python typing and comments

pybuda/pybuda/compile.py

@@ -32,7 +32,7 @@
 from pybuda._C.graph import Graph
 from pybuda._C.runtime import Binary
 import pybuda.ci as ci
-from pybuda.module import PyBudaModule, wrap_module
+from pybuda.module import Module, PyBudaModule, wrap_module
 from pybuda.parameter import Parameter
 from pybuda.pybudaglobal import state_changed, clear_state_changed
 import pybuda.query as query
@@ -95,14 +95,14 @@ def generate_override_config(graph, balancer_solution, placer_solution, nop_inst
 
 @dataclass
 class CompileContext:
-    modules: List[PyBudaModule]
+    modules: List[Module]
     graph_name: str
     compiler_cfg: CompilerConfig
     verify_cfg: VerifyConfig
     microbatch_size: int
     microbatch_count: int
-    inputs: Optional[Tuple[Union[Tensor, List[Any], Dict[str, Any]],...]] = None
-    loss_module: Optional[PyBudaModule] = None
+    inputs: Union[torch.Tensor, List[torch.Tensor]]
+    loss_module: Optional[Module] = None
     optimizer: Optional[torch.optim.Optimizer] = None
     training: bool = False
     graph: Optional[Graph] = None
@@ -164,13 +164,35 @@ def calculate_grads(
 
 def compile_main(
         module: torch.nn.Module | tf.keras.Model | PyBudaModule,
-        sample_inputs: Optional[Tuple[Union[Tensor, List[Any], Dict[str, Any]],...]] = None,
+        sample_inputs: List[torch.Tensor],
         module_name: Optional[str] = None,
         loss: Optional[torch.nn.Module | PyBudaModule] = None,
         optimizer: Optional[torch.optim.Optimizer] = None,
-):
+) -> CompiledModel:
     """
     Main entry point for compiling modules from different frameworks for Tenstorrent devices.
+
+    Parameters
+    ----------
+    module: torch.nn.Module | tf.keras.Model | PyBudaModule
+        Torch, TensorFlow, or PyBuda module to compile
+
+    sample_inputs: List[torch.Tensor]
+        List of sample inputs for the module (used to infer shapes)
+
+    module_name: Optional[str]
+        Name of the module. If not provided, the class name of the provided module will be used.
+
+    loss: Optional[torch.nn.Module | PyBudaModule]
+        Loss module for training.
+
+    optimizer: Optional[torch.optim.Optimizer]
+        Optimizer for training.
+
+    Returns
+    -------
+    CompiledModel - Callable object that can be used to run the compiled module on device.
+
     """
 
     assert isinstance(module, torch.nn.Module) or isinstance(module, tf.keras.Model) or isinstance(module, PyBudaModule), "Only PyTorch, TensorFlow, and PyBuda modules are supported."

pybuda/pybuda/compiled_graph_state.py

@@ -307,7 +307,7 @@ def __call__(self, *inputs: torch.Tensor) -> List[torch.Tensor]:
 
         Parameters
         ----------
-        inputs: Tuple[Tensor, ...]
+        inputs: [Tensor, ...]
             Input tensors
 
         Returns
@@ -329,7 +329,7 @@ def __call__(self, *inputs: torch.Tensor) -> List[torch.Tensor]:
         return outputs
 
     def forward(self, *inputs: torch.Tensor) -> List[torch.Tensor]:
-        return self(*inputs)
+        return self(inputs)
 
     def backward(self, loss_grad: torch.Tensor) -> List[torch.Tensor]:
         assert self.compiled_graph_state.graph.training(), "Model not compiled for training."

pybuda/test/mlir/test_training.py

@@ -2,13 +2,9 @@
 
 # SPDX-License-Identifier: Apache-2.0
 
-import os
-import pytest
 import torch
 import torch.nn as nn
 
-import tensorflow as tf
-
 import pybuda
 import pybuda.config
 
@@ -52,9 +48,9 @@ def forward(self, x1):
         print(f"epoch: {epoch} loss: {loss}")
         print(f"output.grad: {output[0].grad}")
 
-        golden_loss = loss_fn(golden, target)
-
         loss_grad = output[0].grad
+        assert loss_grad is not None
+
         print(f"loss grad: {loss_grad}")
         grad = tt_model.backward(loss_grad)