Phi3.5-Kaveri-Coder-lora

Fine-tuned coding assistant based on microsoft/Phi-3.5-mini-instruct trained on iamtarun/python_code_instructions_18k_alpaca using Kaveri Studio.

• Developer / Organization: Shiva Technologies
• Base Model: microsoft/Phi-3.5-mini-instruct
• Dataset: iamtarun/python_code_instructions_18k_alpaca
• Adapter Type: LoRA (PEFT)
• Target Hardware: Free Kaggle 2x NVIDIA T4 (16GB VRAM) / Google Colab
• Live Dispatch: Dispatched live via Kaveri Studio UI E2E flow

Quick Inference

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

base_model = "microsoft/Phi-3.5-mini-instruct"
adapter_repo = "shiva123782/Phi3.5-Kaveri-Coder-lora"

tokenizer = AutoTokenizer.from_pretrained(base_model, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(base_model, torch_dtype=torch.float16, device_map="auto", trust_remote_code=True)
# model = PeftModel.from_pretrained(model, adapter_repo)

messages = [
    {"role": "system", "content": "You are a coding assistant developed by Shiva Technologies."},
    {"role": "user", "content": "Write a python function to find primes using a sieve."}
]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=128)
print(tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))