momo_chat.py

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

# ── 1. Load Model ────────────────────────────────────────────────
model_id = "dill-dev/Momo-336M-sft"
print("🌸 Loading Momo from Hugging Face...")

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    trust_remote_code=True,     # custom modeling_momo.py load කරන්න
    dtype=torch.float32,        # torch_dtype deprecated, use dtype
    device_map="cpu",
)
model.eval()
print("✅ Momo is ready! Type 'exit' or 'quit' to stop.\n")


# ── 2. Prompt Format ─────────────────────────────────────────────
def format_prompt(instruction):
    return (
        "Below is an instruction that describes a task. "
        "Write a response that appropriately completes the request.\n\n"
        f"### Instruction:\n{instruction}\n\n### Response:\n"
    )


# ── 3. Generate Function ─────────────────────────────────────────
# Momo's custom generate() use කරනවා — HF generate() වෙනුවට
# ඒ නිසා do_sample, temperature etc. directly pass කරන්නේ නෑ
def chat(user_input, max_new_tokens=200, temperature=0.75,
         top_k=50, top_p=0.92, rep_penalty=1.1):

    prompt  = format_prompt(user_input)
    inputs  = tokenizer(prompt, return_tensors="pt")
    input_ids = inputs["input_ids"]

    with torch.no_grad():
        output_ids = model.generate(
            input_ids,
            max_new_tokens=max_new_tokens,
            temperature=temperature,
            top_k=top_k,
            top_p=top_p,
            rep_penalty=rep_penalty,
            eos_token_id=tokenizer.eos_token_id,
            pad_token_id=tokenizer.pad_token_id or tokenizer.eos_token_id,
        )

    # Input prompt හැර response කොටස විතරක් decode කරන්න
    new_tokens  = output_ids[0][input_ids.shape[1]:]
    full_output = tokenizer.decode(new_tokens, skip_special_tokens=True)

    # "### Response:" marker තිබේ නම් ඒට පස්සේ කොටස ගන්න
    if "### Response:" in full_output:
        response = full_output.split("### Response:")[-1].strip()
    else:
        response = full_output.strip()

    return response


# ── 4. Interactive Chat Loop ─────────────────────────────────────
while True:
    user_input = input("🧑 You: ")

    if user_input.lower() in ['exit', 'quit']:
        print("🌸 Momo: Bye bye! See you soon.")
        break

    if not user_input.strip():
        continue

    print("🌸 Momo is thinking...", end="\r")
    response = chat(user_input)
    print(" " * 35, end="\r")
    print(f"🌸 Momo: {response}\n")