EleutherAI/pythia-410m.

Model Details

• Base Model: EleutherAI/pythia-410m
• Training Method: LoRA (Low-Rank Adaptation)
• Parameters: ~410M (base) + LoRA adapters
• Memory Efficient: Uses 4-bit quantization for inference

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

# Load tokenizer and base model
tokenizer = AutoTokenizer.from_pretrained("akp4u/amit-gc")
base_model = AutoModelForCausalLM.from_pretrained(
    "EleutherAI/pythia-410m",
    load_in_4bit=True,
    device_map="auto",
    torch_dtype=torch.float16
)

# Load LoRA adapters
model = PeftModel.from_pretrained(base_model, "akp4u/amit-gc")

# Generate text
prompt = "Your prompt here"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(
    **inputs,
    max_new_tokens=100,
    do_sample=True,
    temperature=0.7,
    top_p=0.9
)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

Training

This model was fine-tuned using the LoRA technique on custom documents.

Limitations

• This is a small model (410M parameters) suitable for experimentation
• Generated text quality may vary depending on the training data
• Best used for specific domain tasks related to the training data