diff options
| -rw-r--r-- | modules/extras.py | 28 |
1 files changed, 14 insertions, 14 deletions
diff --git a/modules/extras.py b/modules/extras.py index 90968352..f6704382 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -150,26 +150,26 @@ def run_modelmerger(primary_model_name, secondary_model_name, interp_method, int alpha = alpha * alpha * (3 - (2 * alpha))
return theta0 + ((theta1 - theta0) * alpha)
- 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:
- secondary_model_filename = 'models/' + secondary_model_name + '.ckpt'
-
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:
primary_model_filename = 'models/' + primary_model_name + '.ckpt'
- print(f"Loading {secondary_model_filename}...")
- model_0 = torch.load(secondary_model_filename, map_location='cpu')
+ 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:
+ secondary_model_filename = 'models/' + secondary_model_name + '.ckpt'
print(f"Loading {primary_model_filename}...")
- model_1 = torch.load(primary_model_filename, map_location='cpu')
-
- theta_0 = model_0['state_dict']
- theta_1 = model_1['state_dict']
+ primary_model = torch.load(primary_model_filename, map_location='cpu')
+
+ 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,
@@ -180,7 +180,7 @@ def run_modelmerger(primary_model_name, secondary_model_name, interp_method, int 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:
@@ -188,7 +188,7 @@ def run_modelmerger(primary_model_name, secondary_model_name, interp_method, int output_modelname = 'models/' + primary_model_name + '_' + str(interp_amount) + '-' + secondary_model_name + '_' + str(float(1.0) - interp_amount) + '-' + 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
|