fix layers/schedule for hybrid parallelization (#111) (#112)

colossalai/engine/schedule/_pipeline_schedule.py

@@ -48,9 +48,13 @@ class PipelineSchedule(BaseSchedule):
         # Pipeline schedule just puts data in memory
         self.batch_data, self.batch_label = super().load_batch(data_iter, to_gpu=False)
         self.microbatch_offset = 0
-        assert self.batch_size % self.num_microbatches == 0, \
+        if isinstance(self.batch_data, torch.Tensor):
+            batch_size = self.batch_data.size(0)
+        else:
+            batch_size = next(iter(self.batch_data.values())).size(0)
+        assert batch_size % self.num_microbatches == 0, \
             "Batch size should divided by the number of microbatches"
-        self.microbatch_size = self.batch_size // self.num_microbatches
+        self.microbatch_size = batch_size // self.num_microbatches
 
     def _get_data_slice(self, data, offset):
         if isinstance(data, torch.Tensor):

colossalai/nn/layer/parallel_1d/layers.py

@@ -71,6 +71,7 @@ class Linear1D(torch.nn.Module):
 @LAYERS.register_module
 class Classifier1D(ParallelLayer):
     """RowLinear with given weight"""
+
     def __init__(self,
                  in_features: int,
                  num_classes: int,
@@ -127,8 +128,8 @@ class Classifier1D(ParallelLayer):
 
         output_parallel = F.linear(input_, self.weight)
         output = reduce_input(output_parallel, ParallelMode.PARALLEL_1D)
-
-        output = output + self.bias
+        if self.bias is not None:
+            output = output + self.bias
         return output
 
 
@@ -152,6 +153,7 @@ class Linear1D_Col(ParallelLayer):
                     which is :math:`Y_i = XA_i`, defaults to False
     :type gather_output: bool, optional
     """
+
     def __init__(self,
                  in_features: int,
                  out_features: int,
@@ -233,6 +235,7 @@ class Linear1D_Row(ParallelLayer):
     :param parallel_input: If set to ``True``, it's assumed that the input is splitted, defaults to False
     :type parallel_input: bool, optional
     """
+
     def __init__(self,
                  in_features: int,
                  out_features: int,
@@ -302,6 +305,7 @@ class Linear1D_Row(ParallelLayer):
 class MixedFusedLayerNorm1D(torch.nn.Module):
     """ Experimental
     """
+
     def __init__(self, normalized_shape, eps=1e-5):
         super(MixedFusedLayerNorm1D, self).__init__()
 

colossalai/nn/layer/parallel_2d/_operation.py

@@ -121,9 +121,10 @@ class classifier_2d(torch.autograd.Function):
             B_grad = torch.matmul(output_grad.reshape(-1, output_grad.shape[-1]).transpose(0, 1), A)
             B_grad = reduce_scatter(B_grad, -1, ctx.col_parallel_mode)
             B_grad = B_grad.reshape(ctx.B_shape)
-
-            bias_grad = torch.sum(output_grad, dim=tuple(range(output_grad.ndim - 1)))
-            bias_grad = all_reduce(bias_grad, ctx.col_parallel_mode)
+            bias_grad = None
+            if ctx.use_bias:
+                bias_grad = torch.sum(output_grad, dim=tuple(range(output_grad.ndim - 1)))
+                bias_grad = all_reduce(bias_grad, ctx.col_parallel_mode)
 
         return A_grad, B_grad, bias_grad, None, None, None, None, None, None, None, None, None, None
 
@@ -174,9 +175,9 @@ class Matmul_AB_2D(torch.autograd.Function):
         col_group = gpc.get_group(col_parallel_mode)
 
         src_a = summa_dim * row_rank + data_parallel_rank * pipeline_parallel_size * tensor_parallel_size + \
-                pipeline_parallel_rank * tensor_parallel_size
+            pipeline_parallel_rank * tensor_parallel_size
         src_b = col_rank + data_parallel_rank * pipeline_parallel_size * tensor_parallel_size + \
-                pipeline_parallel_rank * tensor_parallel_size
+            pipeline_parallel_rank * tensor_parallel_size
 
         opa = [None] * 2
         opb = [None] * 2
@@ -279,9 +280,9 @@ class Matmul_ABT_2D(torch.autograd.Function):
         col_group = gpc.get_group(col_parallel_mode)
 
         src_b = col_rank + data_parallel_rank * pipeline_parallel_size * tensor_parallel_size + \
-                pipeline_parallel_rank * tensor_parallel_size
+            pipeline_parallel_rank * tensor_parallel_size
         src_c = summa_dim * row_rank + data_parallel_rank * pipeline_parallel_size * tensor_parallel_size + \
-                pipeline_parallel_rank * tensor_parallel_size
+            pipeline_parallel_rank * tensor_parallel_size
 
         opb = [None] * 2
         opr = [None] * 2
@@ -393,9 +394,9 @@ class Matmul_ATB_2D(torch.autograd.Function):
         col_group = gpc.get_group(col_parallel_mode)
 
         src_a = summa_dim * row_rank + data_parallel_rank * pipeline_parallel_size * tensor_parallel_size + \
-                pipeline_parallel_rank * tensor_parallel_size
+            pipeline_parallel_rank * tensor_parallel_size
         src_c = col_rank + data_parallel_rank * pipeline_parallel_size * tensor_parallel_size + \
-                pipeline_parallel_rank * tensor_parallel_size
+            pipeline_parallel_rank * tensor_parallel_size
 
         opa = [None] * 2
         opr = [None] * 2

colossalai/nn/layer/wrapper/pipeline_wrapper.py

@@ -38,3 +38,9 @@ class PipelineSharedModuleWrapper:
         for p in module.parameters():
             setattr(p, 'pipeline_shared_module_pg', self.group)
             dist.broadcast(p, src, group=self.group)
+
+    def register_parameter(self, param: nn.Parameter):
+        assert self.ranks_in_group is not None, f'Rank {gpc.get_local_rank(ParallelMode.PIPELINE)} is not in pipeline_ranks {self.pipeline_ranks}'
+        src = self.ranks_in_group[self.pipeline_ranks[0]]
+        setattr(param, 'pipeline_shared_module_pg', self.group)
+        dist.broadcast(param, src, group=self.group)

colossalai/trainer/hooks/_metric_hook.py

@@ -25,6 +25,7 @@ class Metric(ABC):
     :param epoch_only: Whether the metric only read for the full epoch
     :type epoch_only: bool
     """
+
     def __init__(self, epoch_only: bool):
         # is the metric only read for the full epoch
         self._epoch_only = epoch_only
@@ -82,6 +83,7 @@ class LossMetric(Metric):
     :param epoch_only: Whether the metric only read for the full epoch
     :type epoch_only: bool
     """
+
     def __init__(self, epoch_only):
         super().__init__(epoch_only=epoch_only)
         self.last_step_loss = torch.zeros(1, device=get_current_device())
@@ -132,6 +134,7 @@ class LearningRateMetric(Metric):
     :param epoch_only: Whether the metric only read for the full epoch
     :type epoch_only: bool
     """
+
     def __init__(self, epoch_only: bool, initial_lr: float = 0.):
         super().__init__(epoch_only=epoch_only)
         self.lr = initial_lr
@@ -159,6 +162,7 @@ class AccuracyMetric(Metric):
     :param epoch_only: Whether the metric only read for the full epoch
     :type epoch_only: bool
     """
+
     def __init__(self, epoch_only: bool, accuracy_func: Callable):
         super().__init__(epoch_only=epoch_only)
         self.acc = accuracy_func
@@ -217,6 +221,7 @@ class MetricHook(BaseHook):
     :type trainer: Trainer
     :type priority: int
     """
+
     def __init__(
         self,
         priority: int,
@@ -238,6 +243,7 @@ class LossHook(MetricHook):
     :type trainer: Trainer
     :type priority: int, optional
     """
+
     def __init__(self, priority: int = 0):
         super().__init__(priority)
 
@@ -278,6 +284,7 @@ class AccuracyHook(MetricHook):
     :type trainer: Trainer
     :type priority: int
     """
+
     def __init__(self, accuracy_func: Callable, priority: int = 0):
         super().__init__(priority)
         self.accuracy_func = accuracy_func
@@ -351,13 +358,17 @@ class ThroughputHook(MetricHook):
             trainer.states['metrics']['test']['Throughput'] = self.metric
 
     def before_train_epoch(self, trainer):
-        self.metric.reset()
+        if self._is_stage_to_compute:
+            self.metric.reset()
 
     def after_train_iter(self, trainer, *args):
-        self.metric.update(trainer.schedule.batch_size, trainer._timer.get_timer('Train-step').get_elapsed_time())
+        if self._is_stage_to_compute:
+            self.metric.update(trainer.schedule.batch_size, trainer._timer.get_timer('Train-step').get_elapsed_time())
 
     def before_test(self, trainer):
-        self.metric.reset()
+        if self._is_stage_to_compute:
+            self.metric.reset()
 
     def after_test_iter(self, trainer, *args):
-        self.metric.update(trainer.schedule.batch_size, trainer._timer.get_timer('Test-step').get_elapsed_time())
+        if self._is_stage_to_compute:
+            self.metric.update(trainer.schedule.batch_size, trainer._timer.get_timer('Test-step').get_elapsed_time())

model_zoo/gpt/gpt.py

@@ -133,7 +133,7 @@ class GPTBlock(CheckpointModule):
                  dtype: dtype = None,
                  bias: bool = True,
                  checkpoint: bool = False):
-        super().__init__()
+        super().__init__(checkpoint=checkpoint)
         self.norm1 = col_nn.LayerNorm(normalized_shape=dim, eps=1e-6, dtype=dtype)
         self.attn = GPTSelfAttention(dim=dim,
                                      num_heads=num_heads,