fairseq/optim/fused_adam.py

Survived

Mutant 858

--- fairseq/optim/fused_adam.py
+++ fairseq/optim/fused_adam.py
@@ -70,7 +70,7 @@
     def __init__(self, params,
                  lr=1e-3, bias_correction=True,
                  betas=(0.9, 0.999), eps=1e-8, eps_inside_sqrt=False,
-                 weight_decay=0., max_grad_norm=0., amsgrad=False):
+                 weight_decay=1.0, max_grad_norm=0., amsgrad=False):
         global fused_adam_cuda
         import importlib
         fused_adam_cuda = importlib.import_module("fused_adam_cuda")

Mutant 859

--- fairseq/optim/fused_adam.py
+++ fairseq/optim/fused_adam.py
@@ -70,7 +70,7 @@
     def __init__(self, params,
                  lr=1e-3, bias_correction=True,
                  betas=(0.9, 0.999), eps=1e-8, eps_inside_sqrt=False,
-                 weight_decay=0., max_grad_norm=0., amsgrad=False):
+                 weight_decay=0., max_grad_norm=1.0, amsgrad=False):
         global fused_adam_cuda
         import importlib
         fused_adam_cuda = importlib.import_module("fused_adam_cuda")

Mutant 860

--- fairseq/optim/fused_adam.py
+++ fairseq/optim/fused_adam.py
@@ -70,7 +70,7 @@
     def __init__(self, params,
                  lr=1e-3, bias_correction=True,
                  betas=(0.9, 0.999), eps=1e-8, eps_inside_sqrt=False,
-                 weight_decay=0., max_grad_norm=0., amsgrad=False):
+                 weight_decay=0., max_grad_norm=0., amsgrad=True):
         global fused_adam_cuda
         import importlib
         fused_adam_cuda = importlib.import_module("fused_adam_cuda")

Mutant 861

--- fairseq/optim/fused_adam.py
+++ fairseq/optim/fused_adam.py
@@ -88,7 +88,6 @@
         super().__init__(params, defaults)
         self.eps_mode = 0 if eps_inside_sqrt else 1
 
-    @property
     def supports_memory_efficient_fp16(self):
         return True
 

Mutant 862

--- fairseq/optim/fused_adam.py
+++ fairseq/optim/fused_adam.py
@@ -92,7 +92,6 @@
     def supports_memory_efficient_fp16(self):
         return True
 
-    @property
     def supports_flat_params(self):
         return True
 

Mutant 863

--- fairseq/optim/fused_adam.py
+++ fairseq/optim/fused_adam.py
@@ -96,7 +96,6 @@
     def supports_flat_params(self):
         return True
 
-    @property
     def supports_step_with_scale(self):
         return True
 

Mutant 864

--- fairseq/optim/fused_adam.py
+++ fairseq/optim/fused_adam.py
@@ -100,7 +100,7 @@
     def supports_step_with_scale(self):
         return True
 
-    def step(self, closure=None, grads=None, scale=1., grad_norms=None):
+    def step(self, closure=None, grads=None, scale=2.0, grad_norms=None):
         """Performs a single optimization step.
         Arguments:
             closure (callable, optional): A closure that reevaluates the model