app.py

import streamlit as st
import torch
import pandas as pd
import PyPDF2
import pickle
import os
from transformers import AutoTokenizer, PreTrainedModel, PretrainedConfig
from huggingface_hub import login, hf_hub_download
import time
from ch09util import subsequent_mask, create_model

# Device setup
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

# Set page configuration
st.set_page_config(
    page_title="Translator Agent",
    page_icon="🚀",
    layout="centered"
)

# Model repository name
MODEL_NAME = "amiguel/custom-en2fr-transformer-v1"

# Retrieve Hugging Face token from environment variable
HF_TOKEN = os.environ.get("HF_TOKEN")
if not HF_TOKEN:
    st.error("🔐 Hugging Face token not found in environment variables. Please set HF_TOKEN in Space secrets.")
    st.stop()

# Title with rocket emojis
st.title("🚀 English to French Translator 🚀")

# Configure Avatars
USER_AVATAR = "https://raw.githubusercontent.com/achilela/vila_fofoka_analysis/9904d9a0d445ab0488cf7395cb863cce7621d897/USER_AVATAR.png"
BOT_AVATAR = "https://raw.githubusercontent.com/achilela/vila_fofoka_analysis/991f4c6e4e1dc7a8e24876ca5aae5228bcdb4dba/Ataliba_Avatar.jpg"

# Sidebar configuration (removed token input)
with st.sidebar:
    st.header("Upload Documents 📂")
    uploaded_file = st.file_uploader(
        "Choose a PDF or XLSX file to translate",
        type=["pdf", "xlsx"],
        label_visibility="collapsed"
    )

# Initialize chat history
if "messages" not in st.session_state:
    st.session_state.messages = []

# File processing function
@st.cache_data
def process_file(uploaded_file):
    if uploaded_file is None:
        return ""
    
    try:
        if uploaded_file.type == "application/pdf":
            pdf_reader = PyPDF2.PdfReader(uploaded_file)
            return "\n".join([page.extract_text() for page in pdf_reader.pages])
        elif uploaded_file.type == "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet":
            df = pd.read_excel(uploaded_file)
            return df.to_markdown()
    except Exception as e:
        st.error(f"📄 Error processing file: {str(e)}")
        return ""

# Custom model loading function
@st.cache_resource
def load_model_and_resources():
    try:
        login(token=HF_TOKEN)
        
        # Load tokenizer from the model repo
        tokenizer = AutoTokenizer.from_pretrained(
            MODEL_NAME,
            token=HF_TOKEN
        )
        
        # Define Transformer configuration
        class TransformerConfig(PretrainedConfig):
            model_type = "custom_transformer"
            def __init__(self, src_vocab_size=11055, tgt_vocab_size=11239, d_model=256, d_ff=1024, h=8, N=6, dropout=0.1, **kwargs):
                super().__init__(**kwargs)
                self.src_vocab_size = src_vocab_size
                self.tgt_vocab_size = tgt_vocab_size
                self.d_model = d_model
                self.d_ff = d_ff
                self.h = h
                self.N = N
                self.dropout = dropout

        # Define Transformer model
        class CustomTransformer(PreTrainedModel):
            config_class = TransformerConfig
            def __init__(self, config):
                super().__init__(config)
                self.model = create_model(
                    config.src_vocab_size,
                    config.tgt_vocab_size,
                    N=config.N,
                    d_model=config.d_model,
                    d_ff=config.d_ff,
                    h=config.h,
                    dropout=config.dropout
                )
            def forward(self, src, tgt, src_mask, tgt_mask, **kwargs):
                return self.model(src, tgt, src_mask, tgt_mask)

        # Load config with validation from the model repo
        config_dict = TransformerConfig.from_pretrained(MODEL_NAME, token=HF_TOKEN).to_dict()
        if "src_vocab_size" not in config_dict or "tgt_vocab_size" not in config_dict:
            st.warning(
                f"Config at {MODEL_NAME}/config.json is missing 'src_vocab_size' or 'tgt_vocab_size'. "
                "Using defaults (11055, 11239). For accuracy, update the training script to save these values."
            )
            config = TransformerConfig()
        else:
            config = TransformerConfig(**config_dict)

        # Initialize model on meta device and load weights explicitly
        model = CustomTransformer(config)
        weights_path = hf_hub_download(repo_id=MODEL_NAME, filename="model.safetensors", token=HF_TOKEN)
        from safetensors.torch import load_file
        state_dict = load_file(weights_path)
        model.load_state_dict(state_dict)
        
        # Move model to the target device safely
        if DEVICE == "cuda":
            model = model.to_empty(device=DEVICE)  # Move structure to GPU
            model.load_state_dict(state_dict)      # Reload weights on GPU
        else:
            model = model.to(DEVICE)  # CPU can handle direct move after loading weights
        
        model.eval()

        # Load dictionaries from the model repo
        dict_path = hf_hub_download(repo_id=MODEL_NAME, filename="dict.p", token=HF_TOKEN)
        with open(dict_path, "rb") as fb:
            en_word_dict, en_idx_dict, fr_word_dict, fr_idx_dict = pickle.load(fb)
        
        return model, tokenizer, en_word_dict, fr_word_dict, en_idx_dict, fr_idx_dict
        
    except Exception as e:
        st.error(f"🤖 Model loading failed: {str(e)}")
        return None

# Custom streaming generation function
def custom_streaming_generate(input_text, model, tokenizer, en_word_dict, fr_word_dict, fr_idx_dict):
    try:
        model.eval()
        PAD, UNK = 0, 1
        tokenized_en = ["BOS"] + tokenizer.tokenize(input_text) + ["EOS"]
        enidx = [en_word_dict.get(i, UNK) for i in tokenized_en]
        src = torch.tensor(enidx).long().to(DEVICE).unsqueeze(0)
        src_mask = (src != 0).unsqueeze(-2)
        memory = model.model.encode(src, src_mask)
        start_symbol = fr_word_dict["BOS"]
        ys = torch.ones(1, 1).fill_(start_symbol).type_as(src.data)
        for _ in range(100):
            out = model.model.decode(memory, src_mask, ys, subsequent_mask(ys.size(1)).type_as(src.data))
            prob = model.model.generator(out[:, -1])
            _, next_word = torch.max(prob, dim=1)
            next_word = next_word.data[0]
            sym = fr_idx_dict.get(next_word, "UNK")
            if sym != "EOS":
                token = sym.replace("</w>", " ")
                for x in '''?:;.,'("-!&)%''':
                    token = token.replace(f" {x}", f"{x}")
                yield token
            else:
                break
            ys = torch.cat([ys, torch.ones(1, 1).type_as(src.data).fill_(next_word)], dim=1)
        # Yield a final empty token to ensure completion
        yield ""
        
    except Exception as e:
        raise Exception(f"Generation error: {str(e)}")

# Display chat messages
for message in st.session_state.messages:
    try:
        avatar = USER_AVATAR if message["role"] == "user" else BOT_AVATAR
        with st.chat_message(message["role"], avatar=avatar):
            st.markdown(message["content"])
    except:
        with st.chat_message(message["role"]):
            st.markdown(message["content"])

# Chat input handling
if prompt := st.chat_input("Enter text to translate into French..."):
    # Load model and resources if not already loaded
    if "model" not in st.session_state:
        model_data = load_model_and_resources()
        if model_data is None:
            st.error("Failed to load model. Please check the HF_TOKEN in Space secrets and try again.")
            st.stop()
            
        st.session_state.model, st.session_state.tokenizer, \
        st.session_state.en_word_dict, st.session_state.fr_word_dict, \
        st.session_state.en_idx_dict, st.session_state.fr_idx_dict = model_data
    
    model = st.session_state.model
    tokenizer = st.session_state.tokenizer
    en_word_dict = st.session_state.en_word_dict
    fr_word_dict = st.session_state.fr_word_dict
    fr_idx_dict = st.session_state.fr_idx_dict
    
    # Add user message
    with st.chat_message("user", avatar=USER_AVATAR):
        st.markdown(prompt)
    st.session_state.messages.append({"role": "user", "content": prompt})

    # Process file or use prompt directly
    file_context = process_file(uploaded_file)
    input_text = file_context if file_context else prompt
    
    # Generate translation with streaming
    if model and tokenizer:
        try:
            with st.chat_message("assistant", avatar=BOT_AVATAR):
                start_time = time.time()
                
                # Create a placeholder for streaming output
                response_container = st.empty()
                full_response = ""
                
                # Stream translation tokens
                for token in custom_streaming_generate(
                    input_text, model, tokenizer, en_word_dict, fr_word_dict, fr_idx_dict
                ):
                    if token:  # Only append non-empty tokens
                        full_response += token
                        response_container.markdown(full_response)
                
                # Calculate performance metrics
                end_time = time.time()
                input_tokens = len(tokenizer(input_text)["input_ids"])
                output_tokens = len(tokenizer(full_response)["input_ids"])
                speed = output_tokens / (end_time - start_time) if (end_time - start_time) > 0 else 0
                
                # Calculate costs (hypothetical pricing model)
                input_cost = (input_tokens / 1000000) * 5  # $5 per million input tokens
                output_cost = (output_tokens / 1000000) * 15  # $15 per million output tokens
                total_cost_usd = input_cost + output_cost
                total_cost_aoa = total_cost_usd * 1160  # Convert to AOA (Angolan Kwanza)
                
                # Display metrics
                st.caption(
                    f"🤖 Input Tokens: {input_tokens} | Output Tokens: {output_tokens} | "
                    f"🕒 Speed: {speed:.1f}t/s | 💰 Cost (USD): ${total_cost_usd:.4f} | "
                    f"💵 Cost (AOA): {total_cost_aoa:.4f}"
                )
                
                # Store the full response in chat history
                st.session_state.messages.append({"role": "assistant", "content": full_response})
                
        except Exception as e:
            st.error(f"⚡ Translation error: {str(e)}")
    else:
        st.error("🤖 Model not loaded!")