Spaces:
Running
on
Zero
Running
on
Zero
Change spaces.gpu application
Browse files
app.py
CHANGED
@@ -26,19 +26,20 @@ with open("token_probabilities.json") as f:
|
|
26 |
token_probs_dict = json.load(f)
|
27 |
token_probabilities = np.array([token_probs_dict[str(i)] for i in range(len(token_probs_dict))], dtype=np.float32)
|
28 |
|
29 |
-
|
30 |
-
def load_weights():
|
31 |
-
# OK: download & load weights to CPU
|
32 |
ckpt_path = hf_hub_download(
|
33 |
repo_id="ruurd/tini_model",
|
34 |
filename="diffusion-model.pth",
|
35 |
token=os.getenv("HF_TOKEN")
|
36 |
)
|
37 |
-
return torch.load(ckpt_path, map_location="cpu") # ✅ returns only CPU tensors
|
38 |
|
39 |
-
|
40 |
-
model.
|
41 |
-
model
|
|
|
|
|
|
|
|
|
42 |
|
43 |
rng = np.random.default_rng()
|
44 |
|
@@ -82,6 +83,8 @@ def generate_diffusion_text(input_ids, answer_start):
|
|
82 |
return input_ids[:answer_start] + sampled[answer_start:]
|
83 |
|
84 |
# --- Inference Wrapper ---
|
|
|
|
|
85 |
def diffusion_chat(question, eot_weight, max_it, sharpness):
|
86 |
placeholder = "What do you know about the city of New York?"
|
87 |
if question.strip() == "":
|
@@ -144,6 +147,10 @@ def diffusion_chat(question, eot_weight, max_it, sharpness):
|
|
144 |
|
145 |
# --- Gradio Interface ---
|
146 |
|
|
|
|
|
|
|
|
|
147 |
demo = gr.Interface(
|
148 |
fn=diffusion_chat,
|
149 |
inputs=[
|
|
|
26 |
token_probs_dict = json.load(f)
|
27 |
token_probabilities = np.array([token_probs_dict[str(i)] for i in range(len(token_probs_dict))], dtype=np.float32)
|
28 |
|
29 |
+
def load_model():
|
|
|
|
|
30 |
ckpt_path = hf_hub_download(
|
31 |
repo_id="ruurd/tini_model",
|
32 |
filename="diffusion-model.pth",
|
33 |
token=os.getenv("HF_TOKEN")
|
34 |
)
|
|
|
35 |
|
36 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
37 |
+
model = torch.load(ckpt_path, map_location=device)
|
38 |
+
model = disable_dropout(model)
|
39 |
+
model.to(device)
|
40 |
+
model.eval()
|
41 |
+
return model
|
42 |
+
|
43 |
|
44 |
rng = np.random.default_rng()
|
45 |
|
|
|
83 |
return input_ids[:answer_start] + sampled[answer_start:]
|
84 |
|
85 |
# --- Inference Wrapper ---
|
86 |
+
|
87 |
+
@spaces.GPU
|
88 |
def diffusion_chat(question, eot_weight, max_it, sharpness):
|
89 |
placeholder = "What do you know about the city of New York?"
|
90 |
if question.strip() == "":
|
|
|
147 |
|
148 |
# --- Gradio Interface ---
|
149 |
|
150 |
+
print("Loading model...")
|
151 |
+
model = load_model()
|
152 |
+
print("✅ Model loaded.")
|
153 |
+
|
154 |
demo = gr.Interface(
|
155 |
fn=diffusion_chat,
|
156 |
inputs=[
|