try to actually use the GPU
Browse files- handler.py +20 -10
handler.py
CHANGED
@@ -1,27 +1,37 @@
|
|
1 |
from typing import Dict, List, Any
|
2 |
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
|
3 |
-
|
4 |
|
5 |
class EndpointHandler:
|
6 |
def __init__(self, path=""):
|
|
|
|
|
7 |
# load the model
|
8 |
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-1.5B-Instruct")
|
9 |
model = AutoModelForCausalLM.from_pretrained(
|
10 |
"Qwen/Qwen2-1.5B-Instruct",
|
11 |
-
torch_dtype="
|
12 |
device_map="auto"
|
13 |
-
)
|
|
|
14 |
# create inference pipeline
|
15 |
-
self.pipeline = pipeline("text-generation", model=model, tokenizer=tokenizer)
|
16 |
|
17 |
def __call__(self, data: Any) -> List[List[Dict[str, float]]]:
|
18 |
inputs = data.pop("inputs", data)
|
19 |
-
parameters = data.pop("parameters",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
# pass inputs with all kwargs in data
|
22 |
-
|
23 |
-
|
24 |
-
else:
|
25 |
-
prediction = self.pipeline(inputs)
|
26 |
-
# postprocess the prediction
|
27 |
return prediction
|
|
|
1 |
from typing import Dict, List, Any
|
2 |
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
|
3 |
+
import torch
|
4 |
|
5 |
class EndpointHandler:
|
6 |
def __init__(self, path=""):
|
7 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
8 |
+
|
9 |
# load the model
|
10 |
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-1.5B-Instruct")
|
11 |
model = AutoModelForCausalLM.from_pretrained(
|
12 |
"Qwen/Qwen2-1.5B-Instruct",
|
13 |
+
torch_dtype=torch.float16 if device == "cuda" else torch.float32,
|
14 |
device_map="auto"
|
15 |
+
).to(device)
|
16 |
+
|
17 |
# create inference pipeline
|
18 |
+
self.pipeline = pipeline("text-generation", model=model, tokenizer=tokenizer, device=0 if device == "cuda" else -1)
|
19 |
|
20 |
def __call__(self, data: Any) -> List[List[Dict[str, float]]]:
|
21 |
inputs = data.pop("inputs", data)
|
22 |
+
parameters = data.pop("parameters", {})
|
23 |
+
|
24 |
+
# Ensure inputs are on the GPU if available
|
25 |
+
if isinstance(inputs, str):
|
26 |
+
inputs = [inputs]
|
27 |
+
|
28 |
+
# Tensor input handling
|
29 |
+
try:
|
30 |
+
inputs = torch.tensor(inputs).cuda() if torch.cuda.is_available() else torch.tensor(inputs)
|
31 |
+
except:
|
32 |
+
pass # If inputs are not tensors (e.g., strings), continue without conversion
|
33 |
|
34 |
# pass inputs with all kwargs in data
|
35 |
+
prediction = self.pipeline(inputs, **parameters)
|
36 |
+
|
|
|
|
|
|
|
37 |
return prediction
|