inference.py

import os
import torch
import json
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, AutoModel, AutoConfig
from peft import LoraConfig, get_peft_model, PeftModel
from araberta_setting.modeling_bilstm_crf import BERT_BiLSTM_CRF
from seq2seq_inference import infer_t5_prompt
from huggingface_hub import hf_hub_download
from openai import OpenAI

# 🔑 OpenAI client (make sure OPENAI_API_KEY is set in Hugging Face Space secrets)
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

# Your fine-tuned OpenAI model IDs
GPT35_FINETUNED = "ft:gpt-3.5-turbo-0125:asma:gpt-3-5-turbo-absa:Bb6gmwkE"
GPT4O_FINETUNED = "ft:gpt-4o-mini-2024-07-18:asma:gpt4-finetune-absa:BazoEjnp"

MODEL_OPTIONS = {
    "Araberta": {
        "base": "asmashayea/absa-araberta",
        "adapter": "asmashayea/absa-araberta"
    },
    "mT5": {
        "base": "google/mt5-base",
        "adapter": "asmashayea/mt4-absa"
    },
    "GPT3.5": {
        "base": "openai",
        "adapter": GPT35_FINETUNED
    },
    # "GPT4o": {
    #     "base": "openai",
    #     "adapter": GPT4O_FINETUNED
    # }
}

cached_models = {}

# ---------------------------
# Araberta loader
# ---------------------------
def load_araberta():
    path = "asmashayea/absa-arabert"
    device = "cuda" if torch.cuda.is_available() else "cpu"

    tokenizer = AutoTokenizer.from_pretrained(path)
    base_model = AutoModel.from_pretrained(path)

    lora_config = LoraConfig.from_pretrained(path)
    lora_model = get_peft_model(base_model, lora_config)

    local_pt = hf_hub_download(repo_id=path, filename="bilstm_crf_head.pt")

    config = AutoConfig.from_pretrained(path)
    model = BERT_BiLSTM_CRF(lora_model, config)

    state_dict = torch.load(local_pt, map_location=torch.device(device))
    model.load_state_dict(state_dict)
    model.to(device).eval()

    cached_models["Araberta"] = (tokenizer, model)
    return tokenizer, model


def infer_araberta(text):
    if "Araberta" not in cached_models:
        tokenizer, model = load_araberta()
    else:
        tokenizer, model = cached_models["Araberta"]

    device = next(model.parameters()).device
    inputs = tokenizer(
        text,
        return_tensors='pt',
        truncation=True,
        padding='max_length',
        max_length=128
    )
    input_ids = inputs['input_ids'].to(device)
    attention_mask = inputs['attention_mask'].to(device)

    with torch.no_grad():
        outputs = model(input_ids=input_ids, attention_mask=attention_mask)
        predicted_ids = outputs['logits'][0].cpu().tolist()

    tokens = tokenizer.convert_ids_to_tokens(input_ids[0].cpu())
    predicted_labels = [model.id2label.get(p, 'O') for p in predicted_ids]

    # Remove special tokens
    clean_tokens = [t for t in tokens if t not in tokenizer.all_special_tokens]
    clean_labels = [l for t, l in zip(tokens, predicted_labels) if t not in tokenizer.all_special_tokens]

    # Build structured token predictions (pretty JSON)
    token_predictions = [
        {"token": t.replace("##", ""), "label": l}
        for t, l in zip(clean_tokens, clean_labels)
    ]

    # Group tokens by aspect spans
    aspects, current_tokens, current_sentiment = [], [], None

    def join_tokens(tok_list):
        """
        Merge WordPiece tokens correctly:
        - Join subwords without spaces
        - Add spaces between separate words
        """
        merged = ""
        for t in tok_list:
            if t.startswith("##"):
                merged += t[2:]
            else:
                if merged:
                    merged += " "
                merged += t
        return merged

    for token, label in zip(clean_tokens, clean_labels):
        if label.startswith("B-"):
            if current_tokens:
                aspects.append({
                    "aspect": join_tokens(current_tokens),
                    "sentiment": current_sentiment
                })
            current_tokens = [token]
            current_sentiment = label.split("-")[1]
        elif label.startswith("I-") and current_sentiment == label.split("-")[1]:
            current_tokens.append(token)
        else:
            if current_tokens:
                aspects.append({
                    "aspect": join_tokens(current_tokens),
                    "sentiment": current_sentiment
                })
                current_tokens, current_sentiment = [], None

    if current_tokens:
        aspects.append({
            "aspect": join_tokens(current_tokens),
            "sentiment": current_sentiment
        })

    return {
        "aspects": aspects,
        # "token_predictions": token_predictions,
    }




# ---------------------------
# Hugging Face seq2seq loaders
# ---------------------------
def load_model(model_key):
    if model_key in cached_models:
        return cached_models[model_key]

    base_id = MODEL_OPTIONS[model_key]["base"]
    adapter_id = MODEL_OPTIONS[model_key]["adapter"]
    device = "cuda" if torch.cuda.is_available() else "cpu"

    tokenizer = AutoTokenizer.from_pretrained(adapter_id)
    base_model = AutoModelForSeq2SeqLM.from_pretrained(base_id).to(device)
    model = PeftModel.from_pretrained(base_model, adapter_id).to(device)
    model.eval()

    cached_models[model_key] = (tokenizer, model)
    return tokenizer, model

# ---------------------------
# OpenAI inference
# ---------------------------
def infer_openai(text, model_name):
    prompt = f"Extract aspects and their sentiment from this review:\n\n{text}\n\nReturn JSON with 'aspect' and 'sentiment'."
    response = client.chat.completions.create(
        model=model_name,
        messages=[{"role": "user", "content": prompt}],
        max_tokens=512,
        temperature=0
    )
    output = response.choices[0].message.content.strip()
    try:
        return json.loads(output)
    except:
        return {"raw_output": output}

# ---------------------------
# Unified predictor
# ---------------------------
def predict_absa(text, model_choice):
    if model_choice in ['mT5', 'mBART']:
        tokenizer, model = load_model(model_choice)
        return infer_t5_prompt(text, tokenizer, model)
    elif model_choice == 'Araberta':
        return infer_araberta(text)
    elif model_choice in ['GPT3.5', 'GPT4o']:
        return infer_openai(text, MODEL_OPTIONS[model_choice]["adapter"])
    else:
        return {"error": f"Model {model_choice} not supported"}