Spaces:
Sleeping
Sleeping
Update main.py
Browse files
main.py
CHANGED
|
@@ -1,53 +1,104 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
from fastapi import FastAPI
|
| 2 |
from pydantic import BaseModel
|
| 3 |
-
from
|
| 4 |
-
from fastapi.responses import StreamingResponse
|
| 5 |
-
import torch
|
| 6 |
-
import threading
|
| 7 |
|
| 8 |
app = FastAPI()
|
| 9 |
|
| 10 |
-
#
|
| 11 |
-
|
| 12 |
-
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
| 13 |
-
model = AutoModelForCausalLM.from_pretrained(model_id)
|
| 14 |
|
| 15 |
# Modelo de entrada
|
| 16 |
-
class
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
temperature=0.7,
|
| 39 |
-
top_p=0.9,
|
| 40 |
-
do_sample=True,
|
| 41 |
-
streamer=streamer,
|
| 42 |
-
pad_token_id=tokenizer.eos_token_id
|
| 43 |
-
)
|
| 44 |
-
thread = threading.Thread(target=model.generate, kwargs=generation_kwargs)
|
| 45 |
-
thread.start()
|
| 46 |
-
|
| 47 |
-
# StreamingResponse espera un generador que devuelva texto
|
| 48 |
-
async def event_generator():
|
| 49 |
-
for new_text in streamer:
|
| 50 |
-
yield new_text
|
| 51 |
-
|
| 52 |
-
return StreamingResponse(event_generator(), media_type="text/plain")
|
| 53 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# from fastapi import FastAPI
|
| 2 |
+
# from pydantic import BaseModel
|
| 3 |
+
# from transformers import AutoModelForCausalLM, AutoTokenizer, StoppingCriteria, StoppingCriteriaList
|
| 4 |
+
# import torch
|
| 5 |
+
|
| 6 |
+
# app = FastAPI()
|
| 7 |
+
|
| 8 |
+
# model_id = "HuggingFaceTB/SmolLM2-360M"
|
| 9 |
+
# tokenizer = AutoTokenizer.from_pretrained(model_id)
|
| 10 |
+
# model = AutoModelForCausalLM.from_pretrained(model_id)
|
| 11 |
+
|
| 12 |
+
# class ChatRequest(BaseModel):
|
| 13 |
+
# context: str # Historial de la conversación, como texto
|
| 14 |
+
|
| 15 |
+
# class NewlineStoppingCriteria(StoppingCriteria):
|
| 16 |
+
# def __init__(self, prompt_len, tokenizer):
|
| 17 |
+
# super().__init__()
|
| 18 |
+
# self.prompt_len = prompt_len
|
| 19 |
+
# self.tokenizer = tokenizer
|
| 20 |
+
|
| 21 |
+
# def __call__(self, input_ids, scores, **kwargs):
|
| 22 |
+
# # Chequea si después del prompt hay un token de salto de línea
|
| 23 |
+
# gen_tokens = input_ids[0][self.prompt_len:]
|
| 24 |
+
# gen_text = self.tokenizer.decode(gen_tokens, skip_special_tokens=True)
|
| 25 |
+
# return '\n' in gen_text
|
| 26 |
+
|
| 27 |
+
# @app.post("/chat/demo_base")
|
| 28 |
+
# async def chat_demo_base(request: ChatRequest):
|
| 29 |
+
# prompt = (
|
| 30 |
+
# "Conversacion 1:\n"
|
| 31 |
+
# "-Dauro: -Hola Juanjo.\n"
|
| 32 |
+
# "-Juanjo: -¿Qué tal?\n"
|
| 33 |
+
# "-Dauro: -Bien, ¿y tú?\n\n"
|
| 34 |
+
# "Conversacion 2:\n"
|
| 35 |
+
# "-Juanjo: -Oye Asistente, ¿puedes mirar esto?\n"
|
| 36 |
+
# "-Asistente: -Por supuesto, dime.\n\n"
|
| 37 |
+
# f"Conversacion 3:\n{request.context}\n"
|
| 38 |
+
# )
|
| 39 |
+
|
| 40 |
+
# inputs = tokenizer(prompt, return_tensors="pt")
|
| 41 |
+
# input_ids = inputs["input_ids"]
|
| 42 |
+
# attention_mask = inputs["attention_mask"]
|
| 43 |
+
|
| 44 |
+
# stopping_criteria = StoppingCriteriaList([
|
| 45 |
+
# NewlineStoppingCriteria(prompt_len=input_ids.shape[1], tokenizer=tokenizer)
|
| 46 |
+
# ])
|
| 47 |
+
|
| 48 |
+
# output = model.generate(
|
| 49 |
+
# input_ids=input_ids,
|
| 50 |
+
# attention_mask=attention_mask,
|
| 51 |
+
# max_new_tokens=15,
|
| 52 |
+
# temperature=0.9,
|
| 53 |
+
# top_p=0.8,
|
| 54 |
+
# do_sample=True,
|
| 55 |
+
# pad_token_id=tokenizer.eos_token_id if hasattr(tokenizer, "eos_token_id") else None,
|
| 56 |
+
# stopping_criteria=stopping_criteria,
|
| 57 |
+
# )
|
| 58 |
+
|
| 59 |
+
# generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
|
| 60 |
+
# # Solo el fragmento después del prompt
|
| 61 |
+
# continuation = generated_text[len(prompt):].split('\n')[0]
|
| 62 |
+
|
| 63 |
+
# return {"generated_text": generated_text}
|
| 64 |
from fastapi import FastAPI
|
| 65 |
from pydantic import BaseModel
|
| 66 |
+
from typing import List, Optional
|
|
|
|
|
|
|
|
|
|
| 67 |
|
| 68 |
app = FastAPI()
|
| 69 |
|
| 70 |
+
# Almacenamiento en memoria temporal
|
| 71 |
+
registro_actual = {}
|
|
|
|
|
|
|
| 72 |
|
| 73 |
# Modelo de entrada
|
| 74 |
+
class DialogoEntrada(BaseModel):
|
| 75 |
+
enunciado: str
|
| 76 |
+
personajes: List[str] # lista de 3 personajes
|
| 77 |
+
relato_inicial: str
|
| 78 |
+
final_1: str
|
| 79 |
+
final_2: str
|
| 80 |
+
final_3: str
|
| 81 |
+
|
| 82 |
+
# Modelo de salida
|
| 83 |
+
class DialogoSalida(BaseModel):
|
| 84 |
+
enunciado: str
|
| 85 |
+
personajes: List[str]
|
| 86 |
+
relato_inicial: str
|
| 87 |
+
final_1: str
|
| 88 |
+
final_2: str
|
| 89 |
+
final_3: str
|
| 90 |
+
|
| 91 |
+
@app.post("/entrada")
|
| 92 |
+
async def registrar_dialogo(dialogo: DialogoEntrada):
|
| 93 |
+
global registro_actual
|
| 94 |
+
registro_actual = dialogo.dict() # Sobrescribe el contenido anterior
|
| 95 |
+
return {"status": "registro guardado"}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 96 |
|
| 97 |
+
@app.get("/salida", response_model=Optional[DialogoSalida])
|
| 98 |
+
async def obtener_y_limpiar():
|
| 99 |
+
global registro_actual
|
| 100 |
+
if not registro_actual:
|
| 101 |
+
return None
|
| 102 |
+
salida = registro_actual
|
| 103 |
+
registro_actual = {} # Limpia después de devolver
|
| 104 |
+
return salida
|