[cleanup] multiprocess_pipe: dead-code removal and simplification (#335)

fairscale/nn/pipe/async_schedule.py

@@ -18,7 +18,7 @@ from fairscale.nn.model_parallel import get_pipeline_parallel_ranks
 from .messages import Transport
 from .microbatch import Batch
 from .skip.tracker import SkipTrackerThroughPotals
-from .types import EVENT_LOOP_QUEUE, PipelineStyle, PipeMessage, Tensors
+from .types import EVENT_LOOP_QUEUE, PipeMessage, Tensors
 
 
 @dataclass(frozen=True)
@@ -190,17 +190,13 @@ class AsyncEventLoop:
     ) -> Batch:
         """Actually run the forward pass for a given module, and send the result
         to the next stage in the pipeline if needed."""
-        assert self.group
+
+        # We import here to avoid a cyclic dependency.
+        # TODO(msb) Break the cyclic dependency.
         from .multiprocess_pipeline import create_task
 
         task = create_task(
-            PipelineStyle.AsyncSchedule,
-            self.checkpoint_stop,
-            batch.index,
-            self.group.rank(),
-            batch,
-            partition.module,
-            skip_trackers,
+            self.checkpoint_stop, batch.index, self.group.rank(), batch, partition.module, skip_trackers,
         )
         result = task.compute()
         task.finalize(result)
@@ -316,8 +312,6 @@ class AsyncEventLoop:
         calculated. This also handles the first/only stage for the special
         case of a 1-stage pipeline."""
 
-        assert self.group
-
         invocations, activations = self.get_invocations_and_activations()
         expected_invocations = len(invocations) * len(batches)
         actual_invocations = 0
@@ -379,7 +373,6 @@ class AsyncEventLoop:
 
     def event_loop(self, num_microbatch: int, skip_trackers: List[SkipTrackerThroughPotals]) -> None:
         """The event loop for the "middle", i.e. neither the head nor the tail"""
-        assert self.group
 
         invocations, activations = self.get_invocations_and_activations()
 

fairscale/nn/pipe/multiprocess_pipe.py

@@ -36,6 +36,7 @@ from . import microbatch
 from .async_schedule import Invocation, Location, ModuleWrapper
 from .batchnorm import DeferredBatchNorm
 from .multiprocess_pipeline import MultiProcessPipeline
+from .phony import get_phony
 from .skip.layout import SkipLayout, inspect_skip_layout
 from .skip.skippable import Skippable, verify_skippables
 from .types import LazyModule, PipelineStyle
@@ -43,9 +44,6 @@ from .types import LazyModule, PipelineStyle
 __all__ = ["MultiProcessPipe", "LazyModule"]
 
 
-Device = Union[torch.device, int, str]
-Devices = Union[Iterable[Device], List[Device]]
-
 Tensors = Tuple[Tensor, ...]
 TensorOrTensors = Union[Tensor, Tensors]
 
@@ -579,10 +577,6 @@ class MultiProcessPipe(Module):
         """
         microbatch.check(input)
 
-        if not self.group:
-            # Empty sequential module is not illegal.
-            return input
-
         if not self.pipeline:
             # No pipeline is not illegal, more ranks than partitions
             return input
@@ -594,19 +588,12 @@ class MultiProcessPipe(Module):
         with self.lock:
             self.pipeline.run(self.training, batches, event)
 
-            if not self.final_stage:
-                # Don't merge micro-batches to avoid unnecessary edges in autograd
-                # graph
-                # FIXME(tom) should figure out a proper type here
-                return batches  # type: ignore
-            else:
+            if self.final_stage:
                 # Merge the micro-batches into one mini-batch.
                 if self.pipelined_backward:
                     with torch.no_grad():
                         output = microbatch.gather(batches)
 
-                    from .phony import get_phony
-
                     phony = get_phony(
                         torch.device(torch.cuda.current_device() if torch.cuda.is_available() else "cpu"),
                         requires_grad=True,
@@ -614,6 +601,11 @@ class MultiProcessPipe(Module):
                     output = PipelinedBackwardPass.apply(output, batches, phony, True)  # self.retain_graph)
                 else:
                     output = microbatch.gather(batches)
+            else:
+                # Don't merge micro-batches to avoid unnecessary edges in autograd
+                # graph
+                # FIXME(tom) should figure out a proper type here
+                output = batches  # type: ignore
 
             return output
 
@@ -622,7 +614,7 @@ class MultiProcessPipe(Module):
             raise ValueError("back_helper should only be called on non-final stages")
 
         if self.pipeline:
-            self.pipeline.back_helper(list(reversed(output)))
+            self.pipeline.back_helper(output)
 
 
 class PipelinedBackwardPass(torch.autograd.Function):

fairscale/nn/pipe/multiprocess_pipeline.py

@@ -78,7 +78,7 @@ class RecvOperator(torch.autograd.Function):
 
     @staticmethod
     # type: ignore
-    def forward(ctx, dst_rank: int, tensor: Tensor, input_device, transport: Transport, index: int) -> Tensors:
+    def forward(ctx, dst_rank: int, tensor: Tensor, transport: Transport, index: int) -> Tensors:
         assert dst_rank == torch.distributed.get_rank()
         ctx.transport = transport
         ctx.index = index
@@ -120,7 +120,6 @@ else:
 
 
 def create_task(
-    style: PipelineStyle,
     checkpoint_stop: int,
     i: int,
     j: int,
@@ -176,7 +175,7 @@ class MultiProcessPipeline:
         skip_layout: SkipLayout,
         checkpoint_stop: int,
         style: PipelineStyle,
-        group: Optional[torch.distributed.ProcessGroup] = None,
+        group: torch.distributed.ProcessGroup,
         worker_map: Optional[Dict[int, str]] = None,
         input_device: Union[None, int, str, torch.device] = None,
         final_stage: bool = False,
@@ -193,7 +192,6 @@ class MultiProcessPipeline:
             input_device=input_device,
         )
         self.input_device = input_device
-        self.all_at_once = False
         self.callcount = 0
         self.final_stage = final_stage
 
@@ -219,11 +217,9 @@ class MultiProcessPipeline:
         skip_trackers = [SkipTrackerThroughPotals(self.skip_layout, i) for i in range(len(batches))]
 
         if self.style is PipelineStyle.MultiProcess:
-            assert self.group
             schedule = [(i, self.group.rank()) for i in range(m)]
             self.compute(batches, schedule, skip_trackers)
         elif self.style is PipelineStyle.AsyncSchedule:
-            assert self.group
             rank = self.group.rank()
             event_loop = AsyncEventLoop(
                 self.partitions, self.group, self.transport, self.training, self.checkpoint_stop,
@@ -248,7 +244,7 @@ class MultiProcessPipeline:
     ) -> Batch:
 
         phony = torch.empty(0, device=self.input_device, requires_grad=True)
-        result = RecvOperator.apply(torch.distributed.get_rank(), phony, self.input_device, self.transport, i)
+        result = RecvOperator.apply(torch.distributed.get_rank(), phony, self.transport, i)
         if len(result) == 1:
             batch = Batch(result[0], i)
         else:
@@ -261,7 +257,6 @@ class MultiProcessPipeline:
     def send_skip_tensors(
         self, this_rank: int, ranks: List[int], batch: Batch, i: int, skip_trackers: List[SkipTrackerThroughPotals]
     ) -> None:
-        assert self.group
         for next_j, ns, name in self.skip_layout.copy_policy_by_src(self.group.rank()):
             life = skip_trackers[i].portals[(ns, name)].tensor_life
             loaded = skip_trackers[i].load(batch, ns, name)
@@ -302,7 +297,6 @@ class MultiProcessPipeline:
     def execute_task(self, task: Task, i: int, skip_trackers: List[SkipTrackerThroughPotals]) -> Batch:
         batch = task.compute()
 
-        assert self.group
         rank = self.group.rank()
 
         if self.style is PipelineStyle.MultiProcess and not self.final_stage:
@@ -324,9 +318,7 @@ class MultiProcessPipeline:
     ) -> None:
         """Runs tasks with synchronization to copy streams."""
 
-        if self.style is PipelineStyle.MultiProcess:
-            assert self.group
-            n = self.group.size()
+        assert self.style is PipelineStyle.MultiProcess
 
         # With checkpointing, the autograd graph looks like this diagram:
         # ┌─────┸──────┐
@@ -354,17 +346,15 @@ class MultiProcessPipeline:
         # │    Copy    │
         # └─────┰──────┘
         for i, j in schedule:
-            batch = batches[i]
-
-            if self.style is PipelineStyle.MultiProcess:
             assert len(self.partitions) == 1
             partition = self.partitions[0]
 
-                assert self.group
             if self.group.rank() != 0:
                 batch = self.get_batch_from_previous_stage(i, skip_trackers, batches)
+            else:
+                batch = batches[i]
 
-                task = create_task(self.style, self.checkpoint_stop, i, j, batch, partition.module, skip_trackers)
+            task = create_task(self.checkpoint_stop, i, j, batch, partition.module, skip_trackers)
 
             batches[i] = self.execute_task(task, i, skip_trackers)
 
@@ -398,26 +388,10 @@ class MultiProcessPipeline:
         if self.style == PipelineStyle.AsyncSchedule:
             return
 
-        o = list(output)
-
         tensors: Tensors
 
-        if self.all_at_once:
-            # FIXME(tom) allow specifying this branch when constructing Pipe(), add a test
-            grads = []
-            for i, batch in enumerate(o):
-                rank = torch.distributed.get_rank()
-                found = self.transport.get_out_of_order(ACTIVATIONS_GRADS_QUEUE, i)
-                assert len(found) == 1
-                grads.append(found[0])
-                tensors = tuple(x.tensor_or_tensors for x in o)  # type: ignore
-            try:
-                torch.autograd.backward(tensors, grad_tensors=grads, retain_graph=True)
-            except Exception as e:
-                raise RuntimeError("Autograd failed") from e
-        else:
         rank = torch.distributed.get_rank()
-            for batch in o:
+        for batch in reversed(output):
             found = self.transport.get_out_of_order(ACTIVATIONS_GRADS_QUEUE, batch.index)
             if batch.atomic:
                 tensors = tuple([batch.tensor])