Files
FastDeploy/fastdeploy/model_executor/layers/attention/iluvatar_attn_backend.py

290 lines
11 KiB
Python

"""
# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved.
#
# 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.
"""
from __future__ import annotations
from dataclasses import dataclass
from math import sqrt
from typing import TYPE_CHECKING, Optional
import paddle
from fastdeploy.config import FDConfig
from fastdeploy.model_executor.layers.attention.attention import Attention
from fastdeploy.model_executor.layers.attention.base_attention_backend import (
AttentionBackend,
AttentionMetadata,
)
from fastdeploy.model_executor.ops.iluvatar import (
mixed_fused_paged_attention,
paged_attention,
prefill_fused_paged_attention,
)
if TYPE_CHECKING:
from fastdeploy.model_executor.forward_meta import ForwardMeta
@dataclass
class IluvatarAttentionMetadata(AttentionMetadata):
"""
IluvatarAttentionMetadata
"""
alibi_slopes: Optional[paddle.Tensor] = None
window_left: int = -1
window_right: int = -1
softcap: float = 0.0
use_cuda_graph: bool = False
use_sqrt_alibi: bool = False
# qk[seq, h, d], cos/sin [seq, 1, d]
def apply_rope(qk, cos, sin):
rotate_half = paddle.reshape(
paddle.stack([-qk[..., 1::2], qk[..., 0::2]], axis=-1),
paddle.shape(qk),
)
out = paddle.add(paddle.multiply(qk, cos), paddle.multiply(rotate_half, sin))
return paddle.cast(out, qk.dtype)
class IluvatarAttnBackend(AttentionBackend):
"""
The backend class that uses paddle native attention implementation.
Which is used only for testing purpose.
"""
def __init__(self, fd_config: FDConfig, kv_num_heads: int, num_heads: int, head_dim: int):
super().__init__()
self.attention_metadata = IluvatarAttentionMetadata()
self.block_size = fd_config.parallel_config.block_size
assert self.block_size == 16, "Iluvatar paged attn requires block_size must be 16."
self.max_context_len = fd_config.parallel_config.max_model_len
self.causal = getattr(fd_config.model_config, "causal", True)
self.speculate_method = getattr(fd_config.parallel_config, "speculate_method", None)
self.use_speculate = self.speculate_method is not None
self.num_kv_heads = kv_num_heads
self.num_heads = num_heads
self.total_num_heads = num_heads + 2 * kv_num_heads
self.head_dim = head_dim
self.hidden_dim = fd_config.model_config.hidden_size
# note: scale need to change if using MLA
self.scale = 1.0 / sqrt(head_dim)
self.num_layers = fd_config.model_config.num_hidden_layers
self.dtype = paddle.get_default_dtype()
def init_attention_metadata(self, forward_meta: ForwardMeta):
"""Initialize attntion metadata hence all layers in the forward pass can reuse it."""
self.rope_cos = forward_meta.rotary_embs[0, 0, :, :, :]
self.rope_sin = forward_meta.rotary_embs[1, 0, :, :, :]
self.prefill_info_dict = {}
self.decode_info_dict = {}
self.prefill_info_dict["batch_ids"] = paddle.where(forward_meta.seq_lens_encoder)[0]
self.decode_info_dict["batch_ids"] = paddle.where(forward_meta.seq_lens_decoder)[0]
self.prefill_len = len(self.prefill_info_dict["batch_ids"])
self.decode_len = len(self.decode_info_dict["batch_ids"])
# only prefill
if self.decode_len == 0:
cu_seq_ids = list(range(self.prefill_len + 1))
self.prefill_info_dict["cu_seqlens_q"] = forward_meta.cu_seqlens_q[cu_seq_ids]
self.mixed = False
# only decode
elif self.prefill_len == 0:
self.mixed = False
# both prefill and decode
else:
self.mixed = True
self.prefill_num_tokens = paddle.sum(forward_meta.seq_lens_encoder).item()
self.prefill_info_dict["cu_seqlens_q"] = paddle.zeros(
[self.prefill_len + 1], dtype=forward_meta.cu_seqlens_q.dtype
)
self.prefill_info_dict["cu_seqlens_q"][1:] = forward_meta.seq_lens_encoder[
self.prefill_info_dict["batch_ids"], 0
]
self.prefill_info_dict["cu_seqlens_q"] = paddle.cumsum(self.prefill_info_dict["cu_seqlens_q"])
self.tmp_buffer = paddle.zeros(
[self.prefill_num_tokens + self.decode_len, self.hidden_dim], dtype=self.dtype
)
prefill_start, decode_start, start = 0, self.prefill_num_tokens, 0
non_zeros_ids = forward_meta.seq_lens_this_time != 0
non_zeros_seq_lens = forward_meta.seq_lens_this_time[non_zeros_ids]
end = non_zeros_seq_lens[0]
if end > 1:
last_stage = "prefill"
prefill_end = end
decode_end = decode_start
else:
last_stage = "decode"
prefill_end = 0
decode_end = decode_start + end
self.id_group = []
self.reverse_id_group = []
for seq_len in non_zeros_seq_lens[1:]:
if seq_len > 1:
if last_stage == "decode":
self.id_group.append((decode_start, decode_end))
self.reverse_id_group.append((start, end))
decode_start = decode_end
start = end
last_stage = "prefill"
prefill_end += seq_len
end += seq_len
else:
if last_stage == "prefill":
self.id_group.append((prefill_start, prefill_end))
self.reverse_id_group.append((start, end))
prefill_start = prefill_end
start = end
last_stage = "decode"
decode_end += seq_len
end += seq_len
if prefill_start < prefill_end:
self.id_group.append((prefill_start, prefill_end))
self.reverse_id_group.append((start, end))
if decode_start < decode_end:
self.id_group.append((decode_start, decode_end))
self.reverse_id_group.append((start, end))
def get_attntion_meta(self):
"""get_attntion_meta"""
return self.attention_metadata
def get_kv_cache_shape(
self,
max_num_blocks: int,
kv_cache_quant_type: str = None,
):
"""
Calculate kv cache shape
"""
return (
max_num_blocks,
self.num_kv_heads,
self.block_size,
self.head_dim,
)
def transpose(self, hidden_states):
for ids, reverse_ids in zip(self.id_group, self.reverse_id_group):
self.tmp_buffer[ids[0] : ids[1], :] = hidden_states[reverse_ids[0] : reverse_ids[1], :]
return self.tmp_buffer
def reverse_transpose(self, hidden_states):
for ids, reverse_ids in zip(self.id_group, self.reverse_id_group):
self.tmp_buffer[reverse_ids[0] : reverse_ids[1], :] = hidden_states[ids[0] : ids[1], :]
return self.tmp_buffer
def forward_mixed(
self,
q: paddle.Tensor,
k: paddle.Tensor,
v: paddle.Tensor,
qkv: paddle.Tensor,
compressed_kv: paddle.Tensor,
k_pe: paddle.Tensor,
layer: Attention,
forward_meta: ForwardMeta,
):
"""
forward_mixed
"""
layer_id = layer.layer_id
k_cache_id = layer_id * 2
v_cache_id = k_cache_id + 1
k_cache = forward_meta.caches[k_cache_id]
v_cache = forward_meta.caches[v_cache_id]
if self.decode_len == 0:
output = prefill_fused_paged_attention(
qkv,
k_cache,
v_cache,
block_tables=forward_meta.block_tables[self.prefill_info_dict["batch_ids"], :],
cu_seqlens_qkv=self.prefill_info_dict["cu_seqlens_q"],
num_heads=self.num_heads,
head_dim=self.head_dim,
num_kv_heads=self.num_kv_heads,
block_size=self.block_size,
max_seq_len=self.max_context_len,
scale=self.scale,
causal=self.causal,
q_rope=True,
k_rope=True,
v_rope=False,
rope_sin=self.rope_sin,
rope_cos=self.rope_cos,
)
elif self.prefill_len == 0:
output = paged_attention(
qkv,
k_cache,
v_cache,
block_tables=forward_meta.block_tables[self.decode_info_dict["batch_ids"], :],
seq_lens=forward_meta.seq_lens_decoder[self.decode_info_dict["batch_ids"], 0] + 1,
num_heads=self.num_heads,
head_dim=self.head_dim,
num_kv_heads=self.num_kv_heads,
scale=self.scale,
block_size=self.block_size,
max_context_len=self.max_context_len,
alibi_slopes=self.attention_metadata.alibi_slopes,
causal=self.causal,
window_left=self.attention_metadata.window_left,
window_right=self.attention_metadata.window_right,
softcap=self.attention_metadata.softcap,
use_cuda_graph=self.attention_metadata.use_cuda_graph,
use_sqrt_alibi=self.attention_metadata.use_sqrt_alibi,
merged_qkv=True,
k=qkv,
v=qkv,
rope_sin=self.rope_sin,
rope_cos=self.rope_cos,
)
else:
output = mixed_fused_paged_attention(
qkv,
k_cache,
v_cache,
prefill_block_tables=forward_meta.block_tables[self.prefill_info_dict["batch_ids"], :],
decode_block_tables=forward_meta.block_tables[self.decode_info_dict["batch_ids"], :],
cu_seqlens_qkv=self.prefill_info_dict["cu_seqlens_q"],
seq_lens=forward_meta.seq_lens_decoder[self.decode_info_dict["batch_ids"], 0] + 1,
prefill_num_tokens=self.prefill_num_tokens,
num_heads=self.num_heads,
head_dim=self.head_dim,
num_kv_heads=self.num_kv_heads,
block_size=self.block_size,
max_seq_len=self.max_context_len,
scale=self.scale,
causal=self.causal,
q_rope=True,
k_rope=True,
v_rope=False,
window_left=self.attention_metadata.window_left,
window_right=self.attention_metadata.window_right,
softcap=self.attention_metadata.softcap,
use_cuda_graph=self.attention_metadata.use_cuda_graph,
use_sqrt_alibi=self.attention_metadata.use_sqrt_alibi,
rope_sin=self.rope_sin,
rope_cos=self.rope_cos,
)
return output