Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -1,11 +1,11 @@
|
|
1 |
from fastapi import FastAPI, HTTPException
|
2 |
from pydantic import BaseModel
|
3 |
from llama_cpp import Llama
|
4 |
-
from concurrent.futures import
|
|
|
5 |
import uvicorn
|
6 |
from dotenv import load_dotenv
|
7 |
from difflib import SequenceMatcher
|
8 |
-
from tqdm import tqdm
|
9 |
import threading
|
10 |
|
11 |
load_dotenv()
|
@@ -13,30 +13,23 @@ load_dotenv()
|
|
13 |
app = FastAPI()
|
14 |
|
15 |
# Configuración de los modelos
|
16 |
-
|
17 |
{"repo_id": "Ffftdtd5dtft/gpt2-xl-Q2_K-GGUF", "filename": "gpt2-xl-q2_k.gguf"},
|
18 |
{"repo_id": "Ffftdtd5dtft/Meta-Llama-3.1-8B-Instruct-Q2_K-GGUF", "filename": "meta-llama-3.1-8b-instruct-q2_k.gguf"},
|
19 |
{"repo_id": "Ffftdtd5dtft/gemma-2-9b-it-Q2_K-GGUF", "filename": "gemma-2-9b-it-q2_k.gguf"},
|
20 |
{"repo_id": "Ffftdtd5dtft/gemma-2-27b-Q2_K-GGUF", "filename": "gemma-2-27b-q2_k.gguf"},
|
21 |
]
|
22 |
|
23 |
-
#
|
24 |
def load_model(model_config):
|
25 |
return Llama.from_pretrained(repo_id=model_config['repo_id'], filename=model_config['filename'])
|
26 |
|
27 |
-
# Cargar modelos
|
28 |
def load_all_models():
|
29 |
-
with
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
model = future_to_model[future]
|
34 |
-
try:
|
35 |
-
loaded_models[model['repo_id']] = future.result()
|
36 |
-
print(f"Modelo cargado en RAM: {model['repo_id']}")
|
37 |
-
except Exception as exc:
|
38 |
-
print(f"Error al cargar modelo {model['repo_id']}: {exc}")
|
39 |
-
return loaded_models
|
40 |
|
41 |
# Cargar modelos en memoria
|
42 |
llms = load_all_models()
|
@@ -47,7 +40,7 @@ class ChatRequest(BaseModel):
|
|
47 |
top_p: float = 0.95
|
48 |
temperature: float = 0.7
|
49 |
|
50 |
-
# Función
|
51 |
def generate_chat_response(request, llm):
|
52 |
try:
|
53 |
user_input = normalize_input(request.message)
|
@@ -103,8 +96,10 @@ def filter_by_similarity(responses):
|
|
103 |
break
|
104 |
return best_response
|
105 |
|
106 |
-
def worker_function(llm, request):
|
107 |
-
|
|
|
|
|
108 |
|
109 |
@app.post("/generate_chat")
|
110 |
async def generate_chat(request: ChatRequest):
|
@@ -114,26 +109,28 @@ async def generate_chat(request: ChatRequest):
|
|
114 |
print(f"Procesando solicitud: {request.message}")
|
115 |
|
116 |
responses = []
|
117 |
-
|
118 |
-
|
119 |
-
# Crear
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
|
|
|
|
128 |
|
129 |
# Seleccionar la mejor respuesta
|
130 |
-
best_response = select_best_response(
|
131 |
|
132 |
print(f"Mejor respuesta seleccionada: {best_response}")
|
133 |
|
134 |
return {
|
135 |
"best_response": best_response,
|
136 |
-
"all_responses":
|
137 |
}
|
138 |
|
139 |
if __name__ == "__main__":
|
|
|
1 |
from fastapi import FastAPI, HTTPException
|
2 |
from pydantic import BaseModel
|
3 |
from llama_cpp import Llama
|
4 |
+
from concurrent.futures import ThreadPoolExecutor, as_completed
|
5 |
+
from tqdm import tqdm
|
6 |
import uvicorn
|
7 |
from dotenv import load_dotenv
|
8 |
from difflib import SequenceMatcher
|
|
|
9 |
import threading
|
10 |
|
11 |
load_dotenv()
|
|
|
13 |
app = FastAPI()
|
14 |
|
15 |
# Configuración de los modelos
|
16 |
+
model_configs = [
|
17 |
{"repo_id": "Ffftdtd5dtft/gpt2-xl-Q2_K-GGUF", "filename": "gpt2-xl-q2_k.gguf"},
|
18 |
{"repo_id": "Ffftdtd5dtft/Meta-Llama-3.1-8B-Instruct-Q2_K-GGUF", "filename": "meta-llama-3.1-8b-instruct-q2_k.gguf"},
|
19 |
{"repo_id": "Ffftdtd5dtft/gemma-2-9b-it-Q2_K-GGUF", "filename": "gemma-2-9b-it-q2_k.gguf"},
|
20 |
{"repo_id": "Ffftdtd5dtft/gemma-2-27b-Q2_K-GGUF", "filename": "gemma-2-27b-q2_k.gguf"},
|
21 |
]
|
22 |
|
23 |
+
# Cargar un modelo
|
24 |
def load_model(model_config):
|
25 |
return Llama.from_pretrained(repo_id=model_config['repo_id'], filename=model_config['filename'])
|
26 |
|
27 |
+
# Cargar todos los modelos simultáneamente
|
28 |
def load_all_models():
|
29 |
+
with ThreadPoolExecutor(max_workers=len(model_configs)) as executor:
|
30 |
+
futures = [executor.submit(load_model, config) for config in model_configs]
|
31 |
+
models = [future.result() for future in as_completed(futures)]
|
32 |
+
return models
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
# Cargar modelos en memoria
|
35 |
llms = load_all_models()
|
|
|
40 |
top_p: float = 0.95
|
41 |
temperature: float = 0.7
|
42 |
|
43 |
+
# Función para generar respuestas de chat
|
44 |
def generate_chat_response(request, llm):
|
45 |
try:
|
46 |
user_input = normalize_input(request.message)
|
|
|
96 |
break
|
97 |
return best_response
|
98 |
|
99 |
+
def worker_function(llm, request, progress_bar):
|
100 |
+
response = generate_chat_response(request, llm)
|
101 |
+
progress_bar.update(1)
|
102 |
+
return response
|
103 |
|
104 |
@app.post("/generate_chat")
|
105 |
async def generate_chat(request: ChatRequest):
|
|
|
109 |
print(f"Procesando solicitud: {request.message}")
|
110 |
|
111 |
responses = []
|
112 |
+
num_models = len(llms)
|
113 |
+
|
114 |
+
# Crear barra de progreso
|
115 |
+
with tqdm(total=num_models, desc="Generando respuestas", unit="modelo") as progress_bar:
|
116 |
+
# Ejecutar modelos en paralelo
|
117 |
+
with ThreadPoolExecutor(max_workers=num_models) as executor:
|
118 |
+
futures = [executor.submit(worker_function, llm, request, progress_bar) for llm in llms]
|
119 |
+
for future in as_completed(futures):
|
120 |
+
try:
|
121 |
+
response = future.result()
|
122 |
+
responses.append(response['response'])
|
123 |
+
except Exception as exc:
|
124 |
+
print(f"Error en la generación de respuesta: {exc}")
|
125 |
|
126 |
# Seleccionar la mejor respuesta
|
127 |
+
best_response = select_best_response(responses)
|
128 |
|
129 |
print(f"Mejor respuesta seleccionada: {best_response}")
|
130 |
|
131 |
return {
|
132 |
"best_response": best_response,
|
133 |
+
"all_responses": responses
|
134 |
}
|
135 |
|
136 |
if __name__ == "__main__":
|