Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -8,17 +8,9 @@ import time
|
|
8 |
|
9 |
model_id = "google/gemma-2-27b-it"
|
10 |
assistant_id = "google/gemma-2-2b-it"
|
11 |
-
torch_device = "cuda" if torch.cuda.is_available() else "cpu"
|
12 |
-
print("Running on device:", torch_device)
|
13 |
-
print("CPU threads:", torch.get_num_threads())
|
14 |
|
15 |
-
|
16 |
-
|
17 |
-
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16, device_map="auto")
|
18 |
-
assistant_model = AutoModelForCausalLM.from_pretrained(assistant_id, torch_dtype=torch.bfloat16, device_map="auto")
|
19 |
-
else:
|
20 |
-
model = AutoModelForCausalLM.from_pretrained(model_id)
|
21 |
-
assistant_model = AutoModelForCausalLM.from_pretrained(assistant_id)
|
22 |
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
23 |
|
24 |
@spaces.GPU
|
|
|
8 |
|
9 |
model_id = "google/gemma-2-27b-it"
|
10 |
assistant_id = "google/gemma-2-2b-it"
|
|
|
|
|
|
|
11 |
|
12 |
+
model = AutoModelForCausalLM.from_pretrained(model_id).to(dtype=torch.bfloat16, device="cuda")
|
13 |
+
assistant_model = AutoModelForCausalLM.from_pretrained(assistant_id).to(dtype=torch.bfloat16, device="cuda")
|
|
|
|
|
|
|
|
|
|
|
14 |
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
15 |
|
16 |
@spaces.GPU
|