d9d 0.1.1__tar.gz → 0.2.0__tar.gz

d9d-0.2.0/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

{d9d-0.1.1 → d9d-0.2.0}/PKG-INFO

@@ -1,8 +1,9 @@
 Metadata-Version: 2.4
 Name: d9d
-Version: 0.1.1
+Version: 0.2.0
 Summary: d9d - d[istribute]d - distributed training framework based on PyTorch that tries to be efficient yet hackable
 License: Apache-2.0
+License-File: LICENSE
 Author: Maksim Afanasyev
 Author-email: mr.applexz@gmail.com
 Requires-Python: >=3.11,<3.15

{d9d-0.1.1 → d9d-0.2.0}/d9d/kernel/swiglu/op.py

@@ -3,6 +3,15 @@ import triton
 import triton.language as tl
 
 
+def _size_bucket(n_elements: int) -> int:
+    # different auto-tuning for small and asymptotically large kernels
+    # perhaps we could extend this in future?
+    if n_elements < 8192:
+        return 0
+    else:
+        return 1
+
+
 @triton.autotune(
     configs=[
         triton.Config({"BLOCK_SIZE": 1024}, num_warps=4),
@@ -11,7 +20,7 @@ import triton.language as tl
         triton.Config({"BLOCK_SIZE": 4096}, num_warps=8),
         triton.Config({"BLOCK_SIZE": 8192}, num_warps=8),
     ],
-    key=["n_elements"]
+    key=["size_bucket"]
 )
 @triton.jit
 def _silu_mul_kernel(
@@ -19,6 +28,7 @@ def _silu_mul_kernel(
         y_ptr: torch.Tensor,
         out_ptr: torch.Tensor,
         n_elements: int,
+        size_bucket: int,  # used for autotuning
         BLOCK_SIZE: tl.constexpr,
 ):
     # prepare
@@ -72,7 +82,8 @@ def silu_mul_forward(x: torch.Tensor, y: torch.Tensor) -> torch.Tensor:
 
     _silu_mul_kernel[_grid](
         x, y, out,
-        n_elements
+        n_elements,
+        size_bucket=_size_bucket(n_elements)
     )
 
     return out
@@ -86,7 +97,7 @@ def silu_mul_forward(x: torch.Tensor, y: torch.Tensor) -> torch.Tensor:
         triton.Config({"BLOCK_SIZE": 4096}, num_warps=8),
         triton.Config({"BLOCK_SIZE": 8192}, num_warps=8),
     ],
-    key=["n_elements"]
+    key=["size_bucket"]
 )
 @triton.jit
 def _silu_mul_backward_kernel(
@@ -96,6 +107,7 @@ def _silu_mul_backward_kernel(
         grad_x_ptr: torch.Tensor,
         grad_y_ptr: torch.Tensor,
         n_elements: int,
+        size_bucket: int,  # used for autotuning
         BLOCK_SIZE: tl.constexpr
 ):
     # prepare
@@ -161,7 +173,8 @@ def silu_mul_backward(
     _silu_mul_backward_kernel[_grid](
         grad_output, x, y,
         grad_x, grad_y,
-        n_elements
+        n_elements,
+        size_bucket=_size_bucket(n_elements)
     )
 
     return grad_x, grad_y

{d9d-0.1.1 → d9d-0.2.0}/d9d/loop/component/__init__.py

@@ -6,13 +6,13 @@ from .gradient_clipper import GradientClipper
 from .gradient_manager import GradientManager
 from .job_logger import JobLogger
 from .job_profiler import JobProfiler
-from .loss_computer import LossComputer
 from .model_stage_exporter import ModelStageExporter
 from .model_stage_factory import ModelStageFactory, TrackedModules
 from .optimizer_factory import OptimizerFactory
+from .pipeline_result_processing import InferenceProcessor, LossComputer, PipelineOutputsProcessor
 from .stepper import Stepper
+from .task_operator import ForwardResult, InferenceTaskOperator, TrainTaskOperator
 from .timeout_manager import TimeoutManager
-from .train_task_operator import ForwardResult, TrainTaskOperator
 
 __all__ = [
     "BatchMaths",
@@ -20,6 +20,8 @@ __all__ = [
     "ForwardResult",
     "GradientClipper",
     "GradientManager",
+    "InferenceProcessor",
+    "InferenceTaskOperator",
     "JobLogger",
     "JobProfiler",
     "LossComputer",
@@ -27,6 +29,7 @@ __all__ = [
     "ModelStageExporter",
     "ModelStageFactory",
     "OptimizerFactory",
+    "PipelineOutputsProcessor",
     "StateCheckpointer",
     "Stepper",
     "TimeoutManager",

{d9d-0.1.1 → d9d-0.2.0}/d9d/loop/component/model_stage_factory.py

@@ -15,7 +15,7 @@ from d9d.pipelining.api import PipelineStageInfo
 from d9d.pipelining.factory.factory import PipelineScheduleInfo, build_schedule
 
 from .batch_maths import BatchMaths
-from .loss_computer import LossComputer
+from .pipeline_result_processing import PipelineOutputsProcessor
 
 StatefulPredicate = Callable[[str, torch.Tensor], bool]
 """Determines if a specific parameter or buffer should be included in the state dictionary."""
@@ -51,28 +51,6 @@ class TrackedModules(Stateful):
         self._modules = modules
         self._stateful_predicate = stateful_predicate
 
-    def __call__(self, *args: Any, **kwargs: Any) -> Any:
-        """
-        Forwards execution to the only pipeline stage.
-
-        This method is only valid when pipeline parallelism is disabled.
-
-        Args:
-            *args: Positional arguments passed to the module.
-            **kwargs: Keyword arguments passed to the module.
-
-        Returns:
-            The output of the model execution.
-
-        Raises:
-            ValueError: If pipeline parallelism is configured.
-        """
-
-        if self._dist_context.mesh_params.has_pipeline_parallel:
-            raise ValueError("You cannot call tracked modules when using pipelining")
-
-        return self._modules[0](*args, **kwargs)
-
     @property
     def modules(self) -> list[nn.Module]:
         """Returns the list of underlying PyTorch model modules."""
@@ -159,8 +137,8 @@ class ModelStageFactory:
             dist_context: DistributedContext,
             batch_maths: BatchMaths,
             config_model: ModelStageFactoryConfig,
-            config_pipelining: PipeliningConfig | None,
-            loss_computer: LossComputer | None
+            config_pipelining: PipeliningConfig,
+            pipeline_callback: PipelineOutputsProcessor
     ):
         """Constructs a ModelStageFactory object."""
 
@@ -169,7 +147,7 @@ class ModelStageFactory:
         self._config_model = config_model
         self._config_pipelining = config_pipelining
         self._batch_maths = batch_maths
-        self._loss_computer = loss_computer
+        self._pipeline_callback = pipeline_callback
 
     def _build_model_stage(self, stage: PipelineStageInfo) -> nn.Module:
         # create a model with no real memory occupied
@@ -218,21 +196,13 @@ class ModelStageFactory:
 
     def build_pipeline_and_modules(
             self
-    ) -> tuple[PipelineScheduleInfo | None, TrackedModules]:
+    ) -> tuple[PipelineScheduleInfo, TrackedModules]:
         """
         Constructs the execution schedule and the model container.
 
-        If pipeline parallelism is enabled, this orchestrates the creation of a
-        distributed pipeline schedule.
-
-        Otherwise, it simply builds a standalone model stage.
-
         Returns:
-           The pipeline schedule information (or None if no pipelining).
+           The pipeline schedule information.
            The `TrackedModules` instance wrapping the created model stage(s).
-
-        Raises:
-           ValueError: If pipelining configuration is missing but a pipeline is requested.
         """
 
         if self._config_model.checkpoint_only_trainable_parameters:
@@ -240,22 +210,12 @@ class ModelStageFactory:
         else:
             stateful_predicate = _stateful_predicate_always
 
-        if self._dist_context.mesh_params.has_pipeline_parallel:
-            if self._config_pipelining is None:
-                raise ValueError("Pipelining is enabled, but not configured")
-
-            loss_fn = self._loss_computer.compute_loss_mul_weight if self._loss_computer is not None else None
-
-            schedule, modules = build_schedule(
-                dist_context=self._dist_context,
-                n_microbatches=self._batch_maths.num_microbatches_pipelining,
-                schedule_config=self._config_pipelining.schedule,
-                model_provider=self._build_model_stage,
-                loss_fn=loss_fn
-            )
-
-            return schedule, TrackedModules(self._dist_context, modules, stateful_predicate)
-        else:
-            model = self._build_model_stage(PipelineStageInfo(num_stages=1, current_stage=0))
+        schedule, modules = build_schedule(
+            dist_context=self._dist_context,
+            n_microbatches=self._batch_maths.num_microbatches_pipelining,
+            schedule_config=self._config_pipelining.schedule,
+            model_provider=self._build_model_stage,
+            callback=self._pipeline_callback
+        )
 
-            return None, TrackedModules(self._dist_context, [model], stateful_predicate)
+        return schedule, TrackedModules(self._dist_context, modules, stateful_predicate)

d9d-0.1.1/d9d/loop/component/loss_computer.py → d9d-0.2.0/d9d/loop/component/pipeline_result_processing.py

@@ -1,7 +1,10 @@
+import abc
+from typing import Generic, TypeVar
+
 import torch
 
 from d9d.internals.pipeline_state import PipelineStateHandler
-from d9d.loop.control import ComputeLossContext, TrainTask
+from d9d.loop.control import ComputeLossContext, InferenceTask, ProcessOutputsContext, TrainTask
 
 from .stepper import Stepper
 
@@ -9,7 +12,20 @@ STATE_LOSS = "__internal_loss"
 STATE_LOSS_WEIGHT = "__internal_loss_weight"
 
 
-class LossComputer:
+TOutput = TypeVar("TOutput")
+
+
+class PipelineOutputsProcessor(abc.ABC, Generic[TOutput]):
+    @abc.abstractmethod
+    def __call__(
+            self,
+            pipeline_outputs: dict[str, torch.Tensor],
+            microbatch_idx: int
+    ) -> TOutput:
+        ...
+
+
+class LossComputer(PipelineOutputsProcessor[torch.Tensor]):
     """
     Handles the computation of loss values and their integration into the pipeline state.
 
@@ -38,10 +54,10 @@ class LossComputer:
         self._task = task
         self._stepper = stepper
 
-    def compute_loss_mul_weight(
+    def __call__(
             self,
             pipeline_outputs: dict[str, torch.Tensor],
-            microbatch_idx: int | None
+            microbatch_idx: int
     ) -> torch.Tensor:
         """
         Computes the weighted loss for a specific sharded microbatch or the full microbatch.
@@ -61,12 +77,9 @@ class LossComputer:
             The calculated loss multiplied by its weight.
         """
 
-        if microbatch_idx is None:
-            state = self._state.global_state()
-        else:
-            state = self._state.sharded_state(
-                shard_id=microbatch_idx
-            )
+        state = self._state.sharded_state(
+            shard_id=microbatch_idx
+        )
 
         computation = self._task.compute_loss(ComputeLossContext(
             pipeline_results=pipeline_outputs,
@@ -84,3 +97,53 @@ class LossComputer:
         state[STATE_LOSS_WEIGHT] = loss_weight[None]
 
         return loss * loss_weight
+
+
+class InferenceProcessor(PipelineOutputsProcessor[None]):
+    """
+    Handles the processing of model outputs during inference or evaluation.
+
+    This component retrieves the appropriate state context
+    and delegates the output processing logic to the user-defined inference task.
+    """
+
+    def __init__(
+            self,
+            state: PipelineStateHandler,
+            task: InferenceTask
+    ):
+        """
+        Constructs a new ModelOutputsProcessor.
+
+        Args:
+            state: Handler for managing global and sharded pipeline states.
+            task: The user-defined inference task containing processing logic.
+        """
+
+        self._state = state
+        self._task = task
+
+    def __call__(
+            self,
+            pipeline_outputs: dict[str, torch.Tensor],
+            microbatch_idx: int
+    ) -> None:
+        """
+        Processes model outputs for a specific microbatch or full batch.
+
+        This method retrieves the relevant state (scoped by microbatch index if provided)
+        and invokes the task's output processing logic.
+
+        Args:
+            pipeline_outputs: Dictionary containing model output tensors.
+            microbatch_idx: Index of the current microbatch, or None if not using microbatching.
+        """
+
+        state = self._state.sharded_state(
+            shard_id=microbatch_idx
+        )
+
+        self._task.process_outputs(ProcessOutputsContext(
+            pipeline_results=pipeline_outputs,
+            state=state
+        ))