Spaces:
Running
Running
revert and fix
Browse files
app.py
CHANGED
@@ -5,13 +5,15 @@ import torch
|
|
5 |
|
6 |
MODEL_NAME = "osmankoc/llama-2-7b-zoa"
|
7 |
|
|
|
|
|
|
|
|
|
|
|
|
|
8 |
# ZeroGPU için model GPU'ya sadece gerektiğinde yüklenecek
|
9 |
@spaces.GPU
|
10 |
def generate(prompt):
|
11 |
-
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
|
12 |
-
model = AutoModelForCausalLM.from_pretrained(
|
13 |
-
MODEL_NAME, torch_dtype=torch.float16, device_map="auto"
|
14 |
-
)
|
15 |
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
|
16 |
output = model.generate(**inputs, max_length=500)
|
17 |
response = tokenizer.decode(output[0], skip_special_tokens=True)
|
|
|
5 |
|
6 |
MODEL_NAME = "osmankoc/llama-2-7b-zoa"
|
7 |
|
8 |
+
# Model ve tokenizer'ı önceden yükle
|
9 |
+
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
|
10 |
+
model = AutoModelForCausalLM.from_pretrained(
|
11 |
+
MODEL_NAME, torch_dtype=torch.float16, device_map="auto"
|
12 |
+
)
|
13 |
+
|
14 |
# ZeroGPU için model GPU'ya sadece gerektiğinde yüklenecek
|
15 |
@spaces.GPU
|
16 |
def generate(prompt):
|
|
|
|
|
|
|
|
|
17 |
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
|
18 |
output = model.generate(**inputs, max_length=500)
|
19 |
response = tokenizer.decode(output[0], skip_special_tokens=True)
|