Spaces:
Sleeping
Sleeping
igashov
commited on
Commit
·
c1152c1
1
Parent(s):
bc1ef42
handle nan values if linker size is small
Browse files- app.py +43 -32
- src/egnn.py +10 -3
app.py
CHANGED
@@ -40,19 +40,6 @@ if not os.path.exists(diffusion_path):
|
|
40 |
ddpm = DDPM.load_from_checkpoint('models/geom_difflinker.ckpt', map_location=device).eval().to(device)
|
41 |
print('Loaded diffusion model')
|
42 |
|
43 |
-
|
44 |
-
def sample_fn(_data):
|
45 |
-
output, _ = size_nn.forward(_data, return_loss=False)
|
46 |
-
probabilities = torch.softmax(output, dim=1)
|
47 |
-
distribution = torch.distributions.Categorical(probs=probabilities)
|
48 |
-
samples = distribution.sample()
|
49 |
-
sizes = []
|
50 |
-
for label in samples.detach().cpu().numpy():
|
51 |
-
sizes.append(size_nn.linker_id2size[label])
|
52 |
-
sizes = torch.tensor(sizes, device=samples.device, dtype=torch.long)
|
53 |
-
return sizes
|
54 |
-
|
55 |
-
|
56 |
def read_molecule_content(path):
|
57 |
with open(path, "r") as f:
|
58 |
return "".join(f.readlines())
|
@@ -72,7 +59,7 @@ def read_molecule(path):
|
|
72 |
|
73 |
def show_input(input_file):
|
74 |
if input_file is None:
|
75 |
-
return ''
|
76 |
if isinstance(input_file, str):
|
77 |
path = input_file
|
78 |
else:
|
@@ -80,15 +67,24 @@ def show_input(input_file):
|
|
80 |
extension = path.split('.')[-1]
|
81 |
if extension not in ['sdf', 'pdb', 'mol', 'mol2']:
|
82 |
msg = output.INVALID_FORMAT_MSG.format(extension=extension)
|
83 |
-
return
|
|
|
|
|
|
|
84 |
|
85 |
try:
|
86 |
molecule = read_molecule_content(path)
|
87 |
except Exception as e:
|
88 |
-
return
|
|
|
|
|
|
|
89 |
|
90 |
html = output.INITIAL_RENDERING_TEMPLATE.format(molecule=molecule, fmt=extension)
|
91 |
-
return
|
|
|
|
|
|
|
92 |
|
93 |
|
94 |
def draw_sample(idx, out_files):
|
@@ -109,7 +105,7 @@ def draw_sample(idx, out_files):
|
|
109 |
return output.IFRAME_TEMPLATE.format(html=html)
|
110 |
|
111 |
|
112 |
-
def generate(input_file, n_steps):
|
113 |
if input_file is None:
|
114 |
return ''
|
115 |
|
@@ -156,6 +152,21 @@ def generate(input_file, n_steps):
|
|
156 |
ddpm.edm.T = n_steps
|
157 |
assert ddpm.center_of_mass == 'fragments'
|
158 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
159 |
for data in dataloader:
|
160 |
chain, node_mask = ddpm.sample_chain(data, sample_fn=sample_fn, keep_frames=1)
|
161 |
print('Generated linker')
|
@@ -208,6 +219,11 @@ with demo:
|
|
208 |
gr.Markdown('Upload the file with 3D-coordinates of the input fragments in .pdb, .mol2 or .sdf format:')
|
209 |
input_file = gr.File(file_count='single', label='Input Fragments')
|
210 |
n_steps = gr.Slider(minimum=10, maximum=500, label="Number of Denoising Steps", step=10)
|
|
|
|
|
|
|
|
|
|
|
211 |
examples = gr.Dataset(
|
212 |
components=[gr.File(visible=False)],
|
213 |
samples=[['examples/example_1.sdf'], ['examples/example_2.sdf']],
|
@@ -235,21 +251,21 @@ with demo:
|
|
235 |
input_file.change(
|
236 |
fn=show_input,
|
237 |
inputs=[input_file],
|
238 |
-
outputs=[visualization],
|
|
|
|
|
|
|
|
|
|
|
239 |
)
|
240 |
examples.click(
|
241 |
-
fn=lambda idx: [
|
242 |
-
f'examples/example_{idx+1}.sdf',
|
243 |
-
10,
|
244 |
-
show_input(f'examples/example_{idx+1}.sdf'),
|
245 |
-
gr.Radio(value='Sample 1', visible=False)
|
246 |
-
],
|
247 |
inputs=[examples],
|
248 |
-
outputs=[input_file, n_steps, visualization, samples]
|
249 |
)
|
250 |
button.click(
|
251 |
fn=generate,
|
252 |
-
inputs=[input_file, n_steps],
|
253 |
outputs=[visualization, output_files, samples],
|
254 |
)
|
255 |
samples.change(
|
@@ -257,10 +273,5 @@ with demo:
|
|
257 |
inputs=[samples, output_files],
|
258 |
outputs=[visualization],
|
259 |
)
|
260 |
-
input_file.clear(
|
261 |
-
fn=lambda: ['', gr.Radio(value='Sample 1', visible=False)],
|
262 |
-
inputs=[],
|
263 |
-
outputs=[visualization, samples],
|
264 |
-
)
|
265 |
|
266 |
demo.launch(server_name=args.ip)
|
|
|
40 |
ddpm = DDPM.load_from_checkpoint('models/geom_difflinker.ckpt', map_location=device).eval().to(device)
|
41 |
print('Loaded diffusion model')
|
42 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
43 |
def read_molecule_content(path):
|
44 |
with open(path, "r") as f:
|
45 |
return "".join(f.readlines())
|
|
|
59 |
|
60 |
def show_input(input_file):
|
61 |
if input_file is None:
|
62 |
+
return ['', gr.Radio.update(visible=False, value='Sample 1')]
|
63 |
if isinstance(input_file, str):
|
64 |
path = input_file
|
65 |
else:
|
|
|
67 |
extension = path.split('.')[-1]
|
68 |
if extension not in ['sdf', 'pdb', 'mol', 'mol2']:
|
69 |
msg = output.INVALID_FORMAT_MSG.format(extension=extension)
|
70 |
+
return [
|
71 |
+
output.IFRAME_TEMPLATE.format(html=msg),
|
72 |
+
gr.Radio.update(visible=False)
|
73 |
+
]
|
74 |
|
75 |
try:
|
76 |
molecule = read_molecule_content(path)
|
77 |
except Exception as e:
|
78 |
+
return [
|
79 |
+
f'Could not read the molecule: {e}',
|
80 |
+
gr.Radio.update(visible=False)
|
81 |
+
]
|
82 |
|
83 |
html = output.INITIAL_RENDERING_TEMPLATE.format(molecule=molecule, fmt=extension)
|
84 |
+
return [
|
85 |
+
output.IFRAME_TEMPLATE.format(html=html),
|
86 |
+
gr.Radio.update(visible=False)
|
87 |
+
]
|
88 |
|
89 |
|
90 |
def draw_sample(idx, out_files):
|
|
|
105 |
return output.IFRAME_TEMPLATE.format(html=html)
|
106 |
|
107 |
|
108 |
+
def generate(input_file, n_steps, n_atoms):
|
109 |
if input_file is None:
|
110 |
return ''
|
111 |
|
|
|
152 |
ddpm.edm.T = n_steps
|
153 |
assert ddpm.center_of_mass == 'fragments'
|
154 |
|
155 |
+
if n_atoms == 0:
|
156 |
+
def sample_fn(_data):
|
157 |
+
out, _ = size_nn.forward(_data, return_loss=False)
|
158 |
+
probabilities = torch.softmax(out, dim=1)
|
159 |
+
distribution = torch.distributions.Categorical(probs=probabilities)
|
160 |
+
samples = distribution.sample()
|
161 |
+
sizes = []
|
162 |
+
for label in samples.detach().cpu().numpy():
|
163 |
+
sizes.append(size_nn.linker_id2size[label])
|
164 |
+
sizes = torch.tensor(sizes, device=samples.device, dtype=torch.long)
|
165 |
+
return sizes
|
166 |
+
else:
|
167 |
+
def sample_fn(_data):
|
168 |
+
return torch.ones(_data['positions'].shape[0], device=device, dtype=torch.long) * n_atoms
|
169 |
+
|
170 |
for data in dataloader:
|
171 |
chain, node_mask = ddpm.sample_chain(data, sample_fn=sample_fn, keep_frames=1)
|
172 |
print('Generated linker')
|
|
|
219 |
gr.Markdown('Upload the file with 3D-coordinates of the input fragments in .pdb, .mol2 or .sdf format:')
|
220 |
input_file = gr.File(file_count='single', label='Input Fragments')
|
221 |
n_steps = gr.Slider(minimum=10, maximum=500, label="Number of Denoising Steps", step=10)
|
222 |
+
n_atoms = gr.Slider(
|
223 |
+
minimum=0, maximum=20,
|
224 |
+
label="Linker Size: DiffLinker will predict it if set to 0",
|
225 |
+
step=1
|
226 |
+
)
|
227 |
examples = gr.Dataset(
|
228 |
components=[gr.File(visible=False)],
|
229 |
samples=[['examples/example_1.sdf'], ['examples/example_2.sdf']],
|
|
|
251 |
input_file.change(
|
252 |
fn=show_input,
|
253 |
inputs=[input_file],
|
254 |
+
outputs=[visualization, samples],
|
255 |
+
)
|
256 |
+
input_file.clear(
|
257 |
+
fn=lambda: [None, '', gr.Radio.update(visible=False)],
|
258 |
+
inputs=[],
|
259 |
+
outputs=[input_file, visualization, samples],
|
260 |
)
|
261 |
examples.click(
|
262 |
+
fn=lambda idx: [f'examples/example_{idx+1}.sdf', 10, 0] + show_input(f'examples/example_{idx+1}.sdf'),
|
|
|
|
|
|
|
|
|
|
|
263 |
inputs=[examples],
|
264 |
+
outputs=[input_file, n_steps, n_atoms, visualization, samples]
|
265 |
)
|
266 |
button.click(
|
267 |
fn=generate,
|
268 |
+
inputs=[input_file, n_steps, n_atoms],
|
269 |
outputs=[visualization, output_files, samples],
|
270 |
)
|
271 |
samples.change(
|
|
|
273 |
inputs=[samples, output_files],
|
274 |
outputs=[visualization],
|
275 |
)
|
|
|
|
|
|
|
|
|
|
|
276 |
|
277 |
demo.launch(server_name=args.ip)
|
src/egnn.py
CHANGED
@@ -421,13 +421,20 @@ class Dynamics(nn.Module):
|
|
421 |
if self.condition_time:
|
422 |
h_final = h_final[:, :-1]
|
423 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
424 |
vel = vel.view(bs, n_nodes, -1) # (B, N, 3)
|
425 |
h_final = h_final.view(bs, n_nodes, -1) # (B, N, D)
|
426 |
node_mask = node_mask.view(bs, n_nodes, 1) # (B, N, 1)
|
427 |
|
428 |
-
if torch.any(torch.isnan(vel)) or torch.any(torch.isnan(h_final)):
|
429 |
-
raise utils.FoundNaNException(vel, h_final)
|
430 |
-
|
431 |
if self.centering:
|
432 |
vel = utils.remove_mean_with_mask(vel, node_mask)
|
433 |
|
|
|
421 |
if self.condition_time:
|
422 |
h_final = h_final[:, :-1]
|
423 |
|
424 |
+
if torch.any(torch.isnan(vel)):
|
425 |
+
print('Found NaN values in velocities')
|
426 |
+
nan_mask = torch.isnan(vel).float()
|
427 |
+
vel = x * nan_mask + torch.nan_to_num(vel) * (1 - nan_mask)
|
428 |
+
|
429 |
+
if torch.any(torch.isnan(h_final)):
|
430 |
+
print('Found NaN values in features')
|
431 |
+
nan_mask = torch.isnan(h_final).float()
|
432 |
+
h_final = h[:, :h_final.shape[1]] * nan_mask + torch.nan_to_num(h_final) * (1 - nan_mask)
|
433 |
+
|
434 |
vel = vel.view(bs, n_nodes, -1) # (B, N, 3)
|
435 |
h_final = h_final.view(bs, n_nodes, -1) # (B, N, D)
|
436 |
node_mask = node_mask.view(bs, n_nodes, 1) # (B, N, 1)
|
437 |
|
|
|
|
|
|
|
438 |
if self.centering:
|
439 |
vel = utils.remove_mean_with_mask(vel, node_mask)
|
440 |
|