Fix type documentation for optimizer `args`

bitsandbytes/optim/adagrad.py

@@ -38,8 +38,8 @@ class Adagrad(Optimizer1State):
                 The epsilon value prevents division by zero in the optimizer.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -105,8 +105,8 @@ class Adagrad8bit(Optimizer1State):
                 The epsilon value prevents division by zero in the optimizer.
             optim_bits (`int`, defaults to 8):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -173,8 +173,8 @@ class Adagrad32bit(Optimizer1State):
                 The epsilon value prevents division by zero in the optimizer.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):

bitsandbytes/optim/adam.py

@@ -47,8 +47,8 @@ class Adam(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -108,8 +108,8 @@ class Adam8bit(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -169,8 +169,8 @@ class Adam32bit(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -230,8 +230,8 @@ class PagedAdam(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -291,8 +291,8 @@ class PagedAdam8bit(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -352,8 +352,8 @@ class PagedAdam32bit(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):

bitsandbytes/optim/adamw.py

@@ -39,8 +39,8 @@ class AdamW(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -100,8 +100,8 @@ class AdamW8bit(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -161,8 +161,8 @@ class AdamW32bit(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -221,8 +221,8 @@ class PagedAdamW(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -281,8 +281,8 @@ class PagedAdamW8bit(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -341,8 +341,8 @@ class PagedAdamW32bit(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):

bitsandbytes/optim/lamb.py

@@ -45,8 +45,8 @@ class LAMB(Optimizer2State):
                 Whether to use the AdamW variant.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -109,8 +109,8 @@ class LAMB8bit(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             adam_w_mode (`bool`, defaults to `True`):
                 Whether to use the AdamW variant.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -173,8 +173,8 @@ class LAMB32bit(Optimizer2State):
                 Whether to use the [AMSGrad](https://hf.co/papers/1904.09237) variant of Adam that uses the maximum of past squared gradients instead.
             adam_w_mode (`bool`, defaults to `True`):
                 Whether to use the AdamW variant.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):

bitsandbytes/optim/lars.py

@@ -41,8 +41,8 @@ class LARS(Optimizer1State):
                 Whether to use Nesterov momentum.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -98,8 +98,8 @@ class LARS8bit(Optimizer1State):
                 The weight decay value for the optimizer.
             nesterov (`bool`, defaults to `False`):
                 Whether to use Nesterov momentum.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -155,8 +155,8 @@ class LARS32bit(Optimizer1State):
                 The weight decay value for the optimizer.
             nesterov (`bool`, defaults to `False`):
                 Whether to use Nesterov momentum.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):

bitsandbytes/optim/lion.py

@@ -33,8 +33,8 @@ class Lion(Optimizer1State):
                 The weight decay value for the optimizer.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -85,8 +85,8 @@ class Lion8bit(Optimizer1State):
                 The beta values are the decay rates of the first and second-order moment of the optimizer.
             weight_decay (`float`, defaults to 0):
                 The weight decay value for the optimizer.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -137,8 +137,8 @@ class Lion32bit(Optimizer1State):
                 The beta values are the decay rates of the first and second-order moment of the optimizer.
             weight_decay (`float`, defaults to 0):
                 The weight decay value for the optimizer.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -191,8 +191,8 @@ class PagedLion(Optimizer1State):
                 The weight decay value for the optimizer.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -242,8 +242,8 @@ class PagedLion8bit(Optimizer1State):
                 The weight decay value for the optimizer.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -293,8 +293,8 @@ class PagedLion32bit(Optimizer1State):
                 The weight decay value for the optimizer.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):

bitsandbytes/optim/optimizer.py

@@ -373,8 +373,8 @@ class Optimizer2State(Optimizer8bit):
                 The weight decay value for the optimizer.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -596,8 +596,8 @@ class Optimizer1State(Optimizer8bit):
                 The weight decay value for the optimizer.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):

bitsandbytes/optim/rmsprop.py

@@ -41,8 +41,8 @@ class RMSprop(Optimizer1State):
                 Whether the gradients are normalized by the variance. If `True`, it can help training at the expense of additional compute.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -104,8 +104,8 @@ class RMSprop8bit(Optimizer1State):
                 Whether the gradients are normalized by the variance. If `True`, it can help training at the expense of additional compute.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -167,8 +167,8 @@ class RMSprop32bit(Optimizer1State):
                 Whether the gradients are normalized by the variance. If `True`, it can help training at the expense of additional compute.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):

bitsandbytes/optim/sgd.py

@@ -38,8 +38,8 @@ class SGD(Optimizer1State):
                 Whether to use Nesterov momentum.
             optim_bits (`int`, defaults to 32):
                 The number of bits of the optimizer state.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -94,8 +94,8 @@ class SGD8bit(Optimizer1State):
                 The weight decay value for the optimizer.
             nesterov (`bool`, defaults to `False`):
                 Whether to use Nesterov momentum.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):
@@ -150,8 +150,8 @@ class SGD32bit(Optimizer1State):
                 The weight decay value for the optimizer.
             nesterov (`bool`, defaults to `False`):
                 Whether to use Nesterov momentum.
-            args (`dict`, defaults to `None`):
-                A dictionary with additional arguments.
+            args (`object`, defaults to `None`):
+                An object with additional arguments.
             min_8bit_size (`int`, defaults to 4096):
                 The minimum number of elements of the parameter tensors for 8-bit optimization.
             percentile_clipping (`int`, defaults to 100):