Upload main.py
Browse files
main.py
CHANGED
@@ -1,8 +1,10 @@
|
|
|
|
|
|
|
|
1 |
from fastapi import FastAPI, HTTPException
|
|
|
2 |
from pydantic import BaseModel
|
3 |
from transformers import pipeline
|
4 |
-
import uvicorn
|
5 |
-
import os
|
6 |
|
7 |
# Utw贸rz instancj臋 FastAPI
|
8 |
app = FastAPI(
|
@@ -13,7 +15,8 @@ app = FastAPI(
|
|
13 |
|
14 |
# 艢cie偶ka do modelu - Hugging Face automatycznie pobierze model
|
15 |
MODEL_NAME = "speakleash/Bielik-1.5B-v3.0-Instruct"
|
16 |
-
generator = None
|
|
|
17 |
|
18 |
# Model wej艣ciowy dla POST request
|
19 |
class GenerationRequest(BaseModel):
|
@@ -22,6 +25,7 @@ class GenerationRequest(BaseModel):
|
|
22 |
temperature: float = 0.7
|
23 |
top_p: float = 0.9
|
24 |
|
|
|
25 |
@app.on_event("startup")
|
26 |
async def startup_event():
|
27 |
"""
|
@@ -44,7 +48,7 @@ async def startup_event():
|
|
44 |
print(f"B艂膮d 艂adowania modelu: {e}")
|
45 |
# Mo偶esz zdecydowa膰, czy aplikacja ma zako艅czy膰 dzia艂anie, czy kontynuowa膰 bez modelu
|
46 |
# W przypadku b艂臋du 艂adowania modelu, endpoint generacji tekstu b臋dzie zwraca艂 b艂膮d
|
47 |
-
generator = None
|
48 |
|
49 |
|
50 |
@app.get("/")
|
@@ -54,12 +58,12 @@ async def root():
|
|
54 |
"""
|
55 |
return {"message": "Bielik Text Generation API is running!"}
|
56 |
|
|
|
57 |
@app.post("/generate")
|
58 |
async def generate_text(request: GenerationRequest):
|
59 |
"""
|
60 |
Endpoint do generowania tekstu na podstawie promptu.
|
61 |
"""
|
62 |
-
print(request)
|
63 |
if generator is None:
|
64 |
raise HTTPException(status_code=503, detail="Model nie zosta艂 za艂adowany lub wyst膮pi艂 b艂膮d.")
|
65 |
|
@@ -69,13 +73,16 @@ async def generate_text(request: GenerationRequest):
|
|
69 |
max_new_tokens=request.max_new_tokens,
|
70 |
temperature=request.temperature,
|
71 |
top_p=request.top_p,
|
72 |
-
do_sample=True,
|
73 |
)
|
74 |
# Pipeline zwraca list臋 s艂ownik贸w, bierzemy pierwszy wynik
|
75 |
-
|
|
|
|
|
76 |
except Exception as e:
|
77 |
raise HTTPException(status_code=500, detail=f"B艂膮d podczas generowania tekstu: {e}")
|
78 |
|
|
|
79 |
# Uruchamianie serwera Uvicorn bezpo艣rednio (dla Dockera)
|
80 |
if __name__ == "__main__":
|
81 |
-
uvicorn.run(app, host="0.0.0.0", port=int(os.getenv("PORT", 7860)))
|
|
|
1 |
+
import os
|
2 |
+
|
3 |
+
import uvicorn
|
4 |
from fastapi import FastAPI, HTTPException
|
5 |
+
from fastapi.responses import Response
|
6 |
from pydantic import BaseModel
|
7 |
from transformers import pipeline
|
|
|
|
|
8 |
|
9 |
# Utw贸rz instancj臋 FastAPI
|
10 |
app = FastAPI(
|
|
|
15 |
|
16 |
# 艢cie偶ka do modelu - Hugging Face automatycznie pobierze model
|
17 |
MODEL_NAME = "speakleash/Bielik-1.5B-v3.0-Instruct"
|
18 |
+
generator = None # Zostanie za艂adowany p贸藕niej
|
19 |
+
|
20 |
|
21 |
# Model wej艣ciowy dla POST request
|
22 |
class GenerationRequest(BaseModel):
|
|
|
25 |
temperature: float = 0.7
|
26 |
top_p: float = 0.9
|
27 |
|
28 |
+
|
29 |
@app.on_event("startup")
|
30 |
async def startup_event():
|
31 |
"""
|
|
|
48 |
print(f"B艂膮d 艂adowania modelu: {e}")
|
49 |
# Mo偶esz zdecydowa膰, czy aplikacja ma zako艅czy膰 dzia艂anie, czy kontynuowa膰 bez modelu
|
50 |
# W przypadku b艂臋du 艂adowania modelu, endpoint generacji tekstu b臋dzie zwraca艂 b艂膮d
|
51 |
+
generator = None # Ustaw na None, aby sygnalizowa膰 problem
|
52 |
|
53 |
|
54 |
@app.get("/")
|
|
|
58 |
"""
|
59 |
return {"message": "Bielik Text Generation API is running!"}
|
60 |
|
61 |
+
|
62 |
@app.post("/generate")
|
63 |
async def generate_text(request: GenerationRequest):
|
64 |
"""
|
65 |
Endpoint do generowania tekstu na podstawie promptu.
|
66 |
"""
|
|
|
67 |
if generator is None:
|
68 |
raise HTTPException(status_code=503, detail="Model nie zosta艂 za艂adowany lub wyst膮pi艂 b艂膮d.")
|
69 |
|
|
|
73 |
max_new_tokens=request.max_new_tokens,
|
74 |
temperature=request.temperature,
|
75 |
top_p=request.top_p,
|
76 |
+
do_sample=True, # Wa偶ne dla generowania z temperatur膮
|
77 |
)
|
78 |
# Pipeline zwraca list臋 s艂ownik贸w, bierzemy pierwszy wynik
|
79 |
+
#gen_text = {"generated_text": generated_text[0]["generated_text"]}
|
80 |
+
return Response(content=generated_text[0]["generated_text"], media_type="text/plain; charset=utf-8")
|
81 |
+
# return {"generated_text": generated_text[0]["generated_text"]}
|
82 |
except Exception as e:
|
83 |
raise HTTPException(status_code=500, detail=f"B艂膮d podczas generowania tekstu: {e}")
|
84 |
|
85 |
+
|
86 |
# Uruchamianie serwera Uvicorn bezpo艣rednio (dla Dockera)
|
87 |
if __name__ == "__main__":
|
88 |
+
uvicorn.run(app, host="0.0.0.0", port=int(os.getenv("PORT", 7860)))
|