Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -36,24 +36,19 @@ if HF_TOKEN:
|
|
36 |
# ========== DEEPSEEK MODEL LOADING ==========
|
37 |
def load_deepseek_model():
|
38 |
"""Load the DeepSeek model with progress tracking"""
|
39 |
-
progress = gr.Progress()
|
40 |
-
progress(0, desc="Loading DeepSeek model...")
|
41 |
-
|
42 |
try:
|
43 |
start_time = time.time()
|
44 |
tokenizer = AutoTokenizer.from_pretrained(
|
45 |
"deepseek-ai/DeepSeek-V3",
|
46 |
trust_remote_code=True
|
47 |
)
|
48 |
-
progress(0.3, desc="Loading tokenizer...")
|
49 |
|
50 |
model = AutoModelForCausalLM.from_pretrained(
|
51 |
"deepseek-ai/DeepSeek-V3",
|
52 |
trust_remote_code=True,
|
53 |
torch_dtype=torch.float16,
|
54 |
-
device_map="auto"
|
55 |
)
|
56 |
-
progress(0.9, desc="Loading model weights...")
|
57 |
|
58 |
load_time = time.time() - start_time
|
59 |
print(f"DeepSeek model loaded in {load_time:.2f} seconds")
|
@@ -1459,13 +1454,17 @@ def create_interface():
|
|
1459 |
outputs=[tabs, nav_message, quiz_alert]
|
1460 |
)
|
1461 |
|
1462 |
-
# Check model loading status
|
1463 |
def check_model_status():
|
1464 |
if model is not None and tokenizer is not None:
|
1465 |
return gr.update(visible=False)
|
1466 |
return gr.update(visible=True)
|
1467 |
|
1468 |
-
app.load(
|
|
|
|
|
|
|
|
|
1469 |
|
1470 |
return app
|
1471 |
|
@@ -1475,4 +1474,4 @@ app = create_interface()
|
|
1475 |
# For Hugging Face Spaces deployment
|
1476 |
if __name__ == "__main__":
|
1477 |
app.launch()
|
1478 |
-
|
|
|
36 |
# ========== DEEPSEEK MODEL LOADING ==========
|
37 |
def load_deepseek_model():
|
38 |
"""Load the DeepSeek model with progress tracking"""
|
|
|
|
|
|
|
39 |
try:
|
40 |
start_time = time.time()
|
41 |
tokenizer = AutoTokenizer.from_pretrained(
|
42 |
"deepseek-ai/DeepSeek-V3",
|
43 |
trust_remote_code=True
|
44 |
)
|
|
|
45 |
|
46 |
model = AutoModelForCausalLM.from_pretrained(
|
47 |
"deepseek-ai/DeepSeek-V3",
|
48 |
trust_remote_code=True,
|
49 |
torch_dtype=torch.float16,
|
50 |
+
device_map="auto" if torch.cuda.is_available() else None
|
51 |
)
|
|
|
52 |
|
53 |
load_time = time.time() - start_time
|
54 |
print(f"DeepSeek model loaded in {load_time:.2f} seconds")
|
|
|
1454 |
outputs=[tabs, nav_message, quiz_alert]
|
1455 |
)
|
1456 |
|
1457 |
+
# Check model loading status
|
1458 |
def check_model_status():
|
1459 |
if model is not None and tokenizer is not None:
|
1460 |
return gr.update(visible=False)
|
1461 |
return gr.update(visible=True)
|
1462 |
|
1463 |
+
app.load(
|
1464 |
+
fn=check_model_status,
|
1465 |
+
inputs=None,
|
1466 |
+
outputs=model_status
|
1467 |
+
)
|
1468 |
|
1469 |
return app
|
1470 |
|
|
|
1474 |
# For Hugging Face Spaces deployment
|
1475 |
if __name__ == "__main__":
|
1476 |
app.launch()
|
1477 |
+
|