support cos_sin_cache prefetch for qwen2

vllm_ascend/models/qwen2.py

@@ -1,10 +1,30 @@
+#
+# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
+# Copyright 2023 The vLLM team.
+#
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# Adapted from vllm/model_executor/models/qwen2.py
+# This file is a part of the vllm-ascend project.
+
 from collections.abc import Iterable
-from typing import Optional, Union
+from typing import Any, Optional, Union
 
 import torch
 import torch.nn.functional as F
 from torch import nn
 from transformers import Qwen2Config
+from vllm.attention import AttentionType
 from vllm.compilation.decorators import support_torch_compile
 from vllm.config import CacheConfig, VllmConfig
 from vllm.distributed import (get_pp_group, get_tensor_model_parallel_rank,
@@ -13,11 +33,13 @@
                               tensor_model_parallel_all_reduce,
                               tensor_model_parallel_reduce_scatter)
 from vllm.forward_context import get_forward_context
+from vllm.model_executor.layers.layernorm import RMSNorm
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
 from vllm.model_executor.models.interfaces import SupportsLoRA, SupportsPP
-from vllm.model_executor.models.qwen2 import Qwen2DecoderLayer, Qwen2Model
+from vllm.model_executor.models.qwen2 import (Qwen2Attention, Qwen2MLP,
+                                              Qwen2Model)
 from vllm.model_executor.models.utils import (AutoWeightsLoader,
                                               PPMissingLayer, maybe_prefix)
 from vllm.model_executor.sampling_metadata import SamplingMetadata
@@ -47,19 +69,102 @@ def maybe_pad_and_reduce_scatter(
     return hidden_states
 
 
-class CustomQwen2DecoderLayer(Qwen2DecoderLayer):
+class CustomQwen2Attention(Qwen2Attention):
 
     def __init__(
         self,
-        config: Qwen2Config,
+        hidden_size: int,
+        num_heads: int,
+        num_kv_heads: int,
+        max_position: int = 4096 * 32,
+        rope_theta: float = 10000,
         cache_config: Optional[CacheConfig] = None,
         quant_config: Optional[QuantizationConfig] = None,
+        rope_scaling: Optional[tuple] = None,
         prefix: str = "",
+        attn_type: str = AttentionType.DECODER,
+        dual_chunk_attention_config: Optional[dict[str, Any]] = None,
     ) -> None:
-        super().__init__(config=config,
+        super().__init__(hidden_size=hidden_size,
+                         num_heads=num_heads,
+                         num_kv_heads=num_kv_heads,
+                         max_position=max_position,
+                         rope_theta=rope_theta,
                          cache_config=cache_config,
                          quant_config=quant_config,
-                         prefix=prefix)
+                         rope_scaling=rope_scaling,
+                         prefix=prefix,
+                         attn_type=attn_type,
+                         dual_chunk_attention_config=dual_chunk_attention_config)
+
+    def forward(
+        self,
+        positions: torch.Tensor,
+        hidden_states: torch.Tensor,
+        cos: torch.Tensor,
+        sin: torch.Tensor
+    ) -> torch.Tensor:
+        qkv, _ = self.qkv_proj(hidden_states)
+        q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
+        q, k = self.rotary_emb(positions, q, k, cos=cos, sin=sin, skip_index_select=True)
+        attn_output = self.attn(q, k, v)
+        output, _ = self.o_proj(attn_output)
+        return output
+
+
+class CustomQwen2DecoderLayer(nn.Module):
+
+    def __init__(
+        self,
+        config: Qwen2Config,
+        cache_config: Optional[CacheConfig] = None,
+        quant_config: Optional[QuantizationConfig] = None,
+        prefix: str = "",
+    ) -> None:
+        super().__init__()
+
+        self.hidden_size = config.hidden_size
+        # Requires transformers > 4.32.0
+        rope_theta = getattr(config, "rope_theta", 1000000)
+        rope_scaling = getattr(config, "rope_scaling", None)
+        dual_chunk_attention_config = getattr(config,
+                                              "dual_chunk_attention_config",
+                                              None)
+
+        # By default, Qwen2 uses causal attention as it is a decoder-only model.
+        # You can override the HF config with `is_causal=False` to enable
+        # bidirectional attention, which is used in some embedding models
+        # (e.g. Alibaba-NLP/gte-Qwen2-7B-instruct)
+        if getattr(config, "is_causal", True):
+            attn_type = AttentionType.DECODER
+        else:
+            attn_type = AttentionType.ENCODER_ONLY
+
+        self.self_attn = CustomQwen2Attention(
+            hidden_size=self.hidden_size,
+            num_heads=config.num_attention_heads,
+            max_position=config.max_position_embeddings,
+            num_kv_heads=config.num_key_value_heads,
+            rope_theta=rope_theta,
+            cache_config=cache_config,
+            quant_config=quant_config,
+            rope_scaling=rope_scaling,
+            prefix=f"{prefix}.self_attn",
+            attn_type=attn_type,
+            dual_chunk_attention_config=dual_chunk_attention_config,
+        )
+        self.mlp = Qwen2MLP(
+            hidden_size=self.hidden_size,
+            intermediate_size=config.intermediate_size,
+            hidden_act=config.hidden_act,
+            quant_config=quant_config,
+            prefix=f"{prefix}.mlp",
+        )
+        self.input_layernorm = RMSNorm(config.hidden_size,
+                                       eps=config.rms_norm_eps)
+        self.post_attention_layernorm = RMSNorm(config.hidden_size,
+                                                eps=config.rms_norm_eps)
+
         self.tp_rank = get_tensor_model_parallel_rank()
         self.tp_size = get_tensor_model_parallel_world_size()
         self.self_attn.o_proj.reduce_results = False
@@ -68,6 +173,8 @@ def __init__(
     def forward(
         self,
         positions: torch.Tensor,
+        cos: torch.Tensor,
+        sin: torch.Tensor,
         hidden_states: torch.Tensor,
         residual: Optional[torch.Tensor],
         flashcomm_v1_enabled: bool,
@@ -91,6 +198,8 @@ def forward(
         hidden_states = self.self_attn(
             positions=positions,
             hidden_states=hidden_states,
+            cos=cos,
+            sin=sin
         )
         if flashcomm_v1_enabled:
             hidden_states = maybe_pad_and_reduce_scatter(
@@ -133,6 +242,9 @@ def __init__(
                          decoder_layer_type=decoder_layer_type)
         self.tp_size = get_tensor_model_parallel_world_size()
 
+        self.rotary_emb = self.layers[0].self_attn.rotary_emb
+        self.cos_sin_cache = self.rotary_emb.cos_sin_cache
+
     def forward(
         self,
         input_ids: torch.Tensor,
@@ -161,9 +273,19 @@ def forward(
             num_tokens = hidden_states.size(0)
             pad_size = (self.tp_size -
                         (num_tokens % self.tp_size)) % self.tp_size
+
+        cos_sin = self.cos_sin_cache.index_select(0, positions)
+        head_dim = cos_sin.size()[-1]
+        cos, sin = cos_sin.reshape(-1, 2,
+                                   head_dim // 2).repeat(1, 1, 2).chunk(2, dim=-2)
+        cos = cos.view(1, -1, 1, head_dim).contiguous()
+        sin = sin.view(1, -1, 1, head_dim).contiguous()
+
         for layer in self.layers[self.start_layer:self.end_layer]:
             hidden_states, residual = layer(
                 positions,
+                cos,
+                sin,
                 hidden_states,
                 residual,
                 flashcomm_v1_enabled,