patches/modelling_qwen2_infer_gpu.py

import math
import torch
import torch.nn as nn
from typing import List, Optional, Tuple, Union
import transformers.models
from transformers.models.qwen2.modeling_qwen2 import (
    Qwen2RotaryEmbedding,
    Qwen2ForCausalLM,
    Qwen2MLP,
    Qwen2RMSNorm,
    apply_rotary_pos_emb,
    repeat_kv,
    _prepare_4d_causal_attention_mask_with_cache_position,
    )
from transformers.utils import logging
from transformers.modeling_outputs import CausalLMOutputWithPast
from transformers.cache_utils import Cache, StaticCache, SlidingWindowCache
from transformers.models.qwen2.configuration_qwen2 import Qwen2Config
from .utils import InferTaskCode

logger = logging.get_logger(__name__)

_GPU_QWEN_TORCH_COMPILE = True

# ===================================================================
# =============================Attention=============================
# ===================================================================
class GPUQwen2Attention(nn.Module):
    """
    Multi-headed attention from 'Attention Is All You Need' paper. Modified to use sliding window attention: Longformer
    and "Generating Long Sequences with Sparse Transformers".
    """
    def __init__(self, config: Qwen2Config, layer_idx: Optional[int] = None):
        super().__init__()
        self.config = config
        self.layer_idx = layer_idx
        if layer_idx is None:
            logger.warning_once(
                f"Instantiating {self.__class__.__name__} without passing `layer_idx` is not recommended and will "
                "to errors during the forward call, if caching is used. Please make sure to provide a `layer_idx` "
                "when creating this class."
            )

        self.hidden_size = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.head_dim = self.hidden_size // self.num_heads
        self.num_key_value_heads = config.num_key_value_heads
        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
        self.max_position_embeddings = config.max_position_embeddings
        self.rope_theta = config.rope_theta
        self.is_causal = True
        self.attention_dropout = config.attention_dropout

        if (self.head_dim * self.num_heads) != self.hidden_size:
            raise ValueError(
                f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
                f" and `num_heads`: {self.num_heads})."
            )
        self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=True)
        self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=True)
        self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=True)
        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)

        self.rotary_emb = Qwen2RotaryEmbedding(
            self.head_dim,
            max_position_embeddings=self.max_position_embeddings,
            base=self.rope_theta,
        )
   
    # Adapted from Qwen2Attention.forward
    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_value: Optional[Cache] = None,
        output_attentions: bool = False,
        use_cache: bool = False,
        cache_position: Optional[torch.LongTensor] = None,
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
        bsz, q_len, _ = hidden_states.size()

        query_states = self.q_proj(hidden_states)
        key_states = self.k_proj(hidden_states)
        value_states = self.v_proj(hidden_states)

        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)

        # NOTE: RoPE return all embedding (to satisfy torch compile)
        cos, sin = self.rotary_emb(value_states, seq_len=past_key_value.get_max_length())
        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)

        if past_key_value is not None:
            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}  # Specific to RoPE models
            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)

        key_states = repeat_kv(key_states, self.num_key_value_groups)
        value_states = repeat_kv(value_states, self.num_key_value_groups)

        causal_mask = attention_mask
        if attention_mask is not None:  # no matter the length, we just slice it
            causal_mask = attention_mask[:, :, :, : past_key_value.get_max_length()]

        query_states = query_states.contiguous()
        key_states = key_states.contiguous()
        value_states = value_states.contiguous()

        # We dispatch to SDPA's Flash Attention or Efficient kernels via this `is_causal` if statement instead of an inline conditional assignment
        # in SDPA to support both torch.compile's dynamic shapes and full graph options. An inline conditional prevents dynamic shapes from compiling.
        # The q_len > 1 is necessary to match with AttentionMaskConverter.to_causal_4d that does not create a causal mask in case q_len == 1.
        is_causal = True if causal_mask is None and q_len > 1 else False

        attn_output = torch.nn.functional.scaled_dot_product_attention(
            query_states,
            key_states,
            value_states,
            attn_mask=causal_mask,
            dropout_p=0.0,
            is_causal=is_causal,
        )

        attn_output = attn_output.transpose(1, 2).contiguous()
        attn_output = attn_output.view(bsz, q_len, self.hidden_size)

        attn_output = self.o_proj(attn_output)

        return attn_output, None, past_key_value


# ===================================================================
# =============================Layer=================================
# ===================================================================
class GPUQwen2DecoderLayer(nn.Module):
    def __init__(self, config: Qwen2Config, layer_idx: int):
        super().__init__()
        self.hidden_size = config.hidden_size

        if config.sliding_window and config._attn_implementation != "flash_attention_2":
            logger.warning_once(
                f"Sliding Window Attention is enabled but not implemented for `{config._attn_implementation}`; "
                "unexpected results may be encountered."
            )
        self.self_attn = GPUQwen2Attention(config, layer_idx)

        self.mlp = Qwen2MLP(config)
        self.input_layernorm = Qwen2RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.post_attention_layernorm = Qwen2RMSNorm(config.hidden_size, eps=config.rms_norm_eps)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_value: Optional[Tuple[torch.Tensor]] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = False,
        cache_position: Optional[torch.LongTensor] = None,
        **kwargs,
    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
        """
        Args:
            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
            attention_mask (`torch.FloatTensor`, *optional*): attention mask of size
                `(batch, sequence_length)` where padding elements are indicated by 0.
            output_attentions (`bool`, *optional*):
                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
                returned tensors for more detail.
            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`).
            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
            cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
                Indices depicting the position of the input sequence tokens in the sequence.
            kwargs (`dict`, *optional*):
                Arbitrary kwargs to be ignored, used for FSDP and other methods that injects code
                into the model
        """

        residual = hidden_states

        hidden_states = self.input_layernorm(hidden_states)

        # Self Attention
        hidden_states, self_attn_weights, present_key_value = self.self_attn(
            hidden_states=hidden_states,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_value=past_key_value,
            output_attentions=output_attentions,
            use_cache=use_cache,
            cache_position=cache_position,
        )
        hidden_states = residual + hidden_states

        # Fully Connected
        residual = hidden_states
        hidden_states = self.post_attention_layernorm(hidden_states)
        hidden_states = self.mlp(hidden_states)
        hidden_states = residual + hidden_states

        outputs = (hidden_states,)

        if output_attentions:
            outputs += (self_attn_weights,)

        if use_cache:
            outputs += (present_key_value,)

        return outputs
    
# ===================================================================
# ========================Qwen2ForCausalLM===========================
# ===================================================================
class InferQwen2ForCausalLM(Qwen2ForCausalLM):
    def __init__(self, config):
        super().__init__(config)
        self.compile_forward = torch.compile(self.simplify_forward, dynamic=False, fullgraph=True) \
            if _GPU_QWEN_TORCH_COMPILE else self.simplify_forward
        self.text_phase = True
    '''
    NOTE: 重写原Qwen2ForCausalLM forward函数，torchair直接编译原函数在返回CausalLMOutputWithPast时会出现编译错误
    '''
    def simplify_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,
            labels: 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,
            cache_position: Optional[torch.LongTensor] = None,
            ):
        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
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
        outputs = self.model(
            input_ids=input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
            cache_position=cache_position,
        )

        return outputs
    
    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,
            labels: 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,
            cache_position: Optional[torch.LongTensor] = None,
            do_compile = True
        ) -> Union[Tuple, CausalLMOutputWithPast]:
        if past_key_values is not None:
            past_key_values.training = False
        # print(self.text_phase)
        if input_ids is not None:
            if self.text_phase:
                inputs_embeds = self.model.embed_tokens(input_ids)
            else:
                inputs_embeds = self.speech_token_emded(input_ids)
            if torch.isin(input_ids, 151645).any():
                self.text_phase = False
            input_ids = None

        if (inputs_embeds is not None and cache_position[0] == 0) or do_compile==False :
            # prefill branch
            outputs = self.simplify_forward(input_ids,
                            attention_mask,
                            position_ids,
                            past_key_values,
                            inputs_embeds,
                            labels,
                            use_cache,
                            output_attentions,
                            output_hidden_states,
                            return_dict,
                            cache_position)
        else:
            # decoding
            outputs = self.compile_forward(input_ids,
                            attention_mask,
                            position_ids,
                            past_key_values,
                            inputs_embeds,
                            labels,
                            use_cache,
                            output_attentions,
                            output_hidden_states,
                            return_dict,
                            cache_position)
        
        last_hidden_states = outputs.last_hidden_state
        
        if self.text_phase:
            logits = self.lm_head(last_hidden_states)
        else:
            logits = self.speech_head(last_hidden_states)
        
        logits = logits.float()

        return CausalLMOutputWithPast(
            loss=None,
            logits=logits,
            past_key_values=outputs.past_key_values,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )


    def prepare_inputs_for_generation(
        self,
        input_ids,
        past_key_values=None,
        attention_mask=None,
        inputs_embeds=None,
        cache_position=None,
        position_ids=None,
        use_cache=True,
        **kwargs,
    ):
        """
        Mainly add static cache support
        """
        # If we have cache: let's slice `input_ids` through `cache_position`, to keep only the unprocessed tokens
        # Exception 1: when passing input_embeds, input_ids may be missing entries
        # Exception 2: some generation methods do special slicing of input_ids, so we don't need to do it here
        if past_key_values is not None:
            if inputs_embeds is not None:  # Exception 1
                input_ids = input_ids[:, -cache_position.shape[0] :]
            elif input_ids.shape[1] != cache_position.shape[0]:  # Default case (the "else", a no op, is Exception 2)
                input_ids = input_ids[:, cache_position]

        if attention_mask is not None and position_ids is None:
            # create position_ids on the fly for batch generation
            position_ids = attention_mask.long().cumsum(-1) - 1
            position_ids.masked_fill_(attention_mask == 0, 1)
            if past_key_values:
                position_ids = position_ids[:, -input_ids.shape[1] :]
                # This `clone` call is needed to avoid recapturing cuda graphs with `torch.compile`'s  `mode="reduce-overhead`,
                # as otherwise the input `position_ids` would have various stride during the decoding.
                # Here, simply using `.contiguous()` is not sufficient as in the batch size = 1 case,
                # `position_ids` is already contiguous but with varying stride which retriggers a capture.
                position_ids = position_ids.clone(memory_format=torch.contiguous_format)

        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
        if inputs_embeds is not None and cache_position[0] == 0:
            model_inputs = {"inputs_embeds": inputs_embeds}
        else:
            # NOTE: 与上述的position_ids相同，same as position_ids, for torch.compile and cuda graph
            input_ids = input_ids.clone(memory_format=torch.contiguous_format)
            model_inputs = {"input_ids": input_ids}

        if isinstance(past_key_values, StaticCache) and attention_mask.ndim == 2:
            if inputs_embeds is not None and cache_position[0] == 0:
                # prefill phase, inputs_embeds has shape (B,S,H)
                batch_size, sequence_length = inputs_embeds.shape[0], inputs_embeds.shape[1]
                device = inputs_embeds.device
            else:
                # decdoing phase, input_ids has shape (B,S)
                batch_size, sequence_length = input_ids.shape
                device = input_ids.device

            dtype = self.lm_head.weight.dtype
            min_dtype = torch.finfo(dtype).min

            if inputs_embeds is not None and inputs_embeds.ndim == 2 or input_ids is not None and input_ids.size(-1) == 1:
                # we only expand attention mask in docoding mode
                attention_mask = _prepare_4d_causal_attention_mask_with_cache_position(
                    attention_mask,
                    sequence_length=sequence_length,
                    target_length=past_key_values.get_max_length(),
                    dtype=dtype,
                    device=device,
                    min_dtype=min_dtype,
                    cache_position=cache_position,
                    batch_size=batch_size,
                )

        model_inputs.update(
            {
                "position_ids": position_ids,
                "cache_position": cache_position,
                "past_key_values": past_key_values,
                "use_cache": use_cache,
                "attention_mask": attention_mask,
                "do_compile": kwargs['do_compile'],
            }
        )
        return model_inputs

# ===================================================================
print("========================= DO Qwen2 PATCH ===========================")
# ===================================================================
transformers.models.qwen2.modeling_qwen2.Qwen2PreTrainedModel._supports_static_cache = True # enable static cache
transformers.models.qwen2.modeling_qwen2.Qwen2DecoderLayer = GPUQwen2DecoderLayer
transformers.models.qwen2.modeling_qwen2.Qwen2ForCausalLM = InferQwen2ForCausalLM