SmolLM2 Banking LoRA
Fine-tuned SmolLM2 for banking customer service intent classification.
Training details:
- Dataset: atulgupta002/banking_customer_service_query_intent
- LoRA: r=8, alpha=32, target_modules=["q_proj","v_proj"]
- Epochs: 3, batch size: 8, max_seq_length: 192
Usage:
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("username/smollm2-banking-lora")
tokenizer = AutoTokenizer.from_pretrained("username/smollm2-banking-lora")
query = "I forgot my internet banking password"
inputs = tokenizer(f"Instruction: {query}\nResponse:", return_tensors='pt')
generated = model.generate(**inputs, max_new_tokens=16)
print(tokenizer.decode(generated[0], skip_special_tokens=True))
