[moe] deepseek moe sp support

colossalai/shardformer/modeling/deepseek.py

@@ -1,12 +1,18 @@
-from typing import List, Optional
+import warnings
+from typing import List, Optional, Tuple, Union
 
 import torch
 import torch.distributed as dist
 import torch.nn as nn
 from torch.distributed import ProcessGroup
 from torch.nn import CrossEntropyLoss
-from transformers.modeling_attn_mask_utils import _prepare_4d_causal_attention_mask
+from transformers.cache_utils import Cache, DynamicCache
-from transformers.modeling_outputs import CausalLMOutputWithPast
+from transformers.modeling_attn_mask_utils import (
+    _prepare_4d_causal_attention_mask,
+    _prepare_4d_causal_attention_mask_for_sdpa,
+)
+from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast
+from transformers.models.llama.modeling_llama import apply_rotary_pos_emb
 from transformers.utils import is_flash_attn_2_available, logging
 
 from colossalai.lazy import LazyInitContext
@@ -18,6 +24,11 @@ from colossalai.moe._operation import (
     all_to_all_uneven,
 )
 from colossalai.pipeline.stage_manager import PipelineStageManager
+from colossalai.shardformer.layer._operation import (
+    all_to_all_comm,
+    gather_forward_split_backward,
+    split_forward_gather_backward,
+)
 from colossalai.shardformer.layer.linear import Linear1D_Col, Linear1D_Row
 from colossalai.shardformer.shard import ShardConfig
 from colossalai.shardformer.shard.utils import set_tensors_to_none
@@ -362,7 +373,14 @@ class DeepseekPipelineForwards:
         next_cache = next_decoder_cache if use_cache else None
 
         if stage_manager.is_last_stage():
-            return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)
+            if not return_dict:
+                return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)
+            return BaseModelOutputWithPast(
+                last_hidden_state=hidden_states,
+                past_key_values=next_cache,
+                hidden_states=all_hidden_states,
+                attentions=all_self_attns,
+            )
         # always return dict for imediate stage
         return {
             "hidden_states": hidden_states,
@@ -479,3 +497,276 @@ class DeepseekPipelineForwards:
             hidden_states = outputs.get("hidden_states")
             out["hidden_states"] = hidden_states
             return out
+
+
+def get_deepseek_flash_attention_forward(shard_config, sp_mode=None, sp_size=None, sp_group=None):
+    logger = logging.get_logger(__name__)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        **kwargs,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        if sp_mode is not None:
+            assert sp_mode in ["all_to_all", "split_gather", "ring"], "Invalid sp_mode"
+            assert (sp_size is not None) and (
+                sp_group is not None
+            ), "Must specify sp_size and sp_group for sequence parallel"
+
+        # DeepseekFlashAttention2 attention does not support output_attentions
+        if "padding_mask" in kwargs:
+            warnings.warn(
+                "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`"
+            )
+
+            # overwrite attention_mask with padding_mask
+            attention_mask = kwargs.pop("padding_mask")
+
+        output_attentions = False
+
+        bsz, q_len, _ = hidden_states.size()
+
+        # sp: modify sp_len when sequence parallel mode is ring
+        if sp_mode in ["split_gather", "ring"]:
+            q_len *= sp_size
+
+        rank = dist.get_rank()
+        print(f"{rank=}, hidden states:{hidden_states.shape}")
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        rank = dist.get_rank()
+        print(f"{rank=}, before all to all q:{query_states.shape}, k:{key_states.shape}, v:{value_states.shape}")
+        # sp: all-to-all comminucation when introducing sequence parallel
+        if sp_mode == "all_to_all":
+            query_states = all_to_all_comm(query_states, sp_group)
+            key_states = all_to_all_comm(key_states, sp_group)
+            value_states = all_to_all_comm(value_states, sp_group)
+            bsz, q_len, _ = query_states.size()
+        print(f"{rank=}, after all to all q:{query_states.shape}, k:{key_states.shape}, v:{value_states.shape}")
+        # Flash attention requires the input to have the shape
+        # batch_size x seq_length x head_dim x hidden_dim
+        # therefore we just need to keep the original shape
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        print(f"{rank=}, after view to (b,s,h,d) q:{query_states.shape}, k:{key_states.shape}, v:{value_states.shape}")
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+        query_states, key_states = apply_rotary_pos_emb(
+            query_states, key_states, cos, sin, position_ids, unsqueeze_dim=0
+        )
+        print(f"{rank=}, after rope q:{query_states.shape}, k:{key_states.shape}, v:{value_states.shape}")
+
+        if past_key_value is not None:
+            cache_kwargs = {"sin": sin, "cos": cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        # TODO: These transpose are quite inefficient but Flash Attention requires the layout [batch_size, sequence_length, num_heads, head_dim]. We would need to refactor the KV cache
+        # to be able to avoid many of these transpose/reshape/view.
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+        print(
+            f"{rank=}, after transpose to (b, nh, s, d) q:{query_states.shape}, k:{key_states.shape}, v:{value_states.shape}"
+        )
+        dropout_rate = self.attention_dropout if self.training else 0.0
+
+        # In PEFT, usually we cast the layer norms in float32 for training stability reasons
+        # therefore the input hidden states gets silently casted in float32. Hence, we need
+        # cast them back in the correct dtype just to be sure everything works as expected.
+        # This might slowdown training & inference so it is recommended to not cast the LayerNorms
+        # in fp32. (DeepseekRMSNorm handles it correctly)
+
+        input_dtype = query_states.dtype
+        if input_dtype == torch.float32:
+            # Handle the case where the model is quantized
+            if hasattr(self.config, "_pre_quantization_dtype"):
+                target_dtype = self.config._pre_quantization_dtype
+            elif torch.is_autocast_enabled():
+                target_dtype = torch.get_autocast_gpu_dtype()
+            else:
+                target_dtype = self.q_proj.weight.dtype
+
+            logger.warning_once(
+                f"The input hidden states seems to be silently casted in float32, this might be related to"
+                f" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in"
+                f" {target_dtype}."
+            )
+
+            query_states = query_states.to(target_dtype)
+            key_states = key_states.to(target_dtype)
+            value_states = value_states.to(target_dtype)
+        print(f"{rank=}, before flash attn  q:{query_states.shape}, k:{key_states.shape}, v:{value_states.shape}")
+        attn_output = self._flash_attention_forward(
+            query_states, key_states, value_states, attention_mask, q_len, dropout=dropout_rate
+        )
+        # sp: all-to-all comminucation when introducing sequence parallel
+        if sp_mode == "all_to_all":
+            attn_output = attn_output.reshape(bsz, q_len, self.num_heads * self.head_dim).contiguous()  # (1, 8, 128)
+            attn_output = all_to_all_comm(attn_output, sp_group, scatter_dim=1, gather_dim=2)  # (1, 4, 256)
+        else:
+            attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+    return forward
+
+
+def get_deepseek_flash_attention_model_forward(shard_config, sp_mode=None, sp_size=None, sp_group=None):
+    logger = logging.get_logger(__name__)
+
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, BaseModelOutputWithPast]:
+        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
+
+        # retrieve input_ids and inputs_embeds
+        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[:2]
+        elif inputs_embeds is not None:
+            batch_size, seq_length = inputs_embeds.shape[:2]
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        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`transformers."
+                )
+                use_cache = False
+
+        past_key_values_length = 0
+        if use_cache:
+            use_legacy_cache = not isinstance(past_key_values, Cache)
+            if use_legacy_cache:
+                past_key_values = DynamicCache.from_legacy_cache(past_key_values)
+            past_key_values_length = past_key_values.get_usable_length(seq_length)
+
+        if position_ids is None:
+            device = input_ids.device if input_ids is not None else inputs_embeds.device
+            position_ids = torch.arange(
+                past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
+            )
+            position_ids = position_ids.unsqueeze(0)
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if self._use_flash_attention_2:
+            # 2d mask is passed through the layers
+            attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None
+        elif self._use_sdpa and not output_attentions:
+            # output_attentions=True can not be supported when using SDPA, and we fall back on
+            # the manual implementation that requires a 4D causal mask in all cases.
+            attention_mask = _prepare_4d_causal_attention_mask_for_sdpa(
+                attention_mask,
+                (batch_size, seq_length),
+                inputs_embeds,
+                past_key_values_length,
+            )
+        else:
+            # 4d mask is passed through the layers
+            attention_mask = _prepare_4d_causal_attention_mask(
+                attention_mask, (batch_size, seq_length), inputs_embeds, past_key_values_length
+            )
+
+        if sp_mode in ["ring", "split_gather"]:
+            inputs_embeds = split_forward_gather_backward(inputs_embeds, 1, sp_group)
+        elif sp_mode == "all_to_all":
+            inputs_embeds = split_forward_gather_backward(inputs_embeds, 1, sp_group, 1 / sp_size)
+        # embed positions
+        hidden_states = inputs_embeds
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        next_decoder_cache = None
+
+        for decoder_layer in self.layers:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    decoder_layer.__call__,
+                    hidden_states,
+                    attention_mask,
+                    position_ids,
+                    past_key_values,
+                    output_attentions,
+                    use_cache,
+                )
+            else:
+                layer_outputs = decoder_layer(
+                    hidden_states,
+                    attention_mask=attention_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_values,
+                    output_attentions=output_attentions,
+                    use_cache=use_cache,
+                )
+
+            hidden_states = layer_outputs[0]
+
+            if use_cache:
+                next_decoder_cache = layer_outputs[2 if output_attentions else 1]
+
+            if output_attentions:
+                all_self_attns += (layer_outputs[1],)
+
+        hidden_states = self.norm(hidden_states)
+
+        if sp_mode == "ring" or sp_mode == "split_gather":
+            hidden_states = gather_forward_split_backward(hidden_states, 1, sp_group)
+        elif sp_mode == "all_to_all":
+            hidden_states = gather_forward_split_backward(hidden_states, 1, sp_group, grad_scale=sp_size)
+
+        # add hidden states from the last decoder layer
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+
+        next_cache = None
+        if use_cache:
+            next_cache = next_decoder_cache.to_legacy_cache() if use_legacy_cache else next_decoder_cache
+        if not return_dict:
+            return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=next_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+
+    return forward

colossalai/shardformer/policies/deepseek.py

@@ -7,8 +7,14 @@ from torch import Tensor
 from torch.nn import Module
 
 from colossalai.shardformer.layer import FusedRMSNorm, Linear1D_Col
+from colossalai.shardformer.layer.embedding import PaddingEmbedding, VocabParallelEmbedding1D
 from colossalai.shardformer.layer.linear import Linear1D_Row
-from colossalai.shardformer.modeling.deepseek import DeepseekPipelineForwards, EPDeepseekMoE
+from colossalai.shardformer.modeling.deepseek import (
+    DeepseekPipelineForwards,
+    EPDeepseekMoE,
+    get_deepseek_flash_attention_forward,
+    get_deepseek_flash_attention_model_forward,
+)
 from colossalai.shardformer.policies.base_policy import ModulePolicyDescription, Policy, SubModuleReplacementDescription
 
 __all__ = ["DeepseekPolicy", "DeepseekForCausalLMPolicy"]
@@ -19,6 +25,13 @@ class DeepseekPolicy(Policy):
         pass
 
     def preprocess(self):
+        self.tie_weight = self.tie_weight_check()
+        self.origin_attn_implement = self.model.config._attn_implementation
+        """
+        Because transformers library's bug for AutoModel/AutoConfig, who pop “attn_implement” twice from modeling_utils.py and configuration_utils.py.
+        This bug causes attn_cls to be set to sdpa. Here we assign it to "flash_attention_2".
+        """
+        # self.origin_attn_implement =  "flash_attention_2"
         if self.shard_config.enable_tensor_parallelism:
             # Resize embedding
             vocab_size = self.model.config.vocab_size
@@ -31,17 +44,61 @@ class DeepseekPolicy(Policy):
         return self.model
 
     def module_policy(self) -> Dict[Union[str, nn.Module], ModulePolicyDescription]:
-        policy = {}
 
+        ATTN_IMPLEMENTATION = {
+            "eager": "DeepseekAttention",
+            "flash_attention_2": "DeepseekFlashAttention2",
+            "sdpa": "DeepseekSdpaAttention",
+        }
+        policy = {}
+        print(f"{self.origin_attn_implement=}")
+        attn_cls = ATTN_IMPLEMENTATION[self.origin_attn_implement]
+        sp_mode = self.shard_config.sequence_parallelism_mode or None
+        sp_size = self.shard_config.sequence_parallel_size or None
+        sp_group = self.shard_config.sequence_parallel_process_group or None
+        sp_partial_derived = sp_mode in ["split_gather", "ring"]
+        if sp_mode == "all_to_all":
+            decoder_attribute_replacement = {
+                "num_heads": self.model.config.num_attention_heads // sp_size,
+            }
+            if getattr(self.model.config, "num_key_value_heads", False):
+                decoder_attribute_replacement["num_key_value_heads"] = self.model.config.num_key_value_heads // sp_size
+
+            policy[attn_cls] = ModulePolicyDescription(
+                attribute_replacement=decoder_attribute_replacement,
+            )
         if self.shard_config.enable_sequence_parallelism:
             if self.pipeline_stage_manager is not None:
                 # NOTE: we are replacing model forward for both sequence parallelism and pipeline parallelism
                 # if both are enabled, one of them will be ignored
                 raise NotImplementedError("Sequence parallelism is not supported with pipeline parallelism.")
-            raise NotImplementedError(
+            print(f"{attn_cls=}")
-                "Deepseek dosen't support sequence parallelism now, will ignore the sequence parallelism flag."
+            self.append_or_create_method_replacement(
+                description={
+                    "forward": get_deepseek_flash_attention_forward(self.shard_config, sp_mode, sp_size, sp_group),
+                },
+                policy=policy,
+                target_key=attn_cls,
             )
-
+            if self.pipeline_stage_manager is None:
+                self.append_or_create_method_replacement(
+                    description={
+                        "forward": get_deepseek_flash_attention_model_forward(
+                            self.shard_config,
+                            sp_mode=sp_mode,
+                            sp_size=sp_size,
+                            sp_group=sp_group,
+                        ),
+                    },
+                    policy=policy,
+                    target_key="DeepseekModel",
+                )
+        embedding_cls = None
+        if self.shard_config.enable_tensor_parallelism:
+            embedding_cls = VocabParallelEmbedding1D
+        else:
+            if self.tie_weight:
+                embedding_cls = PaddingEmbedding
         if self.shard_config.enable_tensor_parallelism:
             # tensor parallelism for non-moe params
             assert (
@@ -78,6 +135,16 @@ class DeepseekPolicy(Policy):
                     ),
                 ],
             )
+        if embedding_cls is not None:
+            self.append_or_create_submodule_replacement(
+                description=SubModuleReplacementDescription(
+                    suffix="embed_tokens",
+                    target_module=embedding_cls,
+                    kwargs={"make_vocab_size_divisible_by": self.shard_config.make_vocab_size_divisible_by},
+                ),
+                policy=policy,
+                target_key="DeepseekModel",
+            )
 
         if self.shard_config.ep_group:
             # expert parallel
@@ -105,10 +172,12 @@ class DeepseekPolicy(Policy):
                     SubModuleReplacementDescription(
                         suffix="input_layernorm",
                         target_module=FusedRMSNorm,
+                        kwargs={"sp_partial_derived": sp_partial_derived},
                     ),
                     SubModuleReplacementDescription(
                         suffix="post_attention_layernorm",
                         target_module=FusedRMSNorm,
+                        kwargs={"sp_partial_derived": sp_partial_derived},
                     ),
                 ],
                 policy=policy,
@@ -119,6 +188,7 @@ class DeepseekPolicy(Policy):
                 description=SubModuleReplacementDescription(
                     suffix="norm",
                     target_module=FusedRMSNorm,
+                    kwargs={"sp_partial_derived": sp_partial_derived},
                 ),
                 policy=policy,
                 target_key="DeepseekModel",

tests/kit/model_zoo/transformers/__init__.py

@@ -4,6 +4,7 @@ from .blip2 import *
 from .bloom import *
 from .chatglm2 import *
 from .command import *
+from .deepseek import *
 from .falcon import *
 from .gpt import *
 from .gptj import *

tests/kit/model_zoo/transformers/deepseek.py

@@ -0,0 +1,84 @@
+# modified from tests/kit/model_zoo/transformers/mistral.py
+import torch
+import transformers
+from transformers import AutoConfig
+
+from ..registry import ModelAttribute, model_zoo
+
+# ===============================
+# Register single-sentence Mixtral
+# ===============================
+
+
+def data_gen():
+    # Generated from following code snippet
+    #
+    # from transformers import AutoModelForCausalLM, AutoTokenizer
+    # tokenizer = AutoTokenizer.from_pretrained("mixtralai/Mixtral-7B-v0.1")
+    # input = 'My favourite condiment is vinegar' (last two words repeated to satisfy length requirement)
+    # tokenized_input = tokenizer([input], return_tensors="pt")
+    # input_ids = tokenized_input['input_ids']
+    # attention_mask = tokenized_input['attention_mask']
+    input_ids = torch.tensor([[1, 22, 55, 77, 532, 349, 43, 22]], dtype=torch.int64)
+    attention_mask = torch.tensor([[1, 1, 1, 1, 1, 1, 1, 1]], dtype=torch.int64)
+    return dict(input_ids=input_ids, attention_mask=attention_mask)
+
+
+def data_gen_for_lm():
+    # LM data gen
+    # the `labels` of LM is the token of the output, cause no padding, use `input_ids` as `labels`
+    data = data_gen()
+    data["labels"] = data["input_ids"].clone()
+    return data
+
+
+def data_gen_for_sequence_classification():
+    # sequence classification data gen
+    data = data_gen()
+    data["labels"] = torch.tensor([1], dtype=torch.int64)
+    return data
+
+
+# define output transform function
+output_transform_fn = lambda x: x
+
+# define loss function
+loss_fn_for_mixtral_model = lambda x: x[0].mean()
+loss_fn = lambda x: x.loss
+loss_fn_for_seq_classification = lambda output: output.logits.mean()
+
+
+def init_deepseek():
+
+    config = AutoConfig.from_pretrained(
+        "deepseek-ai/deepseek-moe-16b-base",
+        hidden_size=32,
+        intermediate_size=32,
+        moe_intermediate_size=32,
+        num_hidden_layers=2,
+        num_attention_heads=8,
+        num_key_value_heads=8,
+        # vocab_size=2200,
+        first_k_dense_replace=1,
+        attn_implementation="flash_attention_2",
+        torch_dtype="float16",
+        n_routed_experts=8,
+        trust_remote_code=True,
+    )
+
+    if hasattr(config, "pad_token_id"):
+        config.pad_token_id = config.eos_token_id
+    print(config)
+    model = transformers.AutoModel.from_config(config, trust_remote_code=True)
+
+    return model
+
+
+model_zoo.register(
+    name="transformers_deepseek",
+    model_fn=init_deepseek,
+    data_gen_fn=data_gen,
+    output_transform_fn=output_transform_fn,
+    loss_fn=loss_fn_for_mixtral_model,
+    model_attribute=ModelAttribute(has_control_flow=True),
+)

tests/test_shardformer/test_model/test_shard_deepseek.py

@@ -36,8 +36,8 @@ CHECKED_CONFIG = [  # FOR_WORLD=8
     [
         # (2, 1, 2, 1, 1),  # TODO debug deepseek pp
         # (2, 1, 2, 2, 1),  # TODO debug deepseek pp
-        (2, 1, 1, 2, 1),
+        # (2, 1, 1, 2, 1),
-        # (2, 1, 1, 1, 2),  # TODO support deepseek sp
+        (2, 1, 1, 1, 2),
         # (2, 1, 4, 1, 1),  # TODO debug deepseek pp
         # (4, 1, 2, 1, 1),  # TODO debug deepseek pp
     ],
@@ -69,14 +69,22 @@ def run_zero_with_original_model(config: Tuple[int, ...]):
     booster = Booster(plugin=plugin)
 
     assert pp_size <= NUM_LAYERS, "pp_size should be less than or equal to NUM_LAYERS"
-    config = AutoConfig.from_pretrained("deepseek-ai/deepseek-moe-16b-base", trust_remote_code=True)
+    # config = AutoConfig.from_pretrained("deepseek-ai/deepseek-moe-16b-base", trust_remote_code=True)
-    config.hidden_size = HIDDEN_SIZE_PER_HEAD * NUM_HEADS
+    config = AutoConfig.from_pretrained(
-    config.intermediate_size = HIDDEN_SIZE_PER_HEAD * NUM_HEADS * 2
+        "deepseek-ai/deepseek-moe-16b-base",
-    config.num_hidden_layers = 2
+        hidden_size=HIDDEN_SIZE_PER_HEAD * NUM_HEADS,
-    config.num_attention_heads = NUM_HEADS
+        intermediate_size=HIDDEN_SIZE_PER_HEAD * NUM_HEADS * 2,
-    config.num_key_value_heads = NUM_HEADS
+        moe_intermediate_size=HIDDEN_SIZE_PER_HEAD * NUM_HEADS * 2,
-    config.n_routed_experts = NUM_EXPERTS
+        num_hidden_layers=2,
-    config.num_experts_per_tok = TOP_K
+        num_attention_heads=NUM_HEADS,
+        num_key_value_heads=NUM_HEADS,
+        first_k_dense_replace=1,
+        attn_implementation="flash_attention_2",
+        torch_dtype="float16",
+        n_routed_experts=NUM_EXPERTS,
+        num_experts_per_tok=TOP_K,
+        trust_remote_code=True,
+    )
 
     # init model with the same seed
     seed_all(10086)

tests/test_shardformer/test_model/test_shard_deepseek_ghz.py

@@ -0,0 +1,231 @@
+# modified from test_shard_mistral.py
+import os
+
+import pytest
+import torch
+import torch.distributed as dist
+from torch.testing import assert_close
+
+import colossalai
+from colossalai.booster.plugin.moe_hybrid_parallel_plugin import MoeHybridParallelPlugin
+from colossalai.logging import disable_existing_loggers
+from colossalai.shardformer.layer.utils import Randomizer
+from colossalai.tensor.d_tensor.api import clear_layout_converter
+from colossalai.testing import clear_cache_before_run, parameterize, rerun_if_address_is_in_use, spawn
+from tests.kit.model_zoo import model_zoo
+from tests.test_shardformer.test_model._utils import (
+    build_model_from_hybrid_plugin,
+    check_all_grad_tensors,
+    check_loss,
+    check_output_hidden_state,
+    check_weight,
+    get_grad_tensors_for_check,
+    run_forward_backward_with_hybrid_plugin,
+    unwrap_model,
+)
+
+os.environ["TRANSFORMERS_NO_ADVISORY_WARNINGS"] = "true"
+
+
+def check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn, test_config):
+    # TODO: SGD failed for full dp
+    org_model, org_optimizer, sharded_model, sharded_optimizer, criterion, booster = build_model_from_hybrid_plugin(
+        model_fn, loss_fn, test_config, pluggin_cls=MoeHybridParallelPlugin, optim_class=torch.optim.SGD
+    )
+
+    org_model = org_model.to(torch.float16)
+    org_loss, org_output, sharded_loss, sharded_output = run_forward_backward_with_hybrid_plugin(
+        org_model, sharded_model, sharded_optimizer, data_gen_fn, output_transform_fn, criterion, booster
+    )
+    stage_manager = booster.plugin.stage_manager
+    tp_group = booster.plugin.tp_group
+
+    # check last hidden state & loss
+    if stage_manager is None or stage_manager.is_last_stage():
+        if test_config["precision"] == "fp32":
+            atol, rtol = 1e-5, 1e-3
+        else:
+            atol, rtol = 5e-3, 5e-3
+
+        check_loss(org_loss, sharded_loss, atol=atol, rtol=rtol)
+        check_output_hidden_state(org_output, sharded_output, stage_manager, atol, rtol)
+
+    # unwrap model
+    mixtral_model = unwrap_model(org_model, "DeepseekModel", "model")
+    shard_mixtral_model = unwrap_model(sharded_model, "DeepseekModel", "model")
+
+    row_layer_for_check = ["layers[0].self_attn.q_proj", "embed_tokens"]
+    col_layer_for_check = ["layers[0].self_attn.o_proj"]
+
+    name_to_p = {n: p for n, p in mixtral_model.named_parameters()}
+    # Check the grad when using ZeRO-1 and ZeRO-2
+    if (
+        # booster.plugin.zero_stage in [1, 2]
+        booster.plugin.shard_config.enable_sequence_parallelism
+        and booster.plugin.shard_config.sequence_parallelism_mode == "all_to_all"
+    ):
+        rank = dist.get_rank()
+        for n, p in shard_mixtral_model.named_parameters():
+            zero_grad = sharded_optimizer.get_param_grad(p)
+            if name_to_p[n].grad is None:
+                name_to_p[n].grad = torch.zeros_like(name_to_p[n].data)
+                continue
+            assert_close(name_to_p[n].grad, zero_grad, atol=5e-3, rtol=5e-3, check_dtype=False)
+
+    # Save gradient tensors for comparison between the original model and the sharded model before optimizer step.
+    grads_to_check = {}
+    if (stage_manager is None or stage_manager.is_first_stage()) and booster.plugin.zero_stage == 0:
+        if test_config["precision"] == "fp32":
+            atol, rtol = 5e-5, 1e-4
+        else:
+            atol, rtol = 5e-3, 5e-3
+        row_layer_grads = get_grad_tensors_for_check(
+            mixtral_model,
+            shard_mixtral_model,
+            row_layer_for_check,
+            tp_group,
+            atol=atol,
+            rtol=rtol,
+            dim=0,
+            verbose=False,
+        )
+        col_layer_grads = get_grad_tensors_for_check(
+            mixtral_model,
+            shard_mixtral_model,
+            col_layer_for_check,
+            tp_group,
+            atol=atol,
+            rtol=rtol,
+            dim=1,
+            verbose=False,
+        )
+        grads_to_check.update(col_layer_grads)
+        grads_to_check.update(row_layer_grads)
+
+    # check grads
+    check_all_grad_tensors(grads_to_check)
+
+    for n, p in shard_mixtral_model.named_parameters():
+        assert_close(name_to_p[n], p, atol=5e-3, rtol=5e-3, check_dtype=False)
+
+    # optimizer executes step
+    org_optimizer.step()
+    sharded_optimizer.step()
+
+    for n, p in shard_mixtral_model.named_parameters():
+        assert_close(name_to_p[n], p, atol=5e-3, rtol=5e-3, check_dtype=False)
+
+    # check weights
+    if stage_manager is None or stage_manager.is_first_stage():
+        if test_config["precision"] == "fp32":
+            atol, rtol = 2e-4, 1e-3
+        else:
+            atol, rtol = 5e-3, 5e-3
+        try:
+            check_weight(
+                mixtral_model,
+                shard_mixtral_model,
+                col_layer_for_check,
+                tp_group,
+                atol=atol,
+                rtol=rtol,
+                dim=1,
+                verbose=False,
+            )
+        except Exception as e:
+            rank = dist.get_rank()
+            print(f"{rank=}, Failed config: {test_config}")
+            raise e
+
+    torch.cuda.empty_cache()
+
+
+@parameterize(
+    "test_config",
+    [
+        # {
+        #     "tp_size": 1,
+        #     "pp_size": 1,
+        #     "num_microbatches": 2,
+        #     "ep_size": 2,
+        #     "zero_stage": 0,
+        #     "overlap_communication": False,
+        #     "precision": "fp16",
+        # },  # [dp(4)] + [moe_dp(4)]
+        # {
+        #     "tp_size": 1,
+        #     "pp_size": 2,
+        #     "num_microbatches": 2,
+        #     "ep_size": 2,
+        #     "zero_stage": 1,
+        #     "overlap_communication": False,
+        #     "precision": "fp32",
+        # },  # [dp(2) + pp(2)] + [moe_pp(2)]
+        # {
+        #     "tp_size": 1,
+        #     "pp_size": 2,
+        #     "ep_size": 2,
+        #     "num_microbatches": 2,
+        #     "zero_stage": 1,
+        #     "overlap_communication": False,
+        #     "precision": "fp16",
+        #     "initial_scale": 1,
+        #     "find_unused_parameters": True,
+        # },  # [pp(2) + tp(2)] + [pp(2), replicate(2)] pass
+        {  # Ulysess + Flash attention
+            "tp_size": 1,
+            "pp_size": 1,
+            "sp_size": 2,
+            "ep_size": 2,
+            "enable_sequence_parallelism": True,
+            "sequence_parallelism_mode": "all_to_all",
+            "zero_stage": 1,
+            "overlap_communication": False,
+            "precision": "fp16",
+            "initial_scale": 1,
+            "find_unused_parameters": True,
+        },
+        # {
+        #     "tp_size": 1,
+        #     "pp_size": 1,
+        #     "ep_size": 2,
+        #     "zero_stage": 0,
+        #     "overlap_communication": False,
+        #     "precision": "fp32",
+        # },  # [dp(4)] + [ep(2) + moe_tp(2)]
+        # {
+        #     "tp_size": 1,
+        #     "pp_size": 1,
+        #     "ep_size": 4,
+        #     "overlap_communication": False,
+        #     "zero_stage": 0,
+        #     "precision": "fp32"
+        # },  # full dp for non-moe and full ep for moe
+    ],
+)
+def run_deepseek_test(test_config):
+    sub_model_zoo = model_zoo.get_sub_registry("transformers_deepseek")
+
+    for name, (model_fn, data_gen_fn, output_transform_fn, loss_fn, _) in sub_model_zoo.items():
+        check_forward_backward(model_fn, data_gen_fn, output_transform_fn, loss_fn, test_config)
+
+    clear_layout_converter()
+    Randomizer.reset_index()
+    torch.cuda.empty_cache()
+
+
+def check_deepseek(rank, world_size, port):
+    disable_existing_loggers()
+    colossalai.launch(rank=rank, world_size=world_size, host="localhost", port=port, backend="nccl")
+    run_deepseek_test()
+
+
+@pytest.mark.dist
+@rerun_if_address_is_in_use()
+@clear_cache_before_run()
+def test_mixtral():
+    spawn(check_deepseek, 4)
+
+
+if __name__ == "__main__":
+    test_mixtral()