Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -28,37 +28,22 @@ async def predict(request: PredictionRequest):
|
|
28 |
logger.info(f"Loading model: {request.model}")
|
29 |
model_path = MODELS[request.model]
|
30 |
|
31 |
-
|
32 |
-
|
33 |
-
model_path,
|
34 |
-
token=HF_TOKEN,
|
35 |
-
)
|
36 |
-
|
37 |
-
model = AutoModelForSeq2SeqLM.from_pretrained(
|
38 |
-
model_path,
|
39 |
-
token=HF_TOKEN,
|
40 |
-
device_map="auto"
|
41 |
-
)
|
42 |
|
43 |
full_input = "Interpret this dream: " + request.inputs
|
44 |
-
logger.info(f"Processing: {full_input}")
|
45 |
|
46 |
inputs = tokenizer(
|
47 |
full_input,
|
48 |
return_tensors="pt",
|
49 |
truncation=True,
|
50 |
-
max_length=512
|
51 |
-
padding=True
|
52 |
-
)
|
53 |
-
|
54 |
-
outputs = model.generate(
|
55 |
-
**inputs,
|
56 |
-
max_length=200,
|
57 |
-
num_beams=4,
|
58 |
-
no_repeat_ngram_size=2
|
59 |
)
|
60 |
|
|
|
61 |
result = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
|
|
62 |
return PredictionResponse(generated_text=result)
|
63 |
|
64 |
except Exception as e:
|
|
|
28 |
logger.info(f"Loading model: {request.model}")
|
29 |
model_path = MODELS[request.model]
|
30 |
|
31 |
+
tokenizer = AutoTokenizer.from_pretrained(model_path, token=HF_TOKEN)
|
32 |
+
model = AutoModelForSeq2SeqLM.from_pretrained(model_path, token=HF_TOKEN)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
full_input = "Interpret this dream: " + request.inputs
|
35 |
+
logger.info(f"Processing input: {full_input}")
|
36 |
|
37 |
inputs = tokenizer(
|
38 |
full_input,
|
39 |
return_tensors="pt",
|
40 |
truncation=True,
|
41 |
+
max_length=512
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
)
|
43 |
|
44 |
+
outputs = model.generate(**inputs, max_length=200)
|
45 |
result = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
46 |
+
|
47 |
return PredictionResponse(generated_text=result)
|
48 |
|
49 |
except Exception as e:
|