Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -10,6 +10,19 @@ import utils
|
|
10 |
from inversion_utils import inversion_forward_process, inversion_reverse_process
|
11 |
|
12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
def randomize_seed_fn(seed, randomize_seed):
|
14 |
if randomize_seed:
|
15 |
seed = random.randint(0, np.iinfo(np.int32).max)
|
@@ -17,7 +30,7 @@ def randomize_seed_fn(seed, randomize_seed):
|
|
17 |
return seed
|
18 |
|
19 |
|
20 |
-
def invert(x0, prompt_src, num_diffusion_steps, cfg_scale_src): # , ldm_stable):
|
21 |
ldm_stable.model.scheduler.set_timesteps(num_diffusion_steps, device=device)
|
22 |
|
23 |
with inference_mode():
|
@@ -34,7 +47,7 @@ def invert(x0, prompt_src, num_diffusion_steps, cfg_scale_src): # , ldm_stable)
|
|
34 |
|
35 |
|
36 |
|
37 |
-
def sample(zs, wts, steps, prompt_tar, tstart, cfg_scale_tar): # , ldm_stable):
|
38 |
# reverse process (via Zs and wT)
|
39 |
tstart = torch.tensor(tstart, dtype=torch.int)
|
40 |
skip = steps - tstart
|
@@ -79,13 +92,22 @@ def edit(input_audio,
|
|
79 |
t_start=90,
|
80 |
randomize_seed=True):
|
81 |
|
82 |
-
global ldm_stable, current_loaded_model
|
83 |
-
print(f'current loaded model: {ldm_stable.model_id}')
|
84 |
-
if model_id != current_loaded_model:
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
|
90 |
# If the inversion was done for a different model, we need to re-run the inversion
|
91 |
if not do_inversion and (saved_inv_model is None or saved_inv_model != model_id):
|
@@ -94,7 +116,7 @@ def edit(input_audio,
|
|
94 |
x0 = utils.load_audio(input_audio, ldm_stable.get_fn_STFT(), device=device)
|
95 |
|
96 |
if do_inversion or randomize_seed: # always re-run inversion
|
97 |
-
zs_tensor, wts_tensor = invert(x0=x0, prompt_src=source_prompt,
|
98 |
num_diffusion_steps=steps,
|
99 |
cfg_scale_src=cfg_scale_src)
|
100 |
wts = gr.State(value=wts_tensor)
|
@@ -105,16 +127,13 @@ def edit(input_audio,
|
|
105 |
# make sure t_start is in the right limit
|
106 |
t_start = change_tstart_range(t_start, steps)
|
107 |
|
108 |
-
output = sample(zs.value, wts.value, steps, prompt_tar=target_prompt, tstart=t_start,
|
109 |
cfg_scale_tar=cfg_scale_tar)
|
110 |
|
111 |
return output, wts, zs, saved_inv_model, do_inversion
|
112 |
|
113 |
|
114 |
-
|
115 |
-
# current_loaded_model = "cvssp/audioldm2-music"
|
116 |
-
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
117 |
-
ldm_stable = load_model(current_loaded_model, device, 200) # deafult model
|
118 |
|
119 |
|
120 |
def get_example():
|
@@ -267,7 +286,7 @@ with gr.Blocks(css='style.css') as demo:
|
|
267 |
input_audio.change(fn=reset_do_inversion, outputs=[do_inversion])
|
268 |
src_prompt.change(fn=reset_do_inversion, outputs=[do_inversion])
|
269 |
model_id.change(fn=reset_do_inversion, outputs=[do_inversion])
|
270 |
-
|
271 |
|
272 |
gr.Examples(
|
273 |
label="Examples",
|
|
|
10 |
from inversion_utils import inversion_forward_process, inversion_reverse_process
|
11 |
|
12 |
|
13 |
+
# current_loaded_model = "cvssp/audioldm2-music"
|
14 |
+
# # current_loaded_model = "cvssp/audioldm2-music"
|
15 |
+
|
16 |
+
# ldm_stable = load_model(current_loaded_model, device, 200) # deafult model
|
17 |
+
LDM2 = "cvssp/audioldm2"
|
18 |
+
MUSIC = "cvssp/audioldm2-music"
|
19 |
+
LDM2_LARGE = "cvssp/audioldm2-large"
|
20 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
21 |
+
ldm2 = load_model(model_id=LDM2, device=device)
|
22 |
+
ldm2_large = load_model(model_id=LDM2_LARGE, device=device)
|
23 |
+
ldm2_music = load_model(model_id= MUSIC, device=device)
|
24 |
+
|
25 |
+
|
26 |
def randomize_seed_fn(seed, randomize_seed):
|
27 |
if randomize_seed:
|
28 |
seed = random.randint(0, np.iinfo(np.int32).max)
|
|
|
30 |
return seed
|
31 |
|
32 |
|
33 |
+
def invert(ldm_stable, x0, prompt_src, num_diffusion_steps, cfg_scale_src): # , ldm_stable):
|
34 |
ldm_stable.model.scheduler.set_timesteps(num_diffusion_steps, device=device)
|
35 |
|
36 |
with inference_mode():
|
|
|
47 |
|
48 |
|
49 |
|
50 |
+
def sample(ldm_stable, zs, wts, steps, prompt_tar, tstart, cfg_scale_tar): # , ldm_stable):
|
51 |
# reverse process (via Zs and wT)
|
52 |
tstart = torch.tensor(tstart, dtype=torch.int)
|
53 |
skip = steps - tstart
|
|
|
92 |
t_start=90,
|
93 |
randomize_seed=True):
|
94 |
|
95 |
+
# global ldm_stable, current_loaded_model
|
96 |
+
# print(f'current loaded model: {ldm_stable.model_id}')
|
97 |
+
# if model_id != current_loaded_model:
|
98 |
+
# print(f'Changing model to {model_id}...')
|
99 |
+
# current_loaded_model = model_id
|
100 |
+
# ldm_stable = None
|
101 |
+
# ldm_stable = load_model(model_id, device)
|
102 |
+
print(model_id)
|
103 |
+
if model_id == LDM2:
|
104 |
+
ldm_stable = ldm2
|
105 |
+
elif model_id == LDM2_LARGE:
|
106 |
+
ldm_stable = ldm2_large
|
107 |
+
else: # MUSIC
|
108 |
+
ldm_stable = ldm2_music
|
109 |
+
|
110 |
+
|
111 |
|
112 |
# If the inversion was done for a different model, we need to re-run the inversion
|
113 |
if not do_inversion and (saved_inv_model is None or saved_inv_model != model_id):
|
|
|
116 |
x0 = utils.load_audio(input_audio, ldm_stable.get_fn_STFT(), device=device)
|
117 |
|
118 |
if do_inversion or randomize_seed: # always re-run inversion
|
119 |
+
zs_tensor, wts_tensor = invert(ldm_stable=ldm_stable, x0=x0, prompt_src=source_prompt,
|
120 |
num_diffusion_steps=steps,
|
121 |
cfg_scale_src=cfg_scale_src)
|
122 |
wts = gr.State(value=wts_tensor)
|
|
|
127 |
# make sure t_start is in the right limit
|
128 |
t_start = change_tstart_range(t_start, steps)
|
129 |
|
130 |
+
output = sample(ldm_stable, zs.value, wts.value, steps, prompt_tar=target_prompt, tstart=t_start,
|
131 |
cfg_scale_tar=cfg_scale_tar)
|
132 |
|
133 |
return output, wts, zs, saved_inv_model, do_inversion
|
134 |
|
135 |
|
136 |
+
|
|
|
|
|
|
|
137 |
|
138 |
|
139 |
def get_example():
|
|
|
286 |
input_audio.change(fn=reset_do_inversion, outputs=[do_inversion])
|
287 |
src_prompt.change(fn=reset_do_inversion, outputs=[do_inversion])
|
288 |
model_id.change(fn=reset_do_inversion, outputs=[do_inversion])
|
289 |
+
steps.change(fn=reset_do_inversion, outputs=[do_inversion])
|
290 |
|
291 |
gr.Examples(
|
292 |
label="Examples",
|