no message

main.py

@@ -8,7 +8,7 @@ from typing import Generator
 import json  # Asegúrate de que esta línea esté al principio del archivo
 import nltk
 import os
-from transformers import pipeline
+from transformers import pipeline, BartTokenizer
 
 
 nltk.data.path.append(os.getenv('NLTK_DATA'))
@@ -18,6 +18,10 @@ app = FastAPI()
 # Initialize the InferenceClient with your model
 client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.2")
 
+# Assuming you've initialized the tokenizer and model for BART
+tokenizer = BartTokenizer.from_pretrained('facebook/bart-large')
+summarizer = pipeline("summarization", model="facebook/bart-large")
+
 # summarizer = pipeline("summarization", model="sshleifer/distilbart-cnn-12-6")
 
 summarizer = pipeline("summarization", model="sshleifer/distilbart-cnn-12-6")
@@ -78,26 +82,29 @@ async def generate_text(item: Item):
     # Stream response back to the client
     return StreamingResponse(generate_stream(item), media_type="application/x-ndjson")
 
-def split_text(text, max_size=1000):
-    # Splits the text into chunks of approximately `max_size` words
-    words = text.split()
-    for i in range(0, len(words), max_size):
-        yield ' '.join(words[i:i+max_size])
+
+def split_text_by_tokens(text, max_tokens=1024):
+    # Tokenize the text and split into chunks that fit within the model's max token limit
+    tokens = tokenizer.tokenize(text)
+    for i in range(0, len(tokens), max_tokens):
+        # Convert tokens back to string for summarization
+        yield tokenizer.convert_tokens_to_string(tokens[i:i+max_tokens])
 
 def summarize_large_text(text):
-    chunks = list(split_text(text))
+    # Use the new split_text_by_tokens to handle large text
+    chunks = list(split_text_by_tokens(text))
     summaries = [summarizer(chunk, max_length=500, min_length=100, do_sample=False) for chunk in chunks]
-    combined_summary = ' '.join(sum[0]['summary_text'] for sum in summaries)
+    combined_summary = ' '.join(sum['summary_text'] for sum in summaries)  # Adjusted to handle potential output structure
     return combined_summary
 
 @app.post("/summarize")
 async def summarize_text(request: SummarizeRequest):
     try:
-        # Adjusting summarization for very large texts
         summarized_text = summarize_large_text(request.text)
         return JSONResponse(content={"summary": summarized_text})
     except Exception as e:
-        # Handle exceptions that could arise during summarization
+        # Log the error or provide more detail if needed
+        print(f"Error during summarization: {e}")
         raise HTTPException(status_code=500, detail=str(e))