from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained(
    "archit11/qwen-4b-hyperswitch-v1",
    torch_dtype=torch.float16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("archit11/qwen-4b-hyperswitch-v1")

# Generate code
prompt = "use crate::"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_length=150, temperature=0.3)
result = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(result)

Training Details

Base Model: Qwen/Qwen3-4B
Training Data: Hyperswitch repository code
Final Training Loss: 0.4966
Training Progress: 92% complete (excellent results)

This model was fine-tuned on the Hyperswitch codebase to improve code completion and generation for Rust payment processing systems.