[shardformer] bloom support sequence parallel (#4465)

[shardformer] bloom support sequence parallel

colossalai/shardformer/modeling/bloom.py

@@ -23,6 +23,10 @@ from transformers.models.bloom.modeling_bloom import (
 from transformers.utils import logging
 
 from colossalai.pipeline.stage_manager import PipelineStageManager
+from colossalai.shardformer.layer._operation import gather_forward_split_backward, split_forward_gather_backward
+from colossalai.shardformer.shard import ShardConfig
+
+logger = logging.get_logger(__name__)
 
 
 def build_bloom_alibi_tensor_fn(process_group: ProcessGroup) -> torch.Tensor:
@@ -111,6 +115,7 @@ class BloomPipelineForwards:
         stage_manager: Optional[PipelineStageManager] = None,
         hidden_states: Optional[torch.FloatTensor] = None,
         stage_index: Optional[List[int]] = None,
+        shard_config: ShardConfig = None,
         **deprecated_arguments,
     ) -> Union[Tuple[torch.Tensor, ...], 'BaseModelOutputWithPastAndCrossAttentions']:
 
@@ -205,6 +210,13 @@ class BloomPipelineForwards:
             past_key_values_length=past_key_values_length,
         )
 
+        # split the input tensor along sequence dimension
+        # [batch_size, seq_len, hidden_size] -> [batch_size, seq_len/TP_size, hidden_size]
+        if shard_config.enable_sequence_parallelism:
+            hidden_states = split_forward_gather_backward(hidden_states,
+                                                          dim=1,
+                                                          process_group=shard_config.tensor_parallel_process_group)
+
         start_idx, end_idx = stage_index[0], stage_index[1]
         for i, (block, layer_past) in enumerate(zip(self.h[start_idx:end_idx], past_key_values[start_idx:end_idx]),
                                                 start=start_idx):
@@ -248,6 +260,12 @@ class BloomPipelineForwards:
                 all_self_attentions = all_self_attentions + \
                     (outputs[2 if use_cache else 1],)
 
+        # When sequence parallelism done, gather the output tensor in forward and split it in backward
+        if shard_config.enable_sequence_parallelism:
+            hidden_states = gather_forward_split_backward(hidden_states,
+                                                          dim=1,
+                                                          process_group=shard_config.tensor_parallel_process_group)
+
         if stage_manager.is_last_stage():
             # Add last hidden state
             hidden_states = self.ln_f(hidden_states)
@@ -287,6 +305,7 @@ class BloomPipelineForwards:
                                     stage_manager: Optional[PipelineStageManager] = None,
                                     hidden_states: Optional[torch.FloatTensor] = None,
                                     stage_index: Optional[List[int]] = None,
+                                    shard_config: ShardConfig = None,
                                     **deprecated_arguments):
         r"""
         labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
@@ -327,7 +346,8 @@ class BloomPipelineForwards:
                                                                         return_dict=return_dict,
                                                                         stage_manager=stage_manager,
                                                                         hidden_states=hidden_states,
-                                                                        stage_index=stage_index)
+                                                                        stage_index=stage_index,
+                                                                        shard_config=shard_config)
         past_key_values = None
         all_hidden_states = None
         all_self_attentions = None
@@ -380,6 +400,7 @@ class BloomPipelineForwards:
         stage_manager: Optional[PipelineStageManager] = None,
         hidden_states: Optional[torch.FloatTensor] = None,
         stage_index: Optional[List[int]] = None,
+        shard_config: ShardConfig = None,
         **deprecated_arguments,
     ):
         r"""
@@ -424,6 +445,7 @@ class BloomPipelineForwards:
             stage_manager=stage_manager,
             hidden_states=hidden_states,
             stage_index=stage_index,
+            shard_config=shard_config,
         )
         past_key_values = None
         all_hidden_states = None
@@ -503,6 +525,7 @@ class BloomPipelineForwards:
         stage_manager: Optional[PipelineStageManager] = None,
         hidden_states: Optional[torch.FloatTensor] = None,
         stage_index: Optional[List[int]] = None,
+        shard_config: ShardConfig = None,
         **deprecated_arguments,
     ):
         r"""
@@ -547,6 +570,7 @@ class BloomPipelineForwards:
             stage_manager=stage_manager,
             hidden_states=hidden_states,
             stage_index=stage_index,
+            shard_config=shard_config,
         )
         past_key_values = None
         all_hidden_states = None
@@ -597,6 +621,7 @@ class BloomPipelineForwards:
         stage_manager: Optional[PipelineStageManager] = None,
         hidden_states: Optional[torch.FloatTensor] = None,
         stage_index: Optional[List[int]] = None,
+        shard_config: ShardConfig = None,
     ):
         r"""
         start_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
@@ -632,6 +657,7 @@ class BloomPipelineForwards:
             stage_manager=stage_manager,
             hidden_states=hidden_states,
             stage_index=stage_index,
+            shard_config=shard_config,
         )
         past_key_values = None
         all_hidden_states = None
@@ -700,8 +726,7 @@ def get_bloom_flash_attention_forward(enabel_jit_fused=False):
 
         fused_qkv = self.query_key_value(hidden_states)
         (query_layer, key_layer, value_layer) = self._split_heads(fused_qkv)
-        batch_size, tgt_len, _ = hidden_states.size()
+        batch_size, tgt_len, _ = query_layer.size()
-        assert tgt_len % 4 == 0, "Flash Attention Error: The sequence length should be a multiple of 4."
 
         _, kv_length, _, _ = key_layer.size()
 
@@ -896,3 +921,156 @@ def get_jit_fused_bloom_gelu_forward():
             return self.bloom_gelu_forward(x, bias)
 
     return forward
+
+
+def get_bloom_sequence_parallel_forward_fn(shard_config: ShardConfig):
+
+    from transformers import BloomModel
+
+    def forward(
+        self: BloomModel,
+        input_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.LongTensor] = None,
+        inputs_embeds: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **deprecated_arguments,
+    ) -> Union[Tuple[torch.Tensor, ...], BaseModelOutputWithPastAndCrossAttentions]:
+        if deprecated_arguments.pop("position_ids", False) is not False:
+            # `position_ids` could have been `torch.Tensor` or `None` so defaulting pop to `False` allows to detect if users were passing explicitly `None`
+            warnings.warn(
+                "`position_ids` have no functionality in BLOOM and will be removed in v5.0.0. You can safely ignore"
+                " passing `position_ids`.",
+                FutureWarning,
+            )
+        if len(deprecated_arguments) > 0:
+            raise ValueError(f"Got unexpected arguments: {deprecated_arguments}")
+
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (output_hidden_states
+                                if output_hidden_states is not None else self.config.output_hidden_states)
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
+        elif input_ids is not None:
+            batch_size, seq_length = input_ids.shape
+        elif inputs_embeds is not None:
+            batch_size, seq_length, _ = inputs_embeds.shape
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        if past_key_values is None:
+            past_key_values = tuple([None] * len(self.h))
+
+        # Prepare head mask if needed
+        # 1.0 in head_mask indicate we keep the head
+        # attention_probs has shape batch_size x num_heads x N x N
+        # head_mask has shape n_layer x batch x num_heads x N x N
+        head_mask = self.get_head_mask(head_mask, self.config.n_layer)
+
+        if inputs_embeds is None:
+            inputs_embeds = self.word_embeddings(input_ids)
+
+        hidden_states = self.word_embeddings_layernorm(inputs_embeds)
+
+        presents = () if use_cache else None
+        all_self_attentions = () if output_attentions else None
+        all_hidden_states = () if output_hidden_states else None
+
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`...")
+                use_cache = False
+
+        # Compute alibi tensor: check build_alibi_tensor documentation
+        seq_length_with_past = seq_length
+        past_key_values_length = 0
+        if past_key_values[0] is not None:
+            past_key_values_length = past_key_values[0][0].shape[2]
+            seq_length_with_past = seq_length_with_past + past_key_values_length
+        if attention_mask is None:
+            attention_mask = torch.ones((batch_size, seq_length_with_past), device=hidden_states.device)
+        else:
+            attention_mask = attention_mask.to(hidden_states.device)
+
+        alibi = self.build_alibi_tensor(attention_mask, self.num_heads, dtype=hidden_states.dtype)
+
+        causal_mask = self._prepare_attn_mask(
+            attention_mask,
+            input_shape=(batch_size, seq_length),
+            past_key_values_length=past_key_values_length,
+        )
+        # split the input tensor along sequence dimension
+        # [batch_size, seq_len, hidden_size] -> [batch_size, seq_len/TP_size, hidden_size]
+        hidden_states = split_forward_gather_backward(hidden_states,
+                                                      dim=1,
+                                                      process_group=shard_config.tensor_parallel_process_group)
+
+        for i, (block, layer_past) in enumerate(zip(self.h, past_key_values)):
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+
+                def create_custom_forward(module):
+
+                    def custom_forward(*inputs):
+                        # None for past_key_value
+                        return module(*inputs, use_cache=use_cache, output_attentions=output_attentions)
+
+                    return custom_forward
+
+                outputs = torch.utils.checkpoint.checkpoint(
+                    create_custom_forward(block),
+                    hidden_states,
+                    alibi,
+                    causal_mask,
+                    layer_past,
+                    head_mask[i],
+                )
+            else:
+                outputs = block(
+                    hidden_states,
+                    layer_past=layer_past,
+                    attention_mask=causal_mask,
+                    head_mask=head_mask[i],
+                    use_cache=use_cache,
+                    output_attentions=output_attentions,
+                    alibi=alibi,
+                )
+
+            hidden_states = outputs[0]
+            if use_cache is True:
+                presents = presents + (outputs[1],)
+
+            if output_attentions:
+                all_self_attentions = all_self_attentions + (outputs[2 if use_cache else 1],)
+
+        # When sequence parallelism done, gather the output tensor in forward and split it in backward
+        hidden_states = gather_forward_split_backward(hidden_states,
+                                                      dim=1,
+                                                      process_group=shard_config.tensor_parallel_process_group)
+        # Add last hidden state
+        hidden_states = self.ln_f(hidden_states)
+
+        if output_hidden_states:
+            all_hidden_states = all_hidden_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(v for v in [hidden_states, presents, all_hidden_states, all_self_attentions] if v is not None)
+
+        return BaseModelOutputWithPastAndCrossAttentions(
+            last_hidden_state=hidden_states,
+            past_key_values=presents,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attentions,
+        )
+
+    return forward

colossalai/shardformer/policies/bloom.py

@@ -12,6 +12,7 @@ from ..modeling.bloom import (
     BloomPipelineForwards,
     build_bloom_alibi_tensor_fn,
     get_bloom_flash_attention_forward,
+    get_bloom_sequence_parallel_forward_fn,
     get_jit_fused_bloom_attention_forward,
     get_jit_fused_bloom_gelu_forward,
     get_jit_fused_bloom_mlp_forward,
@@ -43,6 +44,7 @@ class BloomPolicy(Policy):
 
         policy = {}
 
+        use_sequence_parallel = self.shard_config.enable_sequence_parallelism
         if self.shard_config.enable_tensor_parallelism:
             policy[BloomBlock] = ModulePolicyDescription(attribute_replacement={
                 "self_attention.hidden_size": self.model.config.hidden_size // self.shard_config.tensor_parallel_size,
@@ -53,11 +55,11 @@ class BloomPolicy(Policy):
                                                              SubModuleReplacementDescription(
                                                                  suffix="self_attention.query_key_value",
                                                                  target_module=col_nn.Linear1D_Col,
-                                                             ),
+                                                                 kwargs={'seq_parallel': use_sequence_parallel}),
                                                              SubModuleReplacementDescription(
                                                                  suffix="self_attention.dense",
                                                                  target_module=col_nn.Linear1D_Row,
-                                                             ),
+                                                                 kwargs={'seq_parallel': use_sequence_parallel}),
                                                              SubModuleReplacementDescription(
                                                                  suffix="self_attention.attention_dropout",
                                                                  target_module=col_nn.DropoutForParallelInput,
@@ -65,11 +67,11 @@ class BloomPolicy(Policy):
                                                              SubModuleReplacementDescription(
                                                                  suffix="mlp.dense_h_to_4h",
                                                                  target_module=col_nn.Linear1D_Col,
-                                                             ),
+                                                                 kwargs={'seq_parallel': use_sequence_parallel}),
                                                              SubModuleReplacementDescription(
                                                                  suffix="mlp.dense_4h_to_h",
                                                                  target_module=col_nn.Linear1D_Row,
-                                                             ),
+                                                                 kwargs={'seq_parallel': use_sequence_parallel}),
                                                          ])
 
             policy[BloomModel] = ModulePolicyDescription(
@@ -116,6 +118,12 @@ class BloomPolicy(Policy):
                                                         policy=policy,
                                                         target_key=BloomBlock)
 
+        if use_sequence_parallel:
+            self.append_or_create_method_replacement(
+                description={'forward': get_bloom_sequence_parallel_forward_fn(self.shard_config)},
+                policy=policy,
+                target_key=BloomModel)
+
         if self.shard_config.enable_flash_attention:
             policy[BloomAttention] = ModulePolicyDescription(method_replacement={
                 'forward': get_bloom_flash_attention_forward(),
@@ -154,7 +162,13 @@ class BloomPolicy(Policy):
 
             layers_per_stage = Policy.distribute_layers(len(module.h), stage_manager.num_stages)
             stage_index = Policy.get_stage_index(layers_per_stage, stage_manager.stage)
-            method_replacement = {'forward': partial(new_forward, stage_manager=stage_manager, stage_index=stage_index)}
+            method_replacement = {
+                'forward':
+                    partial(new_forward,
+                            stage_manager=stage_manager,
+                            stage_index=stage_index,
+                            shard_config=self.shard_config)
+            }
             self.append_or_create_method_replacement(description=method_replacement,
                                                      policy=policy,
                                                      target_key=model_cls)

colossalai/shardformer/shard/shard_config.py

@@ -58,3 +58,4 @@ class ShardConfig:
         self.enable_fused_normalization = True
         self.enable_flash_attention = True
         self.enable_jit_fused = True
+        self.enable_sequence_parallelism = True