Spaces:
Sleeping
Sleeping
afnotes commited on
Commit ·
23342bb
1
Parent(s): e93f31c
Added spaces.GPU (correction)
Browse files
app.py
CHANGED
|
@@ -10,20 +10,9 @@ import os
|
|
| 10 |
BASE_MODEL_ID = "Qwen/Qwen3-4B"
|
| 11 |
ADAPTER_MODEL_ID = "afnotes/Qwen3-4B-YeI-LoRA"
|
| 12 |
|
| 13 |
-
@spaces.GPU
|
| 14 |
-
# Add this GPU check here at the module level
|
| 15 |
-
if not torch.cuda.is_available():
|
| 16 |
-
print("!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!")
|
| 17 |
-
print("WARNING: CUDA (GPU) is NOT available on this Space instance.")
|
| 18 |
-
print("!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!")
|
| 19 |
-
# We will attempt to load on CPU if GPU is not available for now.
|
| 20 |
-
device_to_load = "cpu"
|
| 21 |
-
else:
|
| 22 |
-
print(f"CUDA is available! PyTorch detected GPU: {torch.cuda.get_device_name(0)}")
|
| 23 |
-
device_to_load = "cuda"
|
| 24 |
-
|
| 25 |
# --- Model Loading (Optimized for Space) ---
|
| 26 |
# This function will be called once when the app starts
|
|
|
|
| 27 |
def load_model():
|
| 28 |
print(f"Loading tokenizer from: {ADAPTER_MODEL_ID}")
|
| 29 |
tokenizer = AutoTokenizer.from_pretrained(ADAPTER_MODEL_ID)
|
|
|
|
| 10 |
BASE_MODEL_ID = "Qwen/Qwen3-4B"
|
| 11 |
ADAPTER_MODEL_ID = "afnotes/Qwen3-4B-YeI-LoRA"
|
| 12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
# --- Model Loading (Optimized for Space) ---
|
| 14 |
# This function will be called once when the app starts
|
| 15 |
+
@spaces.GPU
|
| 16 |
def load_model():
|
| 17 |
print(f"Loading tokenizer from: {ADAPTER_MODEL_ID}")
|
| 18 |
tokenizer = AutoTokenizer.from_pretrained(ADAPTER_MODEL_ID)
|