Spaces:
Running
on
Zero
Running
on
Zero
Commit
·
8e9da2c
1
Parent(s):
2d4f4be
Make device optional in load_infinity function; set default to 'cuda' or 'cpu' based on availability and adjust autocast dtype handling
Browse files
app.py
CHANGED
@@ -188,7 +188,7 @@ def load_infinity(
|
|
188 |
model_path='',
|
189 |
scale_schedule=None,
|
190 |
vae=None,
|
191 |
-
device=
|
192 |
model_kwargs=None,
|
193 |
text_channels=2048,
|
194 |
apply_spatial_patchify=0,
|
@@ -196,9 +196,23 @@ def load_infinity(
|
|
196 |
bf16=False,
|
197 |
):
|
198 |
print(f'[Loading Infinity]')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
199 |
text_maxlen = 512
|
200 |
torch.cuda.empty_cache()
|
201 |
-
|
|
|
202 |
infinity_test: Infinity = Infinity(
|
203 |
vae_local=vae, text_channels=text_channels, text_maxlen=text_maxlen,
|
204 |
shared_aln=True, raw_scale_schedule=scale_schedule,
|
@@ -217,6 +231,7 @@ def load_infinity(
|
|
217 |
train_h_div_w_list=[1.0],
|
218 |
**model_kwargs,
|
219 |
).to(device)
|
|
|
220 |
print(f'[you selected Infinity with {model_kwargs=}] model size: {sum(p.numel() for p in infinity_test.parameters())/1e9:.2f}B, bf16={bf16}')
|
221 |
|
222 |
if bf16:
|
@@ -229,7 +244,10 @@ def load_infinity(
|
|
229 |
print(f'[Load Infinity weights]')
|
230 |
state_dict = torch.load(model_path, map_location=device)
|
231 |
print(infinity_test.load_state_dict(state_dict))
|
|
|
|
|
232 |
infinity_test.rng = torch.Generator(device=device)
|
|
|
233 |
return infinity_test
|
234 |
|
235 |
def transform(pil_img, tgt_h, tgt_w):
|
|
|
188 |
model_path='',
|
189 |
scale_schedule=None,
|
190 |
vae=None,
|
191 |
+
device=None, # Make device optional
|
192 |
model_kwargs=None,
|
193 |
text_channels=2048,
|
194 |
apply_spatial_patchify=0,
|
|
|
196 |
bf16=False,
|
197 |
):
|
198 |
print(f'[Loading Infinity]')
|
199 |
+
|
200 |
+
# Set device if not provided
|
201 |
+
if device is None:
|
202 |
+
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
203 |
+
print(f'Using device: {device}')
|
204 |
+
|
205 |
+
# Set autocast dtype based on bf16 and device support
|
206 |
+
if bf16 and device == 'cuda' and torch.cuda.is_bf16_supported():
|
207 |
+
autocast_dtype = torch.bfloat16
|
208 |
+
else:
|
209 |
+
autocast_dtype = torch.float32
|
210 |
+
bf16 = False # Disable bf16 if not supported
|
211 |
+
|
212 |
text_maxlen = 512
|
213 |
torch.cuda.empty_cache()
|
214 |
+
|
215 |
+
with torch.amp.autocast(device_type=device, enabled=bf16, dtype=autocast_dtype, cache_enabled=True), torch.no_grad():
|
216 |
infinity_test: Infinity = Infinity(
|
217 |
vae_local=vae, text_channels=text_channels, text_maxlen=text_maxlen,
|
218 |
shared_aln=True, raw_scale_schedule=scale_schedule,
|
|
|
231 |
train_h_div_w_list=[1.0],
|
232 |
**model_kwargs,
|
233 |
).to(device)
|
234 |
+
|
235 |
print(f'[you selected Infinity with {model_kwargs=}] model size: {sum(p.numel() for p in infinity_test.parameters())/1e9:.2f}B, bf16={bf16}')
|
236 |
|
237 |
if bf16:
|
|
|
244 |
print(f'[Load Infinity weights]')
|
245 |
state_dict = torch.load(model_path, map_location=device)
|
246 |
print(infinity_test.load_state_dict(state_dict))
|
247 |
+
|
248 |
+
# Initialize random number generator on the correct device
|
249 |
infinity_test.rng = torch.Generator(device=device)
|
250 |
+
|
251 |
return infinity_test
|
252 |
|
253 |
def transform(pil_img, tgt_h, tgt_w):
|