hoelee/stable-diffusion-webui
1
0
Fork 0
mirror of https://github.com/AUTOMATIC1111/stable-diffusion-webui.git synced 2025-08-08 05:12:35 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Simplify grad clip

Browse Source
This commit is contained in:
Muhammad Rizqi Nur
2022-11-05 11:48:38 +07:00
parent 3277f90e93
commit bb832d7725
2 changed files with 14 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
modules/hypernetworks/hypernetwork.py
View File

@@ -385,10 +385,10 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
scheduler = LearnRateScheduler(learn_rate, steps, ititial_step)
clip_grad_mode_value = clip_grad_mode == "value"
clip_grad_mode_norm = clip_grad_mode == "norm"
clip_grad_enabled = clip_grad_mode_value or clip_grad_mode_norm
if clip_grad_enabled:
clip_grad = torch.nn.utils.clip_grad_value_ if clip_grad_mode == "value" else \
torch.nn.utils.clip_grad_norm_ if clip_grad_mode == "norm" else \
None
if clip_grad:
clip_grad_sched = LearnRateScheduler(clip_grad_value, steps, ititial_step, verbose=False)
# dataset loading may take a while, so input validations and early returns should be done before this
@@ -433,7 +433,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
if shared.state.interrupted:
break
if clip_grad_enabled:
if clip_grad:
clip_grad_sched.step(hypernetwork.step)
with torch.autocast("cuda"):
@@ -458,10 +458,8 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
steps_without_grad = 0
assert steps_without_grad < 10, 'no gradient found for the trained weight after backward() for 10 steps in a row; this is a bug; training cannot continue'
if clip_grad_mode_value:
torch.nn.utils.clip_grad_value_(weights, clip_value=clip_grad_sched.learn_rate)
elif clip_grad_mode_norm:
torch.nn.utils.clip_grad_norm_(weights, max_norm=clip_grad_sched.learn_rate)
if clip_grad:
clip_grad(weights, clip_grad_sched.learn_rate)
optimizer.step()