models.py

#!/usr/bin/env python
# coding=utf-8
'''
Description  :  
Author       : Azure-Tang
Date         : 2024-07-25 11:25:24
Version      : 1.0.0
LastEditors  : Azure 
LastEditTime : 2024-08-14 14:53:05
Copyright (c) 2024 by KVCache.AI, All Rights Reserved. 
'''

import inspect
import math
from typing import List, Optional, Tuple, Union
import time
import torch
import torch.nn.functional as F
import torch.utils.checkpoint
from torch import nn
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss

from transformers.activations import ACT2FN
from transformers.cache_utils import Cache, DynamicCache, StaticCache
from transformers.modeling_attn_mask_utils import (
    AttentionMaskConverter,
)
from transformers.modeling_outputs import (
    MoeCausalLMOutputWithPast,
    MoeModelOutputWithPast,
    SequenceClassifierOutputWithPast,
    TokenClassifierOutput,
)
from transformers.modeling_utils import PreTrainedModel
from transformers.utils import (
    add_start_docstrings,
    add_start_docstrings_to_model_forward,
    is_flash_attn_2_available,
    is_flash_attn_greater_or_equal_2_10,
    logging,
    replace_return_docstrings,
)
from ktransformers.models.modeling_qwen2_moe import Qwen2MoeSparseMoeBlock, Qwen2MoeMLP, Qwen2MoeDecoderLayer
from ktransformers.models.modeling_deepseek import BaseModelOutputWithPast, DeepseekV2DecoderLayer, DeepseekV2MoE
from transformers.models.qwen2_moe.configuration_qwen2_moe import Qwen2MoeConfig
from ktransformers.operators.base_operator import BaseInjectedModule
from ktransformers.util.utils import InferenceState
from ktransformers.util.custom_gguf import GGUFLoader
from transformers.configuration_utils import PretrainedConfig

if is_flash_attn_2_available():
    from flash_attn import flash_attn_func, flash_attn_varlen_func
    from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input  # noqa

    _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters)

logger = logging.get_logger(__name__)

_CHECKPOINT_FOR_DOC = "Qwen/Qwen1.5-MoE-A2.7B"
_CONFIG_FOR_DOC = "Qwen2MoeConfig"

QWEN2MOE_START_DOCSTRING = r"""
    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
    etc.)

    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
    and behavior.

    Parameters:
        config ([`Qwen2MoeConfig`]):
            Model configuration class with all the parameters of the model. Initializing with a config file does not
            load the weights associated with the model, only the configuration. Check out the
            [`~PreTrainedModel.from_pretrained`] method to load the model weights.
"""

QWEN2MOE_INPUTS_DOCSTRING = r"""
    Args:
        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
            it.

            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.

            [What are input IDs?](../glossary#input-ids)
        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:

            - 1 for tokens that are **not masked**,
            - 0 for tokens that are **masked**.

            [What are attention masks?](../glossary#attention-mask)

            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.

            If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see
            `past_key_values`).

            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
            information on the default strategy.

            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.
        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
            config.n_positions - 1]`.

            [What are position IDs?](../glossary#position-ids)
        past_key_values (`Cache` or `tuple(tuple(torch.FloatTensor))`, *optional*):
            Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
            blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values`
            returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`.

            Two formats are allowed:
            - a [`~cache_utils.Cache`] instance;
            - Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of
            shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`). This is also known as the legacy
            cache format.

            The model will output the same cache format that is fed as input. If no `past_key_values` are passed, the
            legacy cache format will be returned.

            If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't
            have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `input_ids`
            of shape `(batch_size, sequence_length)`.
        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
            model's internal embedding lookup matrix.
        use_cache (`bool`, *optional*):
            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
            `past_key_values`).
        output_attentions (`bool`, *optional*):
            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
            tensors for more detail.
        output_hidden_states (`bool`, *optional*):
            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
            more detail.
        output_router_logits (`bool`, *optional*):
            Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
            should not be returned during inference.
        return_dict (`bool`, *optional*):
            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
        cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
            Indices depicting the position of the input sequence tokens in the sequence. Contrarily to `position_ids`,
            this tensor is not affected by padding. It is used to update the cache in the correct position and to infer
            the complete sequence length.
"""

@add_start_docstrings(
    "The bare Qwen2MoE Model outputting raw hidden-states without any specific head on top.",
    QWEN2MOE_START_DOCSTRING,
)
class KQwen2MoeModel(BaseInjectedModule):
    """
    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`Qwen2MoeDecoderLayer`]

    Args:
        config: Qwen2MoeConfig
    """
    def __init__(
        self,
        key: str,
        gguf_loader : GGUFLoader,
        config: PretrainedConfig,
        orig_module: nn.Module,
        device: str = "cuda",
        per_layer_prefill_intput_threshold: int = 30000, # if None, no per-layer prefill
        transfer_map: dict = None,
        **kwargs,
    ):
        BaseInjectedModule.__init__(self, key, gguf_loader, config, orig_module, device, **kwargs)
        self.per_layer_prefill_intput_threshold = per_layer_prefill_intput_threshold
        self.transfer_map = transfer_map
        self.stream_device_map = dict()

    @add_start_docstrings_to_model_forward(QWEN2MOE_INPUTS_DOCSTRING)
    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,
        output_router_logits: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        cache_position: Optional[torch.LongTensor] = None,
        per_layer_prefill_intput_threshold: int | None = None, # if None or 0, close per-layer prefill
    ) -> Union[Tuple, MoeModelOutputWithPast]:
        # print(f'Total length of input_ids: {input_ids.size(1)}, {input_ids.size()}')

        if per_layer_prefill_intput_threshold is None: per_layer_prefill_intput_threshold = self.per_layer_prefill_intput_threshold
        per_layer_prefill_flag = False
        seq_lenth = inputs_embeds.size(1) if inputs_embeds is not None else input_ids.size(1)
        if per_layer_prefill_intput_threshold and per_layer_prefill_intput_threshold < seq_lenth:
            per_layer_prefill_flag = True
            for layer in self.layers:
                self.load_layer_to(layer, InferenceState.UNLOAD)
        else:
            pass
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_router_logits = (
            output_router_logits if output_router_logits is not None else self.config.output_router_logits
        )
        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 None) ^ (inputs_embeds is not None):
            raise ValueError(
                "You cannot specify both input_ids and inputs_embeds at the same time, and must specify either one"
            )

        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

        use_legacy_cache = False
        if use_cache and not isinstance(past_key_values, Cache):
            use_legacy_cache = True
            past_key_values = DynamicCache.from_legacy_cache(past_key_values)
            logger.warning_once(
                "We detected that you are passing `past_key_values` as a tuple and this is deprecated and will be removed in v4.43. "
                "Please use an appropriate `Cache` class (https://huggingface.co/docs/transformers/v4.41.3/en/internal/generation_utils#transformers.Cache)"
            )

        if inputs_embeds is None:
            input_ids = input_ids.to("cpu")
            inputs_embeds = self.embed_tokens(input_ids)
            inputs_embeds = inputs_embeds.to("cuda")

        if cache_position is None:
            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
            cache_position = torch.arange(
                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
            )
        if position_ids is None:
            position_ids = cache_position.unsqueeze(0)

        causal_mask = self._update_causal_mask(
            attention_mask, inputs_embeds, cache_position, past_key_values, output_attentions
        )

        hidden_states = inputs_embeds

        # decoder layers
        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None
        all_router_logits = () if output_router_logits else None
        next_decoder_cache = None

        for i, decoder_layer in enumerate(self.layers):
            if self.transfer_map is not None and i in self.transfer_map: 
                prev_stream = torch.cuda.current_stream()
                cur_device = self.transfer_map[i]
                if cur_device not in self.stream_device_map:
                    self.stream_device_map[cur_device] = torch.cuda.Stream(cur_device)
                torch.cuda.set_device(cur_device)
                self.stream_device_map[cur_device].wait_stream(prev_stream)
                torch.cuda.set_stream(self.stream_device_map[cur_device])
                hidden_states = hidden_states.to(self.transfer_map[i], non_blocking = True)
                causal_mask = causal_mask.to(self.transfer_map[i], non_blocking = True) if causal_mask is not None else None
                position_ids = position_ids.to(self.transfer_map[i], non_blocking = True) if position_ids is not None else None
                cache_position = cache_position.to(self.transfer_map[i], non_blocking = True) if cache_position is not None else None
                
            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,
                    causal_mask,
                    position_ids,
                    past_key_values,
                    output_attentions,
                    output_router_logits,
                    use_cache,
                    cache_position,
                )
            else:
                if per_layer_prefill_flag:
                    # print(f"to gpu")
                    self.load_layer_to(decoder_layer, InferenceState.PREFILL)
                    torch.cuda.empty_cache()
                layer_outputs = decoder_layer(
                    hidden_states,
                    attention_mask=causal_mask,
                    position_ids=position_ids,
                    past_key_value=past_key_values,
                    output_attentions=output_attentions,
                    output_router_logits=output_router_logits,
                    use_cache=use_cache,
                    cache_position=cache_position,
                )
                if per_layer_prefill_flag:
                    # print(f"to cpu")
                    self.load_layer_to(decoder_layer, InferenceState.UNLOAD)
                    torch.cuda.empty_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],)

            if output_router_logits and layer_outputs[-1] is not None:
                all_router_logits += (layer_outputs[-1],)

        hidden_states = self.norm(hidden_states)


        if per_layer_prefill_flag:
            per_layer_prefill_flag = False
            for layer in self.layers:
                self.load_layer_to(layer, InferenceState.GENERATE)
        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, all_router_logits]
                if v is not None
            )
        return MoeModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=next_cache,
            hidden_states=all_hidden_states,
            attentions=all_self_attns,
            router_logits=all_router_logits,
        )

    def load_layer_to(self,  layer:Qwen2MoeDecoderLayer, target: InferenceState):
        assert isinstance(layer, Qwen2MoeDecoderLayer), "module should be nn.ModuleList of decoder layers"

        # TODO Support restore to original device, not only cuda
        device = "cpu" if target == InferenceState.UNLOAD else "cuda" 

        # attn
        layer.self_attn.q_proj.set_inference_mode(target)
        layer.self_attn.k_proj.set_inference_mode(target)
        layer.self_attn.v_proj.set_inference_mode(target)
        layer.self_attn.o_proj.set_inference_mode(target)
        layer.self_attn.rotary_emb = layer.self_attn.rotary_emb.to(device)

        # mlp
        if isinstance(layer.mlp, Qwen2MoeSparseMoeBlock):
            layer.mlp.gate.set_inference_mode(target)
            layer.mlp.experts.set_inference_mode(target)
            layer.mlp.shared_expert.gate_proj.set_inference_mode(target)
            layer.mlp.shared_expert.up_proj.set_inference_mode(target)
            layer.mlp.shared_expert.down_proj.set_inference_mode(target)
            layer.mlp.shared_expert.act_fn.to(device)
            layer.mlp.shared_expert_gate.to(device)
        else:
            layer.mlp.gate_proj.set_inference_mode(target)
            layer.mlp.up_proj.set_inference_mode(target)
            layer.mlp.down_proj.set_inference_mode(target)
            layer.mlp.act_fn.to(device)
        # layer norm
        layer.input_layernorm.to(device)
        layer.post_attention_layernorm.to(device)


DeepseekV2_INPUTS_DOCSTRING = r"""
    Args:
        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
            it.

            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.

            [What are input IDs?](../glossary#input-ids)
        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:

            - 1 for tokens that are **not masked**,
            - 0 for tokens that are **masked**.

            [What are attention masks?](../glossary#attention-mask)

            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.

            If `past_key_values` is used, optionally only the last `input_ids` have to be input (see
            `past_key_values`).

            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
            information on the default strategy.

            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.
        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
            config.n_positions - 1]`.

            [What are position IDs?](../glossary#position-ids)
        past_key_values (`Cache` or `tuple(tuple(torch.FloatTensor))`, *optional*):
            Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
            blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values`
            returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`.

            Two formats are allowed:
            - a [`~cache_utils.Cache`] instance;
            - Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of
            shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`). This is also known as the legacy
            cache format.

            The model will output the same cache format that is fed as input. If no `past_key_values` are passed, the
            legacy cache format will be returned.

            If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't
            have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `input_ids`
            of shape `(batch_size, sequence_length)`.
        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
            model's internal embedding lookup matrix.
        use_cache (`bool`, *optional*):
            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
            `past_key_values`).
        output_attentions (`bool`, *optional*):
            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
            tensors for more detail.
        output_hidden_states (`bool`, *optional*):
            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
            more detail.
        return_dict (`bool`, *optional*):
            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
"""


class KDeepseekV2Model(BaseInjectedModule):
    """
    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`DeepseekV2DecoderLayer`]

    Args:
        config: DeepseekV2Config
    """
    def __init__(
        self,
        key: str,
        gguf_loader : GGUFLoader,
        config: PretrainedConfig,
        orig_module: nn.Module,
        device: str = "cuda",
        per_layer_prefill_intput_threshold: int = 30000, # if None, no per-layer prefill
        transfer_map: dict = None,
        **kwargs,
    ):
        BaseInjectedModule.__init__(self, key, gguf_loader, config, orig_module, device, **kwargs)
        self.per_layer_prefill_intput_threshold = per_layer_prefill_intput_threshold
        self.transfer_map = transfer_map
        self.stream_device_map = dict()

    @add_start_docstrings_to_model_forward(DeepseekV2_INPUTS_DOCSTRING)
    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,
        cache_position: Optional[torch.LongTensor] = None,
        per_layer_prefill_intput_threshold: int | None = None, # if None, no per-layer prefill
    ) -> Union[Tuple, BaseModelOutputWithPast]:
        if per_layer_prefill_intput_threshold is None: per_layer_prefill_intput_threshold = self.per_layer_prefill_intput_threshold
        per_layer_prefill_flag = False
        seq_lenth = inputs_embeds.size(1) if inputs_embeds is not None else input_ids.size(1)
        if per_layer_prefill_intput_threshold and per_layer_prefill_intput_threshold < seq_lenth:
            per_layer_prefill_flag = True
            for layer in self.layers:
                self.load_layer_to(layer,  InferenceState.UNLOAD)
            torch.cuda.empty_cache()
        else:
            pass
        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 cache_position is None:
            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
            cache_position = torch.arange(
                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
            )

        if position_ids is None:
            position_ids = cache_position.unsqueeze(0)

        if inputs_embeds is None:
            org_device = input_ids.device
            input_ids = input_ids.to("cpu")
            inputs_embeds = self.embed_tokens(input_ids)
            input_ids = input_ids.to(org_device)


        causal_mask = self._update_causal_mask(
            attention_mask, inputs_embeds, cache_position, past_key_values, output_attentions
        )

        # embed positions
        hidden_states = inputs_embeds
        if per_layer_prefill_flag:
            print(f'Total length of input_ids: {hidden_states.size(1)}')

        # decoder layers
        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None
        next_decoder_cache = None

        t_gpu = 0
        t_cpu = 0
        t_f = 0

        for i, decoder_layer in enumerate(self.layers):
            if self.transfer_map is not None and i in self.transfer_map: 
                prev_stream = torch.cuda.current_stream()
                cur_device = self.transfer_map[i]
                if cur_device not in self.stream_device_map:
                    self.stream_device_map[cur_device] = torch.cuda.Stream(cur_device)
                torch.cuda.set_device(cur_device)
                self.stream_device_map[cur_device].wait_stream(prev_stream)
                torch.cuda.set_stream(self.stream_device_map[cur_device])
                hidden_states = hidden_states.to(self.transfer_map[i], non_blocking = True)
                causal_mask = causal_mask.to(self.transfer_map[i], non_blocking = True) if causal_mask is not None else None
                position_ids = position_ids.to(self.transfer_map[i], non_blocking = True) if position_ids is not None else None
                cache_position = cache_position.to(self.transfer_map[i], non_blocking = True) if cache_position is not None else None

            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,
                    causal_mask,
                    position_ids,
                    past_key_values,
                    output_attentions,
                    use_cache,
                    cache_position,
                )
            else:
                t3 = time.time()
                if per_layer_prefill_flag:
                    # print(f"to gpu")
                    self.load_layer_to(decoder_layer, InferenceState.PREFILL)
                    torch.cuda.empty_cache()
                t4 = time.time()
                layer_outputs = decoder_layer(
                    hidden_states,
                    attention_mask=causal_mask,
                    position_ids=position_ids,
                    past_key_value=past_key_values,
                    output_attentions=output_attentions,
                    use_cache=use_cache,
                    cache_position=cache_position,
                )
                t5 = time.time()
                if per_layer_prefill_flag:
                    # print(f"to cpu")
                    self.load_layer_to(decoder_layer,  InferenceState.UNLOAD)
                    torch.cuda.empty_cache()
                t6 = time.time()
            t_gpu += t4-t3
            t_cpu += t6-t5
            t_f += t5-t4

            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 per_layer_prefill_flag:
            t6 = time.time()
            # print(f"restore")
            per_layer_prefill_flag = False
            for layer in self.layers:
                self.load_layer_to(layer, InferenceState.GENERATE)
            torch.cuda.empty_cache()
            t7 = time.time()

            print(f"total time: {t7-t3}, \n layer num{len(self.layers)}, gpu time: {t_gpu}, cpu time: {t_cpu}, forward time: {t_f}, restore time: {t7-t6}")

        # 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,
        )

    def load_layer_to(self,  layer: DeepseekV2DecoderLayer, target: InferenceState):
        assert isinstance(layer, DeepseekV2DecoderLayer), "module should be nn.ModuleList of decoder layers"

        # TODO Support restore to original device, not only cuda
        device = "cpu" if target == InferenceState.UNLOAD else "cuda" 

        # TODO Support DFS to auto use {to, set_inference_mode} according to the module type

        # attn
        layer.self_attn.to(device) #

        # mlp
        if isinstance(layer.mlp, DeepseekV2MoE):
            layer.mlp.gate.to(device)
            layer.mlp.experts.set_inference_mode(target)
            layer.mlp.shared_experts.gate_proj.set_inference_mode(target)
            layer.mlp.shared_experts.up_proj.set_inference_mode(target)
            layer.mlp.shared_experts.down_proj.set_inference_mode(target)
            layer.mlp.shared_experts.act_fn.to(device)
            # layer.mlp.shared_expert_gate.to(device)
        else:
            layer.mlp.gate_proj.set_inference_mode(target)
            layer.mlp.up_proj.set_inference_mode(target)
            layer.mlp.down_proj.set_inference_mode(target)
            layer.mlp.act_fn.to(device)
        # layer norm
        layer.input_layernorm.to(device)
        layer.post_attention_layernorm.to(device)