Spaces:
Sleeping
Sleeping
usmansafdarktk
commited on
Commit
·
14ad9da
1
Parent(s):
b854771
Fix main.py: add os import, health endpoint, memory management, use instruction
Browse files
main.py
CHANGED
@@ -15,51 +15,51 @@ app = FastAPI(title="LaMini-LM API",
|
|
15 |
description="API for text generation using LaMini-GPT-774M", version="1.0.0")
|
16 |
|
17 |
# Define request model
|
18 |
-
|
19 |
-
|
20 |
class TextGenerationRequest(BaseModel):
|
21 |
-
|
22 |
max_length: int = 100
|
23 |
temperature: float = 1.0
|
24 |
top_p: float = 0.9
|
25 |
|
26 |
-
|
27 |
# Load model (cached after first load)
|
28 |
try:
|
29 |
logger.info("Loading LaMini-GPT-774M model...")
|
30 |
-
|
31 |
-
generator = pipeline(
|
32 |
-
'text-generation', model='MBZUAI/LaMini-GPT-774M', device=-1)
|
33 |
logger.info("Model loaded successfully.")
|
|
|
|
|
|
|
34 |
except Exception as e:
|
35 |
logger.error(f"Failed to load model: {str(e)}")
|
36 |
-
|
37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
|
39 |
@app.post("/generate")
|
40 |
async def generate_text(request: TextGenerationRequest):
|
41 |
-
|
42 |
-
|
43 |
-
"""
|
44 |
try:
|
45 |
# Validate inputs
|
46 |
-
if not request.
|
47 |
-
raise HTTPException(
|
48 |
-
status_code=400, detail="Prompt cannot be empty")
|
49 |
if request.max_length < 10 or request.max_length > 500:
|
50 |
-
raise HTTPException(
|
51 |
-
status_code=400, detail="max_length must be between 10 and 500")
|
52 |
if request.temperature <= 0 or request.temperature > 2:
|
53 |
-
raise HTTPException(
|
54 |
-
status_code=400, detail="temperature must be between 0 and 2")
|
55 |
if request.top_p <= 0 or request.top_p > 1:
|
56 |
-
raise HTTPException(
|
57 |
-
status_code=400, detail="top_p must be between 0 and 1")
|
58 |
|
59 |
# Generate text
|
60 |
-
logger.info(f"Generating text for
|
61 |
-
wrapper = "Instruction: You are a helpful assistant. Please respond to the following
|
62 |
-
request.
|
63 |
outputs = generator(
|
64 |
wrapper,
|
65 |
max_length=request.max_length,
|
@@ -68,19 +68,13 @@ async def generate_text(request: TextGenerationRequest):
|
|
68 |
num_return_sequences=1,
|
69 |
do_sample=True
|
70 |
)
|
71 |
-
generated_text = outputs[0]['generated_text'].replace(
|
72 |
-
wrapper, "").strip()
|
73 |
-
|
74 |
return {"generated_text": generated_text}
|
75 |
except Exception as e:
|
76 |
logger.error(f"Error during text generation: {str(e)}")
|
77 |
-
raise HTTPException(
|
78 |
-
status_code=500, detail=f"Text generation failed: {str(e)}")
|
79 |
-
|
80 |
|
81 |
-
|
82 |
-
|
83 |
-
""
|
84 |
-
|
85 |
-
"""
|
86 |
-
return {"message": "Welcome to the LaMini-LM API. Use POST /generate to generate text."}
|
|
|
15 |
description="API for text generation using LaMini-GPT-774M", version="1.0.0")
|
16 |
|
17 |
# Define request model
|
|
|
|
|
18 |
class TextGenerationRequest(BaseModel):
|
19 |
+
instruction: str # Changed from prompt for consistency
|
20 |
max_length: int = 100
|
21 |
temperature: float = 1.0
|
22 |
top_p: float = 0.9
|
23 |
|
|
|
24 |
# Load model (cached after first load)
|
25 |
try:
|
26 |
logger.info("Loading LaMini-GPT-774M model...")
|
27 |
+
generator = pipeline('text-generation', model='MBZUAI/LaMini-GPT-774M', device=-1)
|
|
|
|
|
28 |
logger.info("Model loaded successfully.")
|
29 |
+
if torch.cuda.is_available():
|
30 |
+
torch.cuda.empty_cache()
|
31 |
+
gc.collect()
|
32 |
except Exception as e:
|
33 |
logger.error(f"Failed to load model: {str(e)}")
|
34 |
+
generator = None # Allow server to run for health check
|
35 |
|
36 |
+
@app.get("/health")
|
37 |
+
async def health_check():
|
38 |
+
return {"status": "healthy"}
|
39 |
+
|
40 |
+
@app.get("/")
|
41 |
+
async def root():
|
42 |
+
return {"message": "Welcome to the LaMini-LM API. Use POST /generate to generate text."}
|
43 |
|
44 |
@app.post("/generate")
|
45 |
async def generate_text(request: TextGenerationRequest):
|
46 |
+
if generator is None:
|
47 |
+
raise HTTPException(status_code=503, detail="Model not loaded. Check server logs.")
|
|
|
48 |
try:
|
49 |
# Validate inputs
|
50 |
+
if not request.instruction.strip():
|
51 |
+
raise HTTPException(status_code=400, detail="Instruction cannot be empty")
|
|
|
52 |
if request.max_length < 10 or request.max_length > 500:
|
53 |
+
raise HTTPException(status_code=400, detail="max_length must be between 10 and 500")
|
|
|
54 |
if request.temperature <= 0 or request.temperature > 2:
|
55 |
+
raise HTTPException(status_code=400, detail="temperature must be between 0 and 2")
|
|
|
56 |
if request.top_p <= 0 or request.top_p > 1:
|
57 |
+
raise HTTPException(status_code=400, detail="top_p must be between 0 and 1")
|
|
|
58 |
|
59 |
# Generate text
|
60 |
+
logger.info(f"Generating text for instruction: {request.instruction[:50]}...")
|
61 |
+
wrapper = "Instruction: You are a helpful assistant. Please respond to the following instruction.\n\nInstruction: {}\n\nResponse:".format(
|
62 |
+
request.instruction)
|
63 |
outputs = generator(
|
64 |
wrapper,
|
65 |
max_length=request.max_length,
|
|
|
68 |
num_return_sequences=1,
|
69 |
do_sample=True
|
70 |
)
|
71 |
+
generated_text = outputs[0]['generated_text'].replace(wrapper, "").strip()
|
|
|
|
|
72 |
return {"generated_text": generated_text}
|
73 |
except Exception as e:
|
74 |
logger.error(f"Error during text generation: {str(e)}")
|
75 |
+
raise HTTPException(status_code=500, detail=f"Text generation failed: {str(e)}")
|
|
|
|
|
76 |
|
77 |
+
if __name__ == "__main__":
|
78 |
+
import uvicorn
|
79 |
+
port = int(os.environ.get("PORT", 7860))
|
80 |
+
uvicorn.run(app, host="0.0.0.0", port=port)
|
|
|
|