adv-optm 1.0.0__py3-none-any.whl → 1.0.3__py3-none-any.whl

adv_optm/optim/Simplified_AdEMAMix.py

@@ -46,8 +46,8 @@ class Simplified_AdEMAMix(torch.optim.Optimizer):
             matrices to apply low-rank compression (default: True).
         stochastic_rounding (bool): whether to use stochastic
             rounding for BF16 parameter updates (default: True).
-        use_orthograd (bool): whether to use OrthoGrad. (default: False)
-        factored (bool): whether to use the factorization or disable it to use
+        orthogonal_gradient (bool): whether to use OrthoGrad. (default: False)
+        nnmf_factor (bool): whether to use the factorization or disable it to use
             the uncompressed optimizer. (default: False)
     """
 
@@ -64,8 +64,8 @@ class Simplified_AdEMAMix(torch.optim.Optimizer):
         use_bias_correction: bool = True,
         vector_reshape: bool = True,
         stochastic_rounding: bool = True,
-        use_orthograd: bool = False,
-        factored: bool = False,
+        orthogonal_gradient: bool = False,
+        nnmf_factor: bool = False,
     ):
         if not (lr >= 0.0):
             raise ValueError(f"Learning-rate should be >= 0.0. Got {lr}")
@@ -82,10 +82,10 @@ class Simplified_AdEMAMix(torch.optim.Optimizer):
             "lr": lr, "betas": betas, "eps": eps, "weight_decay": weight_decay,
             "alpha_grad": alpha_grad, "beta1_warmup": beta1_warmup, "min_beta1": min_beta1,
             "vector_reshape": vector_reshape,
-            "use_orthograd": use_orthograd, "use_bias_correction": use_bias_correction,
+            "orthogonal_gradient": orthogonal_gradient, "use_bias_correction": use_bias_correction,
         }
         self.stochastic_rounding = stochastic_rounding
-        self.factored = factored
+        self.factored = nnmf_factor
         super().__init__(params, defaults)
 
     @property
@@ -108,7 +108,7 @@ class Simplified_AdEMAMix(torch.optim.Optimizer):
         grad = p.grad
         if grad.dtype != torch.float32 and self.factored:
             grad = grad.float()
-        if group["use_orthograd"]:
+        if group["orthogonal_gradient"]:
             grad = _orthogonalize_gradient(p, grad)
         state = self.state[p]
 

{adv_optm-1.0.0.dist-info → adv_optm-1.0.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: adv_optm
-Version: 1.0.0
+Version: 1.0.3
 Summary: A family of highly efficient, lightweight yet powerful optimizers.
 Home-page: https://github.com/Koratahiu/Advanced_Optimizers
 Author: Koratahiu

adv_optm-1.0.3.dist-info/RECORD

@@ -0,0 +1,19 @@
+adv_optm/__init__.py,sha256=zL7hnbPAHt7w-0fZQld04Pt58F-aYaRAqz15e-RZh-Y,306
+adv_optm/optim/AdamW_adv.py,sha256=aTuYcJgd_EcZOrs6TDgBrBKw3wtU5LPzE5WvTBDDeEo,14317
+adv_optm/optim/Adopt_adv.py,sha256=lElmraSiIZiGu9W6ELXnIPZNEEYi1ZWuvuemgPZOixk,17484
+adv_optm/optim/Lion_Prodigy_adv.py,sha256=sGzhts9a6gHfCkuHTB5L9IrClo4c6UThzYYErBwqOaA,12844
+adv_optm/optim/Lion_adv.py,sha256=6G1CukJB_pC7l9HwFEuY1ydsNHZFabVmOvcHDsHHVuQ,8295
+adv_optm/optim/Prodigy_adv.py,sha256=8XUpu19BaBmHb-R9K3jgwySDbtVaLU1_Drtttc_zITs,22461
+adv_optm/optim/Simplified_AdEMAMix.py,sha256=tb3d6Cw_nGwcTzYUhDnKqyP7GzjD1hn8k4WqGG5lhmw,9813
+adv_optm/optim/__init__.py,sha256=pcP865H2j1tut2VfTUhzQh7V8TF_tzPjqFnjMfFed2k,382
+adv_optm/util/BF16_Stochastic_Rounding.py,sha256=Q5H0BcogmE4atP65dLoI21HKSf50lRdsBDfeF6v9Tbg,1548
+adv_optm/util/Effective_Shape.py,sha256=TBvIk1V8IuTbbBsxuekJA4e_v8JlR5Nujtut8RTWAm4,318
+adv_optm/util/NNMF.py,sha256=yRf5IP5Sjq0Uf0DxN0Q8NxEGSdD-f1ULziLVDOjY8K4,639
+adv_optm/util/One_Bit_Boolean.py,sha256=Wat49esdwohuN-OHOFMW8D0aOQgV9cP5Rl8z6yfmpos,1068
+adv_optm/util/OrthoGrad.py,sha256=NzInuBQGy_Ja__M1R9XbvqVaQ0fhGbtGgFE9YON7B3I,707
+adv_optm/util/__init__.py,sha256=qoyIF0jcLjs_vSEcsv36clw5LFNBEbifyXrrVxMH-G4,349
+adv_optm-1.0.3.dist-info/licenses/LICENSE,sha256=HrhfyXIkWY2tGFK11kg7vPCqhgh5DcxleloqdhrpyMY,11558
+adv_optm-1.0.3.dist-info/METADATA,sha256=Cx9bqS9VFt2nBey-H7GxVS0AXwNzTy0eW5NtSW6uXKk,8422
+adv_optm-1.0.3.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+adv_optm-1.0.3.dist-info/top_level.txt,sha256=iNfBIIzu-lPrQ7jyC56WBCcbkRwitM2nJ15-MRQ_6fg,9
+adv_optm-1.0.3.dist-info/RECORD,,

adv_optm-1.0.0.dist-info/RECORD

@@ -1,19 +0,0 @@
-adv_optm/__init__.py,sha256=2qu7FpN7btJdmkvyT4uVs8XOIrvSkyZdOf2pHm1JWRU,306
-adv_optm/optim/AdamW_adv.py,sha256=Pu0TB14dOhcq9kwXclMIeKCI6ef_P0emwzxPu6xuBM0,14252
-adv_optm/optim/Adopt_adv.py,sha256=71o9BHV3XFefJX21G37PKG96D09x-PSU0eW3Q7WkAjs,17427
-adv_optm/optim/Lion_Prodigy_adv.py,sha256=kIAGXoMbDNRg5reKXtUC_vQQ2gyM-NXPB-Pv9zSpiE8,12787
-adv_optm/optim/Lion_adv.py,sha256=05j_j6LIzHW5b79DVwMIf1FZHVNB8xnStNVjlOdVkCE,8256
-adv_optm/optim/Prodigy_adv.py,sha256=0BsCNYnq5O0UBvvyUuV385oMTJmNOyv5IQ3wEFuNfeM,21929
-adv_optm/optim/Simplified_AdEMAMix.py,sha256=opIZjnGJ03-DDAIHTZyJBMReVfgusGDb8FZSWMU3-UM,9774
-adv_optm/optim/__init__.py,sha256=pcP865H2j1tut2VfTUhzQh7V8TF_tzPjqFnjMfFed2k,382
-adv_optm/util/BF16_Stochastic_Rounding.py,sha256=Q5H0BcogmE4atP65dLoI21HKSf50lRdsBDfeF6v9Tbg,1548
-adv_optm/util/Effective_Shape.py,sha256=TBvIk1V8IuTbbBsxuekJA4e_v8JlR5Nujtut8RTWAm4,318
-adv_optm/util/NNMF.py,sha256=yRf5IP5Sjq0Uf0DxN0Q8NxEGSdD-f1ULziLVDOjY8K4,639
-adv_optm/util/One_Bit_Boolean.py,sha256=Wat49esdwohuN-OHOFMW8D0aOQgV9cP5Rl8z6yfmpos,1068
-adv_optm/util/OrthoGrad.py,sha256=NzInuBQGy_Ja__M1R9XbvqVaQ0fhGbtGgFE9YON7B3I,707
-adv_optm/util/__init__.py,sha256=qoyIF0jcLjs_vSEcsv36clw5LFNBEbifyXrrVxMH-G4,349
-adv_optm-1.0.0.dist-info/licenses/LICENSE,sha256=HrhfyXIkWY2tGFK11kg7vPCqhgh5DcxleloqdhrpyMY,11558
-adv_optm-1.0.0.dist-info/METADATA,sha256=sxxTU-62diwjtJPWHFeL6gMWuX8xmlWzEzqk9KfpXvA,8422
-adv_optm-1.0.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-adv_optm-1.0.0.dist-info/top_level.txt,sha256=iNfBIIzu-lPrQ7jyC56WBCcbkRwitM2nJ15-MRQ_6fg,9
-adv_optm-1.0.0.dist-info/RECORD,,