diff options
Diffstat (limited to 'modules/extras.py')
| -rw-r--r-- | modules/extras.py | 40 |
1 files changed, 20 insertions, 20 deletions
diff --git a/modules/extras.py b/modules/extras.py index c4ee2b62..b8ebc619 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -140,7 +140,7 @@ def run_pnginfo(image): return '', geninfo, info
-def run_modelmerger(modelname_0, modelname_1, interp_method, interp_amount):
+def run_modelmerger(primary_model_name, secondary_model_name, interp_method, interp_amount):
# Linear interpolation (https://en.wikipedia.org/wiki/Linear_interpolation)
def weighted_sum(theta0, theta1, alpha):
return ((1 - alpha) * theta0) + (alpha * theta1)
@@ -150,26 +150,26 @@ def run_modelmerger(modelname_0, modelname_1, interp_method, interp_amount): alpha = alpha * alpha * (3 - (2 * alpha))
return theta0 + ((theta1 - theta0) * alpha)
- if os.path.exists(modelname_0):
- model0_filename = modelname_0
- modelname_0 = os.path.splitext(os.path.basename(modelname_0))[0]
+ if os.path.exists(primary_model_name):
+ primary_model_filename = primary_model_name
+ primary_model_name = os.path.splitext(os.path.basename(primary_model_name))[0]
else:
- model0_filename = 'models/' + modelname_0 + '.ckpt'
+ primary_model_filename = 'models/' + primary_model_name + '.ckpt'
- if os.path.exists(modelname_1):
- model1_filename = modelname_1
- modelname_1 = os.path.splitext(os.path.basename(modelname_1))[0]
+ if os.path.exists(secondary_model_name):
+ secondary_model_filename = secondary_model_name
+ secondary_model_name = os.path.splitext(os.path.basename(secondary_model_name))[0]
else:
- model1_filename = 'models/' + modelname_1 + '.ckpt'
+ secondary_model_filename = 'models/' + secondary_model_name + '.ckpt'
- print(f"Loading {model0_filename}...")
- model_0 = torch.load(model0_filename, map_location='cpu')
+ print(f"Loading {primary_model_filename}...")
+ primary_model = torch.load(primary_model_filename, map_location='cpu')
- print(f"Loading {model1_filename}...")
- model_1 = torch.load(model1_filename, map_location='cpu')
-
- theta_0 = model_0['state_dict']
- theta_1 = model_1['state_dict']
+ print(f"Loading {secondary_model_filename}...")
+ secondary_model = torch.load(secondary_model_filename, map_location='cpu')
+
+ theta_0 = primary_model['state_dict']
+ theta_1 = secondary_model['state_dict']
theta_funcs = {
"Weighted Sum": weighted_sum,
@@ -180,15 +180,15 @@ def run_modelmerger(modelname_0, modelname_1, interp_method, interp_amount): print(f"Merging...")
for key in tqdm.tqdm(theta_0.keys()):
if 'model' in key and key in theta_1:
- theta_0[key] = theta_func(theta_0[key], theta_1[key], interp_amount)
+ theta_0[key] = theta_func(theta_0[key], theta_1[key], (float(1.0) - interp_amount)) # Need to reverse the interp_amount to match the desired mix ration in the merged checkpoint
for key in theta_1.keys():
if 'model' in key and key not in theta_0:
theta_0[key] = theta_1[key]
- output_modelname = 'models/' + modelname_0 + '-' + modelname_1 + '-' + interp_method.replace(" ", "_") + '-' + str(interp_amount) + '-merged.ckpt'
+ output_modelname = 'models/' + primary_model_name + '_' + str(round(interp_amount,2)) + '-' + secondary_model_name + '_' + str(round((float(1.0) - interp_amount),2)) + '-' + interp_method.replace(" ", "_") + '-merged.ckpt'
print(f"Saving to {output_modelname}...")
- torch.save(model_0, output_modelname)
+ torch.save(primary_model, output_modelname)
print(f"Checkpoint saved.")
- return "Checkpoint saved to " + output_modelname
+ return "Checkpoint saved to " + output_modelname
\ No newline at end of file |