[shardformer] fix gathering output when using tensor parallelism (#5431)

* fix

* padding vocab_size when using pipeline parallellism

padding vocab_size when using pipeline parallellism

fix

fix

* fix

* fix

fix

fix

* fix gather output

* fix

* fix

* fix

fix resize embedding

fix resize embedding

* fix resize embedding

fix

* revert

* revert

* revert

colossalai/booster/plugin/hybrid_parallel_plugin.py

@@ -199,7 +199,12 @@ def get_param_info(optim: Optimizer):
 
     if optim is None:
         return {}
-    param_info = {"param_groups": [], "param2id": {}, "id2param": {}, "param2shape": {}}
+    param_info = {
+        "param_groups": [],
+        "param2id": {},
+        "id2param": {},
+        "param2shape": {},
+    }
     start_index = 0
     for group in optim.param_groups:
         packed_group = {k: v for k, v in group.items() if k != "params"}
@@ -899,6 +904,7 @@ class HybridParallelPlugin(PipelinePluginBase):
         enable_jit_fused (bool, optional): Whether to switch on JIT in Shardformer. Default to False.
         enable_sequence_parallelism (bool): Whether to turn on sequence parallelism in Shardformer. Defaults to False.
         enable_sequence_overlap (bool): Whether to turn on sequence overlap in Shardformer. Defaults to False.
+        parallel_output (bool): Whether to keep the output parallel when enabling tensor parallelism. Default to True.
         num_microbatches (int, optional): Number of microbatches when using pipeline parallelism. Defaults to None.
         microbatch_size (int, optional): Microbatch size when using pipeline parallelism.
             Either ``num_microbatches`` or ``microbatch_size`` should be provided if using pipeline.
@@ -939,6 +945,7 @@ def __init__(
         enable_jit_fused: bool = False,
         enable_sequence_parallelism: bool = False,
         enable_sequence_overlap: bool = False,
+        parallel_output: bool = True,
         num_microbatches: Optional[int] = None,
         microbatch_size: Optional[int] = None,
         initial_scale: float = 2**16,
@@ -1035,6 +1042,7 @@ def __init__(
             enable_jit_fused=self.enable_jit_fused,
             enable_sequence_parallelism=enable_sequence_parallelism,
             enable_sequence_overlap=enable_sequence_overlap,
+            parallel_output=parallel_output,
         )
         self.amp_config = dict(
             initial_scale=initial_scale,

colossalai/shardformer/modeling/gpt2.py

@@ -25,6 +25,7 @@
 from colossalai.shardformer.shard import ShardConfig
 
 from ..layer import cross_entropy_1d
+from ..layer._operation import gather_forward_split_backward
 
 
 class GPT2PipelineForwards:
@@ -337,6 +338,9 @@ def gpt2_lmhead_model_forward(
             else:
                 loss = loss_fct(shift_logits, shift_labels)
 
+            if not shard_config.parallel_output:
+                lm_logits = gather_forward_split_backward(lm_logits, -1, shard_config.tensor_parallel_process_group)
+
         if not return_dict:
             output = (lm_logits,) + outputs[1:]
             return ((loss,) + output) if loss is not None else output
@@ -793,11 +797,12 @@ def forward(
             scale = scale * (1 / float(self.layer_idx + 1))
 
         # use coloattention
-        attention = ColoAttention(
-            embed_dim=self.embed_dim, num_heads=self.num_heads, dropout=self.attn_dropout.p, scale=scale
-        )
+        if not hasattr(self, "attention"):
+            self.attention = ColoAttention(
+                embed_dim=self.embed_dim, num_heads=self.num_heads, dropout=self.attn_dropout.p, scale=scale
+            )
 
-        attn_output = attention(query, key, value, attn_mask=flash_attention_mask, attn_mask_type=attn_mask_type)
+        attn_output = self.attention(query, key, value, attn_mask=flash_attention_mask, attn_mask_type=attn_mask_type)
 
         attn_output = self.c_proj(attn_output)
         attn_output = self.resid_dropout(attn_output)
@@ -1083,6 +1088,9 @@ def forward(
             else:
                 loss = loss_fct(shift_logits, shift_labels)
 
+        if not shard_config.parallel_output:
+            lm_logits = gather_forward_split_backward(lm_logits, -1, shard_config.tensor_parallel_process_group)
+
         if not return_dict:
             output = (lm_logits,) + transformer_outputs[1:]
             return ((loss,) + output) if loss is not None else output

colossalai/shardformer/modeling/llama.py

@@ -16,7 +16,7 @@
 from colossalai.shardformer.shard import ShardConfig
 
 from ..layer import cross_entropy_1d
-from ..layer._operation import _gather
+from ..layer._operation import gather_forward_split_backward
 
 try:
     from transformers.models.llama.modeling_llama import _prepare_4d_causal_attention_mask
@@ -290,7 +290,7 @@ def llama_for_causal_lm_forward(
                     loss = loss_fct(shift_logits, shift_labels)
 
             if not shard_config.parallel_output:
-                logits = _gather(logits, -1, shard_config.tensor_parallel_process_group)
+                logits = gather_forward_split_backward(logits, -1, shard_config.tensor_parallel_process_group)
 
             if not return_dict:
                 output = (logits,) + outputs[1:]
@@ -485,8 +485,9 @@ def forward(
             flash_attention_mask = ~(attention_mask[:, :, -1].squeeze(1).to(torch.bool)).contiguous()
             attn_mask_type = AttnMaskType.paddedcausal
 
-        attention = ColoAttention(embed_dim=self.hidden_size, num_heads=self.num_heads)
-        attn_output = attention(
+        if not hasattr(self, "attention"):
+            self.attention = ColoAttention(embed_dim=self.hidden_size, num_heads=self.num_heads)
+        attn_output = self.attention(
             query_states,
             key_states,
             value_states,
@@ -593,7 +594,7 @@ def forward(
                 loss = loss_fct(shift_logits, shift_labels)
 
         if not shard_config.parallel_output:
-            logits = _gather(logits, -1, shard_config.tensor_parallel_process_group)
+            logits = gather_forward_split_backward(logits, -1, shard_config.tensor_parallel_process_group)
 
         if not return_dict:
             output = (logits,) + outputs[1:]

colossalai/shardformer/policies/base_policy.py

@@ -242,4 +242,4 @@ def get_stage_index(
             end_idx = num_layers_per_stage_accumulated[stage + model_chunk * num_stages + 1]
             stage_indices.append([start_idx, end_idx])
 
-        return stage_indices[0] if num_model_chunks == 1 else stage_indices
+        return stage_indices[0] if num_model_chunks == 1 else stage_indices

colossalai/shardformer/shard/shard_config.py

@@ -34,8 +34,10 @@ class ShardConfig:
     enable_all_optimization: bool = False
     enable_sequence_parallelism: bool = False
     enable_sequence_overlap: bool = False
-    parallel_output = True
+    parallel_output: bool = True
     extra_kwargs: Dict[str, Any] = field(default_factory=dict)
+    # TODO padding vocab
+    # make_vocab_size_divisible_by: int = 128
     # pipeline_parallel_size: int
     # data_parallel_size: int
     # tensor_parallel_mode: Literal['1d', '2d', '2.5d', '3d']

tests/test_booster/test_plugin/test_3d_plugin.py

@@ -260,7 +260,7 @@ def run_grad_acc_test(test_args):
         origin_model, origin_optimizer, dataloader=dataloader
     )
     for p1, p2 in zip(model.unwrap().parameters(), origin_model.unwrap().parameters()):
-        assert_close(p1.to(p2.dtype), p2, atol=1e-2, rtol=1e-2)
+        assert_close(p1.to(p2.dtype), p2, atol=1e-2, rtol=1e-2)      
 
 
 def run_dist(rank, world_size, port, early_stop: bool = True):