ayo/ipex-llm
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 1

Update vllm patch for fix telechat2 and baichuan2 error(#13150)

Browse source
This commit is contained in:
Wang, Jian4 2025-05-12 10:54:22 +08:00 committed by GitHub
parent 9da1c56fa8
commit 5df03ced2c
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 409 additions and 141 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

550
docker/llm/serving/xpu/docker/vllm_for_multi_arc.patch
View file

@ -7078,61 +7078,61 @@ index 000000000..93c64d759
--- /dev/null --- /dev/null
+++ b/csrc/xpu/reduction_utils.h +++ b/csrc/xpu/reduction_utils.h
@@ -0,0 +1,56 @@ @@ -0,0 +1,56 @@
+/* +/*
+ * Copyright (c) 2023, The vLLM team. + * Copyright (c) 2023, The vLLM team.
+ * + *
+ * Licensed under the Apache License, Version 2.0 (the "License"); + * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License. + * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at + * You may obtain a copy of the License at
+ * + *
+ * http://www.apache.org/licenses/LICENSE-2.0 + * http://www.apache.org/licenses/LICENSE-2.0
+ * + *
+ * Unless required by applicable law or agreed to in writing, software + * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS, + * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and + * See the License for the specific language governing permissions and
+ * limitations under the License. + * limitations under the License.
+ */ + */
+#pragma once +#pragma once
+ +
+#include <dpct/dpct.hpp> +#include <dpct/dpct.hpp>
+#include <stdint.h> +#include <stdint.h>
+#include <sycl/sycl.hpp> +#include <sycl/sycl.hpp>
+ +
+namespace vllm { +namespace vllm {
+ +
+template <typename T> +template <typename T>
+__inline__ T warpReduceSum(T val, const sycl::nd_item<3>& item_ct1) { +__inline__ T warpReduceSum(T val, const sycl::nd_item<3>& item_ct1) {
+#pragma unroll +#pragma unroll
+ for (int mask = 16; mask > 0; mask >>= 1) + for (int mask = 16; mask > 0; mask >>= 1)
+ val += dpct::permute_sub_group_by_xor( + val += dpct::permute_sub_group_by_xor(
+ item_ct1.get_sub_group(), val, mask, 32); + item_ct1.get_sub_group(), val, mask, 32);
+ return val; + return val;
+} +}
+ +
+/* Calculate the sum of all elements in a block */ +/* Calculate the sum of all elements in a block */
+template<typename T> +template<typename T>
+__inline__ T blockReduceSum(T val, const sycl::nd_item<3> &item_ct1, T *shared) { +__inline__ T blockReduceSum(T val, const sycl::nd_item<3> &item_ct1, T *shared) {
+ +
+ int lane = item_ct1.get_local_id(2) & 0x1f; + int lane = item_ct1.get_local_id(2) & 0x1f;
+ int wid = item_ct1.get_local_id(2) >> 5; + int wid = item_ct1.get_local_id(2) >> 5;
+ +
+ val = warpReduceSum<T>(val, item_ct1); + val = warpReduceSum<T>(val, item_ct1);
+ +
+ if (lane == 0) { + if (lane == 0) {
+ shared[wid] = val; + shared[wid] = val;
+ } + }
+ item_ct1.barrier(); + item_ct1.barrier();
+ +
+ // Modify from blockDim.x << 5 to blockDim.x / 32. to prevent + // Modify from blockDim.x << 5 to blockDim.x / 32. to prevent
+ // blockDim.x is not divided by 32 + // blockDim.x is not divided by 32
+ val = (item_ct1.get_local_id(2) < (item_ct1.get_local_range(2) / 32.f)) + val = (item_ct1.get_local_id(2) < (item_ct1.get_local_range(2) / 32.f))
+ ? shared[lane] + ? shared[lane]
+ : (T)(0.0f); + : (T)(0.0f);
+ val = warpReduceSum<T>(val, item_ct1); + val = warpReduceSum<T>(val, item_ct1);
+ return val; + return val;
+} +}
+ +
+} // namespace vllm +} // namespace vllm
\ No newline at end of file \ No newline at end of file
diff --git a/csrc/xpu/utils.cpp b/csrc/xpu/utils.cpp diff --git a/csrc/xpu/utils.cpp b/csrc/xpu/utils.cpp
@ -8692,7 +8692,7 @@ index 000000000..e98db9b65
+ tensor_parallel_size=1, + tensor_parallel_size=1,
+ ) + )
diff --git a/vllm/_ipex_ops.py b/vllm/_ipex_ops.py diff --git a/vllm/_ipex_ops.py b/vllm/_ipex_ops.py
index c3d210c27..c3b6ca7eb 100644 index c3d210c27..8dd101608 100644
--- a/vllm/_ipex_ops.py --- a/vllm/_ipex_ops.py
+++ b/vllm/_ipex_ops.py +++ b/vllm/_ipex_ops.py
@@ -1,6 +1,4 @@ @@ -1,6 +1,4 @@
@ -8780,10 +8780,10 @@ index c3d210c27..c3b6ca7eb 100644
+ # todo: ipex will refactor namespace + # todo: ipex will refactor namespace
+ import vllm._C.ops + import vllm._C.ops
+ vllm._C.ops.paged_attention_v1(out, query, + vllm._C.ops.paged_attention_v1(out, query,
+ key_cache.view_as(value_cache), + key_cache.view_as(value_cache),
+ value_cache, num_kv_heads, scale, + value_cache, num_kv_heads, scale,
+ block_tables, context_lens, block_size, + block_tables, context_lens, block_size,
+ max_context_len, alibi_slopes, kv_cache_dtype, k_scale, logits_soft_cap) + max_context_len, alibi_slopes, kv_cache_dtype, k_scale, logits_soft_cap)
@staticmethod @staticmethod
def paged_attention_v2( def paged_attention_v2(
@ -8929,7 +8929,7 @@ index c3d210c27..c3b6ca7eb 100644
@staticmethod @staticmethod
def varlen_attention( def varlen_attention(
@@ -220,22 +262,233 @@ class ipex_ops: @@ -220,22 +262,250 @@ class ipex_ops:
kv_cache_dtype: str, kv_cache_dtype: str,
k_scale: float, k_scale: float,
v_scale: float, v_scale: float,
@ -9044,30 +9044,47 @@ index c3d210c27..c3b6ca7eb 100644
+ p_dropout: float, + p_dropout: float,
+ softmax_scale: float, + softmax_scale: float,
+ zero_tensors: bool, + zero_tensors: bool,
+ is_caual: bool, + is_casual: bool,
+ return_softmax: bool, + return_softmax: bool,
+ gen_: Optional[torch.Generator], + gen_: Optional[torch.Generator],
+ ): + ):
+ return torch.ops.torch_ipex.chunked_prefill( + return ipex.llm.modules.PagedAttention.flash_attn_varlen_func(
+ output,
+ query.contiguous(), + query.contiguous(),
+ key_cache, + key_cache,
+ value_cache, + value_cache,
+ output,
+ cu_seqlens_q, + cu_seqlens_q,
+ cu_seqlens_k, + cu_seqlens_k,
+ seq_used_k,
+ block_table,
+ alibi_slopes,
+ max_seqlen_q, + max_seqlen_q,
+ max_seqlen_k, + max_seqlen_k,
+ p_dropout,
+ softmax_scale, + softmax_scale,
+ zero_tensors, + is_casual,
+ is_caual, + block_table,
+ return_softmax, + alibi_slopes,
+ gen_, + k_scale=1.0,
+ v_scale=1.0,
) )
+ # return torch.ops.torch_ipex.chunked_prefill(
+ # query.contiguous(),
+ # key_cache,
+ # value_cache,
+ # output,
+ # cu_seqlens_q,
+ # cu_seqlens_k,
+ # seq_used_k,
+ # block_table,
+ # alibi_slopes,
+ # max_seqlen_q,
+ # max_seqlen_k,
+ # p_dropout,
+ # softmax_scale,
+ # zero_tensors,
+ # is_caual,
+ # return_softmax,
+ # gen_,
+ # )
+
+
+ @staticmethod + @staticmethod
+ def copy_blocks(key_caches: List[torch.Tensor], + def copy_blocks(key_caches: List[torch.Tensor],
+ value_caches: List[torch.Tensor], + value_caches: List[torch.Tensor],
@ -9078,7 +9095,7 @@ index c3d210c27..c3b6ca7eb 100644
+ # block_mapping, + # block_mapping,
+ # ) + # )
+ vllm._C.cache_ops.copy_blocks(key_caches, value_caches, block_mapping) + vllm._C.cache_ops.copy_blocks(key_caches, value_caches, block_mapping)
+
@staticmethod @staticmethod
def swap_blocks(src: torch.Tensor, dst: torch.Tensor, def swap_blocks(src: torch.Tensor, dst: torch.Tensor,
block_mapping: torch.Tensor) -> None: block_mapping: torch.Tensor) -> None:
@ -11666,6 +11683,143 @@ index 5649cf2dd..66e30984e 100644
if isinstance(load_config.load_format, type): if isinstance(load_config.load_format, type):
return load_config.load_format(load_config) return load_config.load_format(load_config)
diff --git a/vllm/model_executor/models/baichuan.py b/vllm/model_executor/models/baichuan.py
index 6a3112b5f..7e2b7c862 100644
--- a/vllm/model_executor/models/baichuan.py
+++ b/vllm/model_executor/models/baichuan.py
@@ -47,7 +47,7 @@ from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.sequence import IntermediateTensors
from .interfaces import SupportsLoRA, SupportsPP, SupportsQuant
-from .utils import (AutoWeightsLoader, is_pp_missing_parameter,
+from .utils import (is_pp_missing_parameter,
make_empty_intermediate_tensors_factory, make_layers)
@@ -321,45 +321,6 @@ class BaiChuanModel(nn.Module):
hidden_states, _ = self.norm(hidden_states, residual)
return hidden_states
- def load_weights(self, weights: Iterable[Tuple[str,
- torch.Tensor]]) -> Set[str]:
- stacked_params_mapping = [
- # (param_name, shard_name, shard_id)
- ("gate_up_proj", "gate_proj", 0),
- ("gate_up_proj", "up_proj", 1),
- ]
- params_dict = dict(self.named_parameters())
- loaded_params: Set[str] = set()
- for name, loaded_weight in weights:
- if "rotary_emb.inv_freq" in name:
- continue
-
- for (param_name, weight_name, shard_id) in stacked_params_mapping:
- if weight_name not in name:
- continue
- name = name.replace(weight_name, param_name)
- # Skip loading extra bias for GPTQ models.
- if name.endswith(".bias") and name not in params_dict:
- continue
- if is_pp_missing_parameter(name, self):
- continue
- param = params_dict[name]
- weight_loader = param.weight_loader
- weight_loader(param, loaded_weight, shard_id)
- break
- else:
- # Skip loading extra bias for GPTQ models.
- if name.endswith(".bias") and name not in params_dict:
- continue
- if is_pp_missing_parameter(name, self):
- continue
- param = params_dict[name]
- weight_loader = getattr(param, "weight_loader",
- default_weight_loader)
- weight_loader(param, loaded_weight)
- loaded_params.add(name)
- return loaded_params
-
class BaiChuanBaseForCausalLM(nn.Module, SupportsLoRA, SupportsPP,
SupportsQuant):
@@ -392,7 +353,6 @@ class BaiChuanBaseForCausalLM(nn.Module, SupportsLoRA, SupportsPP,
self.lm_head = ParallelLMHead(config.vocab_size,
config.hidden_size,
quant_config=quant_config)
- self.lm_head.weight.weight_loader = self.lm_head_weight_loader
if self.config.tie_word_embeddings:
self.lm_head.weight = self.model.embed_tokens.weight
self.logits_processor = LogitsProcessor(config.vocab_size)
@@ -433,22 +393,53 @@ class BaiChuanBaseForCausalLM(nn.Module, SupportsLoRA, SupportsPP,
def load_weights(self, weights: Iterable[Tuple[str,
torch.Tensor]]) -> Set[str]:
- loader = AutoWeightsLoader(self)
- return loader.load_weights(weights)
-
- def lm_head_weight_loader(self, param: nn.Parameter,
- loaded_weight: torch.Tensor):
- # Unlike Baichuan, Baichuan2 normalizes the head weights.
- # Refer to:
- # https://huggingface.co/baichuan-inc/Baichuan2-7B-Chat/blob/84603cde5ebffb6084e476cfaeceaf0b8b91fe54/modeling_baichuan.py#L508
- # Distinguish between Baichuan and Baichuan2 by checking the
- # vocab size. This is suggested by
- # https://github.com/vllm-project/vllm/pull/1022#discussion_r1325652704
- is_baichuan2 = self.config.vocab_size == 125696
- if is_baichuan2:
- loaded_weight = torch.nn.functional.normalize(loaded_weight)
-
- default_weight_loader(param, loaded_weight)
+ stacked_params_mapping = [
+ # (param_name, shard_name, shard_id)
+ ("gate_up_proj", "gate_proj", 0),
+ ("gate_up_proj", "up_proj", 1),
+ ]
+ params_dict = dict(self.named_parameters())
+ loaded_params: Set[str] = set()
+ for name, loaded_weight in weights:
+ if "rotary_emb.inv_freq" in name:
+ continue
+ if name == "lm_head.weight":
+ # Unlike Baichuan, Baichuan2 normalizes the head weights.
+ # Refer to:
+ # https://huggingface.co/baichuan-inc/Baichuan2-7B-Chat/blob/84603cde5ebffb6084e476cfaeceaf0b8b91fe54/modeling_baichuan.py#L508
+ # Distinguish between Baichuan and Baichuan2 by checking the
+ # vocab size. This is suggested by
+ # https://github.com/vllm-project/vllm/pull/1022#discussion_r1325652704
+ is_baichuan2 = self.config.vocab_size == 125696
+ if is_baichuan2:
+ loaded_weight = torch.nn.functional.normalize(
+ loaded_weight)
+
+ for (param_name, weight_name, shard_id) in stacked_params_mapping:
+ if weight_name not in name:
+ continue
+ name = name.replace(weight_name, param_name)
+ # Skip loading extra bias for GPTQ models.
+ if name.endswith(".bias") and name not in params_dict:
+ continue
+ if is_pp_missing_parameter(name, self):
+ continue
+ param = params_dict[name]
+ weight_loader = param.weight_loader
+ weight_loader(param, loaded_weight, shard_id)
+ break
+ else:
+ # Skip loading extra bias for GPTQ models.
+ if name.endswith(".bias") and name not in params_dict:
+ continue
+ if is_pp_missing_parameter(name, self):
+ continue
+ param = params_dict[name]
+ weight_loader = getattr(param, "weight_loader",
+ default_weight_loader)
+ weight_loader(param, loaded_weight)
+ loaded_params.add(name)
+ return loaded_params
class BaichuanForCausalLM(BaiChuanBaseForCausalLM):
diff --git a/vllm/model_executor/models/chatglm.py b/vllm/model_executor/models/chatglm.py diff --git a/vllm/model_executor/models/chatglm.py b/vllm/model_executor/models/chatglm.py
index 1b1738f88..2c2ed67b9 100644 index 1b1738f88..2c2ed67b9 100644
--- a/vllm/model_executor/models/chatglm.py --- a/vllm/model_executor/models/chatglm.py
@ -14147,7 +14301,7 @@ index c0a3c59ba..8614c2273 100644
"Zamba2ForCausalLM": ("zamba2", "Zamba2ForCausalLM"), "Zamba2ForCausalLM": ("zamba2", "Zamba2ForCausalLM"),
# [Encoder-decoder] # [Encoder-decoder]
diff --git a/vllm/model_executor/models/siglip.py b/vllm/model_executor/models/siglip.py diff --git a/vllm/model_executor/models/siglip.py b/vllm/model_executor/models/siglip.py
index cecad9e89..df4cf4776 100644 index cecad9e89..7eaabd1db 100644
--- a/vllm/model_executor/models/siglip.py --- a/vllm/model_executor/models/siglip.py
+++ b/vllm/model_executor/models/siglip.py +++ b/vllm/model_executor/models/siglip.py
@@ -140,6 +140,74 @@ class SiglipVisionEmbeddings(nn.Module): @@ -140,6 +140,74 @@ class SiglipVisionEmbeddings(nn.Module):
@ -14195,9 +14349,9 @@ index cecad9e89..df4cf4776 100644
+ +
+ query, key, value = (x.transpose(1, 2) + query, key, value = (x.transpose(1, 2)
+ for x in (query, key, value)) + for x in (query, key, value))
+ from ipex_llm.transformers.models.utils import use_sdp_causal
+ from vllm.attention.backends.ipex_attn import use_sdp_causal + from vllm.attention.backends.ipex_attn import use_sdp_causal
+ import xe_addons, math + import xe_addons, math
+ from vllm.attention.backends.abstract import AttentionType
+ mask = None + mask = None
+ scale = 1 / math.sqrt(self.head_size) if self.scale is None else self.scale + scale = 1 / math.sqrt(self.head_size) if self.scale is None else self.scale
+ from ipex_llm.transformers.models.common import padding_qkv_hd + from ipex_llm.transformers.models.common import padding_qkv_hd
@ -14209,7 +14363,7 @@ index cecad9e89..df4cf4776 100644
+ query, key, value, + query, key, value,
+ self.head_size, num + self.head_size, num
+ ) + )
+ if use_sdp_causal(query.shape[-1], query, 0): + if use_sdp_causal(query.shape[-1], query, 0, AttentionType.DECODER):
+ out = xe_addons.sdp_non_causal(query.contiguous(), key.contiguous(), value.contiguous(), mask, scale)[:, :, :, :self.head_size].transpose(1, 2) + out = xe_addons.sdp_non_causal(query.contiguous(), key.contiguous(), value.contiguous(), mask, scale)[:, :, :, :self.head_size].transpose(1, 2)
+ # import torch.nn.functional as F + # import torch.nn.functional as F
+ # out = F.scaled_dot_product_attention(query, + # out = F.scaled_dot_product_attention(query,
@ -14239,10 +14393,23 @@ index cecad9e89..df4cf4776 100644
def forward( def forward(
self, self,
diff --git a/vllm/model_executor/models/telechat2.py b/vllm/model_executor/models/telechat2.py diff --git a/vllm/model_executor/models/telechat2.py b/vllm/model_executor/models/telechat2.py
index a38035e37..9631fbd83 100644 index a38035e37..570f2bcdd 100644
--- a/vllm/model_executor/models/telechat2.py --- a/vllm/model_executor/models/telechat2.py
+++ b/vllm/model_executor/models/telechat2.py +++ b/vllm/model_executor/models/telechat2.py
@@ -44,9 +44,9 @@ class TeleChat2Model(LlamaModel): @@ -22,10 +22,12 @@
from typing import Iterable, Set, Tuple
import torch
+import torch.nn as nn
from vllm.config import VllmConfig
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.model_executor.models.llama import LlamaForCausalLM, LlamaModel
+from .llama import LlamaDecoderLayer
from .utils import (AutoWeightsLoader, PPMissingLayer, WeightsMapper,
is_pp_missing_parameter)
@@ -44,9 +46,9 @@ class TeleChat2Model(LlamaModel):
for layer in self.layers: for layer in self.layers:
if not isinstance(layer, PPMissingLayer): if not isinstance(layer, PPMissingLayer):
layer.self_attn.qkv_proj.bias = None layer.self_attn.qkv_proj.bias = None
@ -14254,6 +14421,18 @@ index a38035e37..9631fbd83 100644
def load_weights(self, weights: Iterable[Tuple[str, def load_weights(self, weights: Iterable[Tuple[str,
torch.Tensor]]) -> Set[str]: torch.Tensor]]) -> Set[str]:
@@ -120,7 +122,10 @@ class TeleChat2ForCausalLM(LlamaForCausalLM):
},
)
- def _init_model(self, vllm_config: VllmConfig, prefix: str = ""):
+ def _init_model(self,
+ vllm_config: VllmConfig,
+ prefix: str = "",
+ layer_type: type[nn.Module] = LlamaDecoderLayer):
return TeleChat2Model(vllm_config=vllm_config, prefix=prefix)
def load_weights(self, weights: Iterable[Tuple[str,
diff --git a/vllm/multimodal/utils.py b/vllm/multimodal/utils.py diff --git a/vllm/multimodal/utils.py b/vllm/multimodal/utils.py
index fc0fb8929..6454e7006 100644 index fc0fb8929..6454e7006 100644
--- a/vllm/multimodal/utils.py --- a/vllm/multimodal/utils.py
@ -14319,7 +14498,7 @@ index b6f6029de..b90fea9fd 100644
def is_neuron(self) -> bool: def is_neuron(self) -> bool:
return self._enum == PlatformEnum.NEURON return self._enum == PlatformEnum.NEURON
diff --git a/vllm/platforms/xpu.py b/vllm/platforms/xpu.py diff --git a/vllm/platforms/xpu.py b/vllm/platforms/xpu.py
index 225e756cd..4fd7fe220 100644 index 225e756cd..25b83549a 100644
--- a/vllm/platforms/xpu.py --- a/vllm/platforms/xpu.py
+++ b/vllm/platforms/xpu.py +++ b/vllm/platforms/xpu.py
@@ -4,6 +4,7 @@ from typing import TYPE_CHECKING, Optional @@ -4,6 +4,7 @@ from typing import TYPE_CHECKING, Optional
@ -14330,7 +14509,17 @@ index 225e756cd..4fd7fe220 100644
from vllm.logger import init_logger from vllm.logger import init_logger
from .interface import DeviceCapability, Platform, PlatformEnum, _Backend from .interface import DeviceCapability, Platform, PlatformEnum, _Backend
@@ -33,8 +34,13 @@ class XPUPlatform(Platform): @@ -25,6 +26,9 @@ class XPUPlatform(Platform):
# see https://github.com/ray-project/ray/blob/6a5eb5865eeb9ccf058a79b44f107e327e360673/python/ray/_private/accelerators/intel_gpu.py#L20 # noqa: E501
ray_device_key: str = "GPU"
device_control_env_var: str = "ONEAPI_DEVICE_SELECTOR"
+ additional_env_vars: list[str] = [
+ "IPEX_LLM_LOWBIT",
+ ]
@classmethod
def get_attn_backend_cls(cls, selected_backend: _Backend, head_size: int,
@@ -33,8 +37,13 @@ class XPUPlatform(Platform):
use_mla: bool) -> str: use_mla: bool) -> str:
if selected_backend != _Backend.IPEX: if selected_backend != _Backend.IPEX:
logger.info("Cannot use %s backend on XPU.", selected_backend) logger.info("Cannot use %s backend on XPU.", selected_backend)
@ -14346,7 +14535,7 @@ index 225e756cd..4fd7fe220 100644
@staticmethod @staticmethod
def get_device_capability( def get_device_capability(
@@ -63,6 +69,8 @@ class XPUPlatform(Platform): @@ -63,6 +72,8 @@ class XPUPlatform(Platform):
@classmethod @classmethod
def check_and_update_config(cls, vllm_config: VllmConfig) -> None: def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
cache_config = vllm_config.cache_config cache_config = vllm_config.cache_config
@ -14355,7 +14544,7 @@ index 225e756cd..4fd7fe220 100644
if cache_config and cache_config.block_size is None: if cache_config and cache_config.block_size is None:
cache_config.block_size = 16 cache_config.block_size = 16
@@ -87,31 +95,46 @@ class XPUPlatform(Platform): @@ -87,31 +98,46 @@ class XPUPlatform(Platform):
raise NotImplementedError( raise NotImplementedError(
"XPU does not support speculative decoding") "XPU does not support speculative decoding")
@ -14412,6 +14601,15 @@ index 225e756cd..4fd7fe220 100644
@classmethod @classmethod
def is_pin_memory_available(cls): def is_pin_memory_available(cls):
@@ -140,3 +166,7 @@ class XPUPlatform(Platform):
@classmethod
def get_device_communicator_cls(cls) -> str:
return "vllm.distributed.device_communicators.xpu_communicator.XpuCommunicator" # noqa
+
+ @classmethod
+ def use_all_gather(cls) -> bool:
+ return False
\ No newline at end of file
diff --git a/vllm/transformers_utils/configs/__init__.py b/vllm/transformers_utils/configs/__init__.py diff --git a/vllm/transformers_utils/configs/__init__.py b/vllm/transformers_utils/configs/__init__.py
index 53699341b..6bc039068 100644 index 53699341b..6bc039068 100644
--- a/vllm/transformers_utils/configs/__init__.py --- a/vllm/transformers_utils/configs/__init__.py
@ -14432,6 +14630,35 @@ index 53699341b..6bc039068 100644
"ChatGLMConfig", "ChatGLMConfig",
"Cohere2Config", "Cohere2Config",
"DbrxConfig", "DbrxConfig",
diff --git a/vllm/utils.py b/vllm/utils.py
index 5f32f8cb6..2ee0c1906 100644
--- a/vllm/utils.py
+++ b/vllm/utils.py
@@ -128,6 +128,8 @@ STR_NOT_IMPL_ENC_DEC_ERR_STRS = {
"STR_NOT_IMPL_ENC_DEC_PROMPT_ADAPTER": STR_NOT_IMPL_ENC_DEC_PROMPT_ADAPTER,
}
+BMG_TARGET_IDS = ["0xe20b", "0xe210"]
+
# Constants related to forcing the attention backend selection
# String name of register which may be set in order to
@@ -2564,3 +2566,14 @@ def sha256(input) -> int:
input_bytes = pickle.dumps(input, protocol=pickle.HIGHEST_PROTOCOL)
return int.from_bytes(hashlib.sha256(input_bytes).digest(),
byteorder="big")
+
+@cache
+def is_bmg_platform():
+ if not torch.xpu.is_available():
+ raise ValueError("Cannot detect the usage of XPU!")
+ device_index = torch.xpu.current_device()
+ device_name = torch.xpu.get_device_name(device_index)
+ for target_id in BMG_TARGET_IDS:
+ if target_id in device_name:
+ return True
+ return False
\ No newline at end of file
diff --git a/vllm/v1/attention/backends/flash_attn.py b/vllm/v1/attention/backends/flash_attn.py diff --git a/vllm/v1/attention/backends/flash_attn.py b/vllm/v1/attention/backends/flash_attn.py
index c271f438e..cf7180606 100755 index c271f438e..cf7180606 100755
--- a/vllm/v1/attention/backends/flash_attn.py --- a/vllm/v1/attention/backends/flash_attn.py
@ -14457,10 +14684,10 @@ index c271f438e..cf7180606 100755
assert sliding_window == (-1, -1), ( assert sliding_window == (-1, -1), (
diff --git a/vllm/v1/attention/backends/ipex_attn.py b/vllm/v1/attention/backends/ipex_attn.py diff --git a/vllm/v1/attention/backends/ipex_attn.py b/vllm/v1/attention/backends/ipex_attn.py
new file mode 100644 new file mode 100644
index 000000000..29cde02f3 index 000000000..f4a435eaa
--- /dev/null --- /dev/null
+++ b/vllm/v1/attention/backends/ipex_attn.py +++ b/vllm/v1/attention/backends/ipex_attn.py
@@ -0,0 +1,358 @@ @@ -0,0 +1,392 @@
+from dataclasses import dataclass +from dataclasses import dataclass
+from typing import Any, Dict, List, Optional, Tuple, Type +from typing import Any, Dict, List, Optional, Tuple, Type
+ +
@ -14474,6 +14701,7 @@ index 000000000..29cde02f3
+from vllm.attention.ops.paged_attn import (PagedAttention, +from vllm.attention.ops.paged_attn import (PagedAttention,
+ PagedAttentionMetadata) + PagedAttentionMetadata)
+from vllm.attention.backends.ipex_attn import use_gqa_kernel +from vllm.attention.backends.ipex_attn import use_gqa_kernel
+from vllm.utils import is_bmg_platform
+import os +import os
+ +
+@dataclass +@dataclass
@ -14509,9 +14737,9 @@ index 000000000..29cde02f3
+ # if block_size % 16 != 0: + # if block_size % 16 != 0:
+ # raise ValueError("Block size must be a multiple of 16.") + # raise ValueError("Block size must be a multiple of 16.")
+ # This needs to be changed... + # This needs to be changed...
+ # return (2, num_blocks, block_size, num_kv_heads, head_size) + return (2, num_blocks, block_size, num_kv_heads, head_size)
+ return PagedAttention.get_kv_cache_shape(num_blocks, block_size, + # return PagedAttention.get_kv_cache_shape(num_blocks, block_size,
+ num_kv_heads, head_size) + # num_kv_heads, head_size)
+ +
+ +
+ +
@ -14557,6 +14785,8 @@ index 000000000..29cde02f3
+ self.num_queries_per_kv = self.num_heads // self.num_kv_heads + self.num_queries_per_kv = self.num_heads // self.num_kv_heads
+ +
+ support_head_sizes = IPEXAttentionBackend.get_supported_head_sizes() + support_head_sizes = IPEXAttentionBackend.get_supported_head_sizes()
+ self.using_gqa_kernel = use_gqa_kernel(num_heads, num_kv_heads, head_size, logits_soft_cap)
+ self.is_bmg_platform = is_bmg_platform()
+ if head_size not in support_head_sizes: + if head_size not in support_head_sizes:
+ raise ValueError( + raise ValueError(
+ f"Head size {head_size} is not supported by FlashAttention. " + f"Head size {head_size} is not supported by FlashAttention. "
@ -14567,7 +14797,6 @@ index 000000000..29cde02f3
+ "are not implemented for " + "are not implemented for "
+ "IpexAttnBackendImpl") + "IpexAttnBackendImpl")
+ +
+ # TODO(gc): Refine this logic..., because of bad performance...
+ def forward( + def forward(
+ self, + self,
+ layer: AttentionLayer, + layer: AttentionLayer,
@ -14610,6 +14839,8 @@ index 000000000..29cde02f3
+ k_scale, + k_scale,
+ v_scale, + v_scale,
+ self.scale, + self.scale,
+ self.using_gqa_kernel,
+ self.is_bmg_platform,
+ self.sliding_window, + self.sliding_window,
+ self.alibi_slopes, + self.alibi_slopes,
+ self.logits_soft_cap, + self.logits_soft_cap,
@ -14682,6 +14913,8 @@ index 000000000..29cde02f3
+ k_scale: float, + k_scale: float,
+ v_scale: float, + v_scale: float,
+ scale: float, + scale: float,
+ using_gqa_kernel: bool,
+ is_bmg_platform: bool,
+ sliding_window: Optional[List[int]] = None, + sliding_window: Optional[List[int]] = None,
+ alibi_slopes: Optional[torch.Tensor] = None, + alibi_slopes: Optional[torch.Tensor] = None,
+ logits_soft_cap: Optional[float] = None, + logits_soft_cap: Optional[float] = None,
@ -14700,54 +14933,82 @@ index 000000000..29cde02f3
+ key = key.view(-1, num_kv_heads, head_size) + key = key.view(-1, num_kv_heads, head_size)
+ value = value.view(-1, num_kv_heads, head_size) + value = value.view(-1, num_kv_heads, head_size)
+ +
+ using_gqa_kernel = use_gqa_kernel(num_heads, num_kv_heads, head_size, logits_soft_cap) + if is_bmg_platform:
+ + key_cache, value_cache = kv_cache.unbind(0)
+ + ipex_ops.reshape_and_cache_flash(
+ if using_gqa_kernel: + key[:num_actual_tokens],
+ key_cache, value_cache = split_kv_cache_ipexllm( + value[:num_actual_tokens],
+ key_cache,
+ value_cache,
+ attn_metadata.slot_mapping,
+ kv_cache_dtype,
+ k_scale,
+ v_scale,
+ )
+ ipex_ops.chunked_prefill(
+ query[:num_actual_tokens].contiguous(),
+ key_cache,
+ value_cache,
+ output[:num_actual_tokens],
+ attn_metadata.query_start_loc,
+ attn_metadata.seq_start_loc,
+ None,
+ attn_metadata.block_table,
+ alibi_slopes,
+ attn_metadata.max_query_len,
+ attn_metadata.max_seq_len,
+ 0.0,
+ scale,
+ False,
+ True,
+ False,
+ None,
+ )
+ else:
+ if using_gqa_kernel:
+ key_cache, value_cache = split_kv_cache_ipexllm(
+ kv_cache, num_kv_heads, head_size)
+ ipex_ops.reshape_and_cache_ipexllm(
+ key[:num_actual_tokens],
+ value[:num_actual_tokens],
+ key_cache,
+ value_cache,
+ attn_metadata.slot_mapping.flatten(),
+ kv_cache_dtype,
+ k_scale,
+ v_scale,
+ )
+ else:
+ key_cache, value_cache = split_kv_cache(
+ kv_cache, num_kv_heads, head_size) + kv_cache, num_kv_heads, head_size)
+ ipex_ops.reshape_and_cache_ipexllm( + ipex_ops.reshape_and_cache(
+ key[:num_actual_tokens], + key[:num_actual_tokens],
+ value[:num_actual_tokens], + value[:num_actual_tokens],
+ key_cache, + key_cache,
+ value_cache, + value_cache,
+ attn_metadata.slot_mapping.flatten(), + attn_metadata.slot_mapping.flatten(),
+ kv_cache_dtype, + kv_cache_dtype,
+ k_scale, + k_scale,
+ v_scale, + v_scale,
+ ) + )
+ else: + # Invoke chunked prefill method...
+ key_cache, value_cache = split_kv_cache( + import vllm._C.ops
+ kv_cache, num_kv_heads, head_size) + assert head_size == 128 or head_size == 64
+ ipex_ops.reshape_and_cache( + value = os.environ.get('USE_CONTEXT_V1')
+ key[:num_actual_tokens], + query_len = attn_metadata.query_start_loc[1:] - attn_metadata.query_start_loc[:-1]
+ value[:num_actual_tokens], + seq_len = attn_metadata.seq_start_loc[1:] - attn_metadata.seq_start_loc[:-1]
+ key_cache, + context_len = seq_len - query_len
+ value_cache, + if using_gqa_kernel:
+ attn_metadata.slot_mapping.flatten(), + # if using_gqa_kernel, then only the v1 kernel can be used
+ kv_cache_dtype, + out = vllm._C.ops.context_attention_forward_v1(query[:num_actual_tokens], key_cache, value_cache, attn_metadata.block_table, attn_metadata.query_start_loc, seq_len, context_len, attn_metadata.max_seq_len, torch.amax(context_len).item())
+ k_scale, + elif value is None:
+ v_scale, + # Otherwise, by default use v2 attention forward kernel...
+ ) + out = vllm._C.ops.context_attention_forward_v2(query[:num_actual_tokens], key_cache, value_cache, attn_metadata.block_table, attn_metadata.query_start_loc, seq_len, context_len, attn_metadata.max_seq_len, torch.amax(context_len).item(), torch.amax(query_len).item())
+ # Invoke chunked prefill method... + else:
+ import vllm._C.ops + out = vllm._C.ops.context_attention_forward_v1(query[:num_actual_tokens], key_cache, value_cache, attn_metadata.block_table, attn_metadata.query_start_loc, seq_len, context_len, attn_metadata.max_seq_len, torch.amax(context_len).item())
+ assert head_size == 128 or head_size == 64
+ value = os.environ.get('USE_CONTEXT_V1')
+ query_len = attn_metadata.query_start_loc[1:] - attn_metadata.query_start_loc[:-1]
+ seq_len = attn_metadata.seq_start_loc[1:] - attn_metadata.seq_start_loc[:-1]
+ context_len = seq_len - query_len
+ if using_gqa_kernel:
+ # if using_gqa_kernel, then only the v1 kernel can be used
+ out = vllm._C.ops.context_attention_forward_v1(query[:num_actual_tokens], key_cache, value_cache, attn_metadata.block_table, attn_metadata.query_start_loc, seq_len, context_len, attn_metadata.max_seq_len, torch.amax(context_len).item())
+ elif value is None:
+ # Otherwise, by default use v2 attention forward kernel...
+ out = vllm._C.ops.context_attention_forward_v2(query[:num_actual_tokens], key_cache, value_cache, attn_metadata.block_table, attn_metadata.query_start_loc, seq_len, context_len, attn_metadata.max_seq_len, torch.amax(context_len).item(), torch.amax(query_len).item())
+ else:
+ out = vllm._C.ops.context_attention_forward_v1(query[:num_actual_tokens], key_cache, value_cache, attn_metadata.block_table, attn_metadata.query_start_loc, seq_len, context_len, attn_metadata.max_seq_len, torch.amax(context_len).item())
+
+ # output[:num_actual_tokens] = out
+ output[:num_actual_tokens] = out.view(out.shape[0], -1)
+ +
+ # output[:num_actual_tokens] = out
+ output[:num_actual_tokens] = out.view(out.shape[0], -1)
+ +
+ +
+ +
@ -15648,10 +15909,10 @@ index 000000000..8612d3d77
+ self.kv_caches) + self.kv_caches)
diff --git a/vllm/v1/worker/xpu_worker.py b/vllm/v1/worker/xpu_worker.py diff --git a/vllm/v1/worker/xpu_worker.py b/vllm/v1/worker/xpu_worker.py
new file mode 100644 new file mode 100644
index 000000000..1bc531e28 index 000000000..1fb0dca87
--- /dev/null --- /dev/null
+++ b/vllm/v1/worker/xpu_worker.py +++ b/vllm/v1/worker/xpu_worker.py
@@ -0,0 +1,168 @@ @@ -0,0 +1,175 @@
+# SPDX-License-Identifier: Apache-2.0 +# SPDX-License-Identifier: Apache-2.0
+import os +import os
+from typing import Optional +from typing import Optional
@ -15685,9 +15946,16 @@ index 000000000..1bc531e28
+ assert device_config.device_type == "xpu" + assert device_config.device_type == "xpu"
+ assert current_platform.is_xpu() + assert current_platform.is_xpu()
+ +
+ def load_model(self) -> None: + import os
+ self.model_runner.load_model() + lowbit = os.getenv("IPEX_LLM_LOWBIT", None)
+ if lowbit is not None:
+ from ipex_llm.vllm.xpu.model_convert import _ipex_llm_convert
+ _ipex_llm_convert(lowbit)
+ +
+
+ def compile_or_warm_up_model(self) -> None:
+ pass
+
+ # we provide this function due to `torch.xpu.mem_get_info()` doesn't + # we provide this function due to `torch.xpu.mem_get_info()` doesn't
+ # return correct free_gpu_memory on intel client GPU. We need to + # return correct free_gpu_memory on intel client GPU. We need to
+ # calculate/estiamte it. + # calculate/estiamte it.
@ -15838,7 +16106,7 @@ index 86e6d9752..ad80bf54e 100644
@dataclass(frozen=True) @dataclass(frozen=True)
diff --git a/vllm/worker/xpu_model_runner.py b/vllm/worker/xpu_model_runner.py diff --git a/vllm/worker/xpu_model_runner.py b/vllm/worker/xpu_model_runner.py
index 9d49b4385..67f07f5b1 100644 index 9d49b4385..7396b0c89 100644
--- a/vllm/worker/xpu_model_runner.py --- a/vllm/worker/xpu_model_runner.py
+++ b/vllm/worker/xpu_model_runner.py +++ b/vllm/worker/xpu_model_runner.py
@@ -5,8 +5,8 @@ import time @@ -5,8 +5,8 @@ import time
@ -16163,7 +16431,7 @@ index 9d49b4385..67f07f5b1 100644
+ slot_mapping_tensor = torch.tensor(slot_mapping, + slot_mapping_tensor = torch.tensor(slot_mapping,
+ dtype=torch.long, + dtype=torch.long,
+ device=self.device) + device=self.device)
+ if need_block_table: + if need_block_table or "bge" in self.runner.model_config.model.lower():
+ seq_lens_tensor = torch.tensor(seq_lens, + seq_lens_tensor = torch.tensor(seq_lens,
+ dtype=torch.int, + dtype=torch.int,
+ device=self.device) + device=self.device)