adv-optm 1.1.2__tar.gz → 1.1.3__tar.gz

{adv_optm-1.1.2 → adv_optm-1.1.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: adv_optm
-Version: 1.1.2
+Version: 1.1.3
 Summary: A family of highly efficient, lightweight yet powerful optimizers.
 Home-page: https://github.com/Koratahiu/Advanced_Optimizers
 Author: Koratahiu

{adv_optm-1.1.2 → adv_optm-1.1.3}/adv_optm/__init__.py

@@ -16,4 +16,4 @@ __all__ = [
     "Lion_Prodigy_adv",
 ]
 
-__version__ = "1.1.2"
+__version__ = "1.1.3"

{adv_optm-1.1.2 → adv_optm-1.1.3}/adv_optm/optim/AdamW_adv.py

@@ -209,7 +209,7 @@ class AdamW_adv(torch.optim.Optimizer):
         beta1, beta2 = group['betas']
 
         current_step = state['step']
-        if group['kourkoutas_beta']:
+        if group.get('kourkoutas_beta', False):
             # Call prepare_step() once at the beginning of the step for all params
             self.kourkoutas_helper.maybe_prepare_step(current_step)
             # Accumulate current grad's norm for the *next* step
@@ -220,7 +220,7 @@ class AdamW_adv(torch.optim.Optimizer):
         step = state['step'] + 1
         if group['use_bias_correction']:
             bias_correction1 = 1.0 - beta1 ** step
-            if group['kourkoutas_beta']:
+            if group.get('kourkoutas_beta', False):
                 bias_correction2 = 1.0 - group['betas'][1] ** step
                 # Use beta2_max for bias correction
             else:

{adv_optm-1.1.2 → adv_optm-1.1.3}/adv_optm/optim/Adopt_adv.py

@@ -240,7 +240,7 @@ class Adopt_adv(torch.optim.Optimizer):
         beta1, beta2 = group['betas']
 
         current_step = state['step']
-        if group['kourkoutas_beta']:
+        if group.get('kourkoutas_beta', False):
             # Call prepare_step() once at the beginning of the step for all params
             self.kourkoutas_helper.maybe_prepare_step(current_step)
             # Accumulate current grad's norm for the *next* step

{adv_optm-1.1.2 → adv_optm-1.1.3}/adv_optm/optim/Lion_Prodigy_adv.py

@@ -325,7 +325,7 @@ class Lion_Prodigy_adv(torch.optim.Optimizer):
             d_hat = self.d
             if global_d_denom > 0:
                 d_hat = d_coef * global_d_numerator / global_d_denom
-                if g_group['d_limiter']:
+                if g_group.get('d_limiter', False):
                     d_hat = min(self.d * (2 ** 0.25), d_hat)
                 if self.d == g_group['d0']:
                     self.d = max(self.d, d_hat)

{adv_optm-1.1.2 → adv_optm-1.1.3}/adv_optm/optim/Prodigy_adv.py

@@ -304,7 +304,7 @@ class Prodigy_adv(torch.optim.Optimizer):
                 state['p0'] = torch.tensor(0, device=device, dtype=p.dtype)
 
         current_step = state['step']
-        if group['kourkoutas_beta']:
+        if group.get('kourkoutas_beta', False):
             # Call prepare_step() once at the beginning of the step for all params
             self.kourkoutas_helper.maybe_prepare_step(current_step)
             # Accumulate current grad's norm for the *next* step
@@ -515,7 +515,7 @@ class Prodigy_adv(torch.optim.Optimizer):
             d_hat = self.d
             if global_d_denom > 0:
                 d_hat = d_coef * global_d_numerator / global_d_denom
-                if g_group['d_limiter']:
+                if g_group.get('d_limiter', False):
                     d_hat = min(self.d * (2 ** 0.25), d_hat)
                 if self.d == g_group['d0']:
                     self.d = max(self.d, d_hat)

{adv_optm-1.1.2 → adv_optm-1.1.3}/adv_optm/optim/Simplified_AdEMAMix.py

@@ -191,7 +191,7 @@ class Simplified_AdEMAMix(torch.optim.Optimizer):
         beta1_final, beta2 = group["betas"]
 
         current_step = state['step']
-        if group['kourkoutas_beta']:
+        if group.get('kourkoutas_beta', False):
             # Call prepare_step() once at the beginning of the step for all params
             self.kourkoutas_helper.maybe_prepare_step(current_step)
             # Accumulate current grad's norm for the *next* step
@@ -210,7 +210,7 @@ class Simplified_AdEMAMix(torch.optim.Optimizer):
 
         if group['use_bias_correction']:
             state['num_sum'] = beta1 * state['num_sum'] + 1.0
-            if group['kourkoutas_beta']:
+            if group.get('kourkoutas_beta', False):
                 state['den_sum'] = group['betas'][1] * state['den_sum'] + (1.0 - group['betas'][1])
             else:
                 state['den_sum'] = beta2 * state['den_sum'] + (1.0 - beta2)

{adv_optm-1.1.2 → adv_optm-1.1.3}/adv_optm.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: adv_optm
-Version: 1.1.2
+Version: 1.1.3
 Summary: A family of highly efficient, lightweight yet powerful optimizers.
 Home-page: https://github.com/Koratahiu/Advanced_Optimizers
 Author: Koratahiu

{adv_optm-1.1.2 → adv_optm-1.1.3}/setup.py

@@ -5,7 +5,7 @@ with open("README.md", "r", encoding="utf-8") as fh:
 
 setup(
     name="adv_optm",
-    version="1.1.2",
+    version="1.1.3",
     author="Koratahiu",
     author_email="hiuhonor@gmail.com",
     license='Apache 2.0',