refactor qwen2 (#11087)
This commit is contained in:
Yishuo Wang
GitHub
parent
492ed3fd41
commit
f00625f9a4
2 changed files with 97 additions and 518 deletions
|
|
@ -717,6 +717,10 @@ def _optimize_pre(model):
|
|||
# baichuan2-7B
|
||||
from ipex_llm.transformers.models.baichuan2 import pre_compute_inv_freq
|
||||
model.apply(pre_compute_inv_freq)
|
||||
# for qwen2
|
||||
if model.config.model_type == "qwen2":
|
||||
from ipex_llm.transformers.models.qwen2 import merge_qkv
|
||||
model.apply(merge_qkv)
|
||||
if model.config.model_type == "stablelm":
|
||||
# For stablelm-zephyr-3b and stablelm-2-zephyr-1_6b
|
||||
from ipex_llm.transformers.models.stablelm import merge_qkv
|
||||
|
|
|
|||
|
|
@ -42,59 +42,24 @@ import warnings
|
|||
from typing import TYPE_CHECKING, Optional, Tuple, Union, Callable, List
|
||||
|
||||
import torch
|
||||
import torch.nn as nn
|
||||
import torch.nn.functional as F
|
||||
from torch.nn.functional import scaled_dot_product_attention as sdpa
|
||||
|
||||
from ipex_llm.transformers.models.llama import repeat_kv
|
||||
from ipex_llm.transformers.models.utils import extend_kv_cache, append_kv_cache
|
||||
from ipex_llm.transformers.models.utils import should_use_fuse_rope
|
||||
from ipex_llm.transformers.models.utils import use_quantize_kv_cache, restore_fp8_kv_cache
|
||||
from ipex_llm.transformers.models.utils import is_enough_kv_cache_room_4_36
|
||||
from ipex_llm.transformers.models.utils import apply_rotary_pos_emb_cache_freq_xpu
|
||||
from ipex_llm.transformers.kv import DynamicFp8Cache
|
||||
from ipex_llm.transformers.models.utils import use_flash_attention, use_sdp, use_sdp_causal
|
||||
from ipex_llm.transformers.kv import DynamicFp8Cache, DynamicNormalCache
|
||||
from ipex_llm.utils.common import invalidInputError
|
||||
from ipex_llm.transformers.models.utils import use_flash_attention, use_sdp
|
||||
from transformers.models.qwen2.modeling_qwen2 import Qwen2Model, apply_rotary_pos_emb
|
||||
|
||||
from transformers.models.qwen2.modeling_qwen2 import Qwen2Attention, apply_rotary_pos_emb, repeat_kv
|
||||
from transformers.models.qwen2.modeling_qwen2 import _prepare_4d_causal_attention_mask_for_sdpa
|
||||
from transformers.models.qwen2.modeling_qwen2 import _prepare_4d_causal_attention_mask
|
||||
from transformers.modeling_outputs import BaseModelOutputWithPast
|
||||
from ipex_llm.transformers.models.utils import use_decoding_fast_path
|
||||
|
||||
try:
|
||||
from transformers.cache_utils import Cache, DynamicCache
|
||||
except ImportError:
|
||||
Cache = Tuple[torch.Tensor]
|
||||
import logging
|
||||
from transformers import logging
|
||||
|
||||
|
||||
logger = logging.get_logger(__name__)
|
||||
|
||||
import os
|
||||
|
||||
KV_CACHE_ALLOC_BLOCK_LENGTH = int(os.environ.get("KV_CACHE_ALLOC_BLOCK_LENGTH", 256))
|
||||
|
||||
|
||||
def should_split_qkv_tensor(query_states, bsz, num_heads, q_len, kv_seq_len, output_attentions):
|
||||
if not output_attentions:
|
||||
if os.environ.get("IPEX_LLM_SPLIT_QKV", None) is not None:
|
||||
return os.environ.get("IPEX_LLM_SPLIT_QKV", None) == "1"
|
||||
elif query_states.dtype == torch.float16 and \
|
||||
query_states.shape[2] >= 5000:
|
||||
# split tensor for memory block limitation
|
||||
# support fp16 and set input length threshold at 5000 for now
|
||||
return True
|
||||
elif query_states.element_size()*bsz*num_heads*q_len*kv_seq_len >= 4*1024**3:
|
||||
# attn_weight size larger than memory block limitation 4GB
|
||||
return True
|
||||
return False
|
||||
|
||||
|
||||
def should_use_fuse_rope(self, query_states, position_ids):
|
||||
use_fuse_rope = query_states.device.type == "xpu"
|
||||
use_fuse_rope = use_fuse_rope and not (self.training and query_states.requires_grad)
|
||||
use_fuse_rope = use_fuse_rope and position_ids is not None
|
||||
return use_fuse_rope
|
||||
|
||||
|
||||
def qwen2_model_forward(
|
||||
self,
|
||||
|
|
@ -109,9 +74,12 @@ def qwen2_model_forward(
|
|||
return_dict: Optional[bool] = None,
|
||||
):
|
||||
use_cache = use_cache if use_cache is not None else self.config.use_cache
|
||||
if use_cache and use_quantize_kv_cache(self.layers[0].mlp.up_proj, input_ids):
|
||||
if not isinstance(past_key_values, DynamicFp8Cache):
|
||||
use_quantize_kv = use_quantize_kv_cache(self.layers[0].mlp.up_proj, input_ids)
|
||||
if use_cache:
|
||||
if use_quantize_kv and not isinstance(past_key_values, DynamicFp8Cache):
|
||||
past_key_values = DynamicFp8Cache.from_legacy_cache(past_key_values)
|
||||
if not use_quantize_kv and not isinstance(past_key_values, DynamicNormalCache):
|
||||
past_key_values = DynamicNormalCache.from_legacy_cache(past_key_values)
|
||||
return qwen2_model_forward_internal(
|
||||
self=self,
|
||||
input_ids=input_ids,
|
||||
|
|
@ -248,13 +216,13 @@ def qwen2_model_forward_internal(
|
|||
use_cache,
|
||||
)
|
||||
else:
|
||||
# bigdl-llm changes
|
||||
# ipex-llm changes
|
||||
curr_device = decoder_layer.input_layernorm.weight.device
|
||||
if attention_mask is not None:
|
||||
attention_mask = attention_mask.to(curr_device)
|
||||
if position_ids is not None:
|
||||
position_ids = position_ids.to(curr_device)
|
||||
# bigdl-llm changes end
|
||||
# ipex-llm changes end
|
||||
layer_outputs = decoder_layer(
|
||||
hidden_states,
|
||||
attention_mask=attention_mask,
|
||||
|
|
@ -294,510 +262,117 @@ def qwen2_model_forward_internal(
|
|||
)
|
||||
|
||||
|
||||
def merge_qkv(module: torch.nn.Module):
|
||||
if isinstance(module, Qwen2Attention):
|
||||
new_weight = torch.cat([
|
||||
module.q_proj.weight.data,
|
||||
module.k_proj.weight.data,
|
||||
module.v_proj.weight.data,
|
||||
], dim=0)
|
||||
new_bias = torch.cat([
|
||||
module.q_proj.bias.data,
|
||||
module.k_proj.bias.data,
|
||||
module.v_proj.bias.data,
|
||||
], dim=-1)
|
||||
|
||||
qkv_proj = torch.nn.Linear(0, 0, bias=True)
|
||||
qkv_proj.weight = torch.nn.Parameter(new_weight, requires_grad=False)
|
||||
qkv_proj.bias = torch.nn.Parameter(new_bias, requires_grad=False)
|
||||
qkv_proj.in_features = new_weight.size(1)
|
||||
qkv_proj.out_features = new_weight.size(0)
|
||||
module.qkv_proj = qkv_proj
|
||||
|
||||
del module.q_proj, module.k_proj, module.v_proj
|
||||
|
||||
|
||||
def qwen2_attention_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,
|
||||
past_key_value: Optional[Cache] = None,
|
||||
output_attentions: bool = False,
|
||||
use_cache: bool = False,
|
||||
**kwargs,
|
||||
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
||||
if use_quantize_kv_cache(self.q_proj, hidden_states):
|
||||
forward_function = qwen2_attention_forward_quantized
|
||||
elif hidden_states.device.type == "cpu":
|
||||
forward_function = qwen2_sdpa_attention_forward
|
||||
else:
|
||||
forward_function = qwen2_attention_forward_origin
|
||||
return forward_function(
|
||||
self=self,
|
||||
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,
|
||||
**kwargs,
|
||||
)
|
||||
|
||||
|
||||
def qwen2_attention_forward_quantized(
|
||||
self,
|
||||
hidden_states: torch.Tensor,
|
||||
attention_mask: Optional[torch.Tensor] = None,
|
||||
position_ids: Optional[torch.LongTensor] = None,
|
||||
past_key_value: Optional[DynamicFp8Cache] = None,
|
||||
output_attentions: bool = False,
|
||||
use_cache: bool = False,
|
||||
**kwargs,
|
||||
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
||||
if "padding_mask" in kwargs:
|
||||
warnings.warn(
|
||||
"Passing `padding_mask` is deprecated and will be removed in v4.37. "
|
||||
"Please make sure use `attention_mask` instead.`"
|
||||
)
|
||||
use_fuse_rope = should_use_fuse_rope(self, hidden_states, position_ids)
|
||||
bsz, q_len, _ = hidden_states.size()
|
||||
device = hidden_states.device
|
||||
|
||||
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)
|
||||
qkv = self.qkv_proj(hidden_states)
|
||||
qkv = qkv.view(bsz, q_len, self.num_heads + 2 * self.num_key_value_heads, self.head_dim)
|
||||
qkv = qkv.transpose(1, 2)
|
||||
query_states, key_states, value_states = qkv.split([self.num_heads,
|
||||
self.num_key_value_heads,
|
||||
self.num_key_value_heads], dim=1)
|
||||
|
||||
kv_seq_len = key_states.shape[-2]
|
||||
if past_key_value is not None:
|
||||
invalidInputError(self.layer_idx is not None,
|
||||
"The cache structure has changed since version v4.36. "
|
||||
f"If you are using {self.__class__.__name__} "
|
||||
"for auto-regressive decoding with k/v caching, "
|
||||
"please make sure to initialize the attention class "
|
||||
"with a layer index.")
|
||||
kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
|
||||
cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
|
||||
|
||||
if use_fuse_rope:
|
||||
query_states, key_states = apply_rotary_pos_emb_cache_freq_xpu(query_states, key_states,
|
||||
sin, cos, "qwen2",
|
||||
position_ids)
|
||||
if should_use_fuse_rope(hidden_states, position_ids, self.training):
|
||||
import linear_q4_0
|
||||
linear_q4_0.rotary_half_inplaced(self.rotary_emb.inv_freq, position_ids,
|
||||
query_states, key_states)
|
||||
else:
|
||||
cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
|
||||
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} # Specific to RoPE models
|
||||
key_states, value_states = past_key_value.update(key_states, value_states,
|
||||
self.layer_idx, cache_kwargs)
|
||||
self.layer_idx, None)
|
||||
|
||||
if q_len == 1 and query_states.device.type == 'xpu' and not self.training \
|
||||
and not hidden_states.requires_grad:
|
||||
import linear_q4_0
|
||||
attn_output = linear_q4_0.sdp_fp8(query_states, key_states, value_states,
|
||||
attention_mask)
|
||||
attn_weights = None
|
||||
else:
|
||||
key, value = restore_fp8_kv_cache(key_states, value_states, query_states.dtype)
|
||||
key = repeat_kv(key, self.num_key_value_groups)
|
||||
value = repeat_kv(value, self.num_key_value_groups)
|
||||
if should_split_qkv_tensor(query_states, bsz, self.num_heads,
|
||||
q_len, kv_seq_len, output_attentions):
|
||||
attn_output, attn_weights = native_sdp_split_qkv_tensor(query_states, key,
|
||||
value, attention_mask,
|
||||
bsz, q_len, kv_seq_len,
|
||||
self.head_dim, self.num_heads,
|
||||
self.attention_dropout,
|
||||
self.training)
|
||||
else:
|
||||
attn_weights = torch.matmul(query_states, key.transpose(2, 3))
|
||||
attn_weights = attn_weights / math.sqrt(self.head_dim)
|
||||
|
||||
invalidInputError(attn_weights.size() == (bsz, self.num_heads, q_len, kv_seq_len),
|
||||
("Attention weights should be of size "
|
||||
f"{(bsz, self.num_heads, q_len, kv_seq_len)},"
|
||||
"but is {attn_weights.size()}"))
|
||||
|
||||
if attention_mask is not None:
|
||||
invalidInputError(attention_mask.size() == (bsz, 1, q_len, kv_seq_len),
|
||||
(f"Attention mask should be of size "
|
||||
f"{(bsz, 1, q_len, kv_seq_len)},"
|
||||
f" but is {attention_mask.size()}"))
|
||||
|
||||
attn_weights = attn_weights + attention_mask
|
||||
|
||||
if kv_seq_len >= 2048 or bsz >= 64:
|
||||
# for memory considerations, do not upcast attention to fp32
|
||||
# for long sequences or large batches
|
||||
attn_weights = nn.functional.softmax(attn_weights, dim=-1)
|
||||
else:
|
||||
# upcast attention to fp32
|
||||
attn_weights = nn.functional.softmax(attn_weights, dim=-1,
|
||||
dtype=torch.float32).to(query_states.dtype)
|
||||
attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout,
|
||||
training=self.training)
|
||||
|
||||
attn_output = torch.matmul(attn_weights, value)
|
||||
|
||||
invalidInputError(attn_output.size() == (bsz, self.num_heads, q_len, self.head_dim),
|
||||
"`attn_output` should be of size "
|
||||
f"{(bsz, self.num_heads, q_len, self.head_dim)},"
|
||||
f" but is {attn_output.size()}")
|
||||
|
||||
attn_output = attn_output.transpose(1, 2).contiguous()
|
||||
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
|
||||
|
||||
attn_output = self.o_proj(attn_output)
|
||||
|
||||
if not output_attentions:
|
||||
attn_weights = None
|
||||
|
||||
return attn_output, attn_weights, past_key_value
|
||||
from ipex_llm.ggml.quantize import ggml_tensor_qtype
|
||||
SYM_INT4 = ggml_tensor_qtype["sym_int4"]
|
||||
FP8E5 = ggml_tensor_qtype["fp8_e5m2"]
|
||||
|
||||
|
||||
def qwen2_attention_forward_origin(
|
||||
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: bool = False,
|
||||
use_cache: bool = False,
|
||||
**kwargs,
|
||||
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
||||
|
||||
use_fuse_rope = should_use_fuse_rope(self, hidden_states, position_ids)
|
||||
|
||||
if "padding_mask" in kwargs:
|
||||
warnings.warn(
|
||||
"Passing `padding_mask` is deprecated and will be removed in v4.37. "
|
||||
"Please make sure use `attention_mask` instead.`"
|
||||
)
|
||||
bsz, q_len, _ = hidden_states.size()
|
||||
device = hidden_states.device
|
||||
|
||||
enough_kv_room = is_enough_kv_cache_room_4_36(past_key_value, self.layer_idx)
|
||||
decoding_fast_path = use_decoding_fast_path(self.q_proj,
|
||||
use_fuse_rope,
|
||||
enough_kv_room,
|
||||
bsz * q_len)
|
||||
if decoding_fast_path:
|
||||
hidden_states = hidden_states.view(1, -1)
|
||||
cache_k = past_key_value.key_cache[self.layer_idx]
|
||||
cache_v = past_key_value.value_cache[self.layer_idx]
|
||||
kv_seq_len = cache_k.shape[-2]
|
||||
import linear_q4_0
|
||||
args = [hidden_states, self.q_proj.weight, self.k_proj.weight, self.v_proj.weight,
|
||||
self.q_proj.bias, self.k_proj.bias, self.v_proj.bias, position_ids, cache_k,
|
||||
cache_v, self.q_proj.weight.qtype, self.v_proj.weight.qtype, kv_seq_len,
|
||||
self.head_dim, self.rotary_emb.base]
|
||||
query_states, key_states, value_states = linear_q4_0.forward_qkv_bias(*args)
|
||||
kv_seq_len += 1
|
||||
if self.layer_idx == 0:
|
||||
past_key_value.seen_tokens = kv_seq_len
|
||||
past_key_value.key_cache[self.layer_idx] = key_states
|
||||
past_key_value.value_cache[self.layer_idx] = value_states
|
||||
|
||||
else:
|
||||
|
||||
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)
|
||||
|
||||
kv_seq_len = key_states.shape[-2]
|
||||
if past_key_value is not None:
|
||||
if self.layer_idx is None:
|
||||
invalidInputError(
|
||||
False,
|
||||
"The cache structure has changed since version v4.36. "
|
||||
f"If you are using {self.__class__.__name__} "
|
||||
"for auto-regressive decoding with k/v caching, "
|
||||
"please make sure to initialize the attention class with a layer index."
|
||||
)
|
||||
kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
|
||||
cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
|
||||
if use_fuse_rope:
|
||||
query_states, key_states = apply_rotary_pos_emb_cache_freq_xpu(query_states, key_states,
|
||||
sin, cos, "qwen2",
|
||||
position_ids)
|
||||
else:
|
||||
query_states, key_states = apply_rotary_pos_emb(query_states, key_states,
|
||||
cos, sin, position_ids)
|
||||
|
||||
if past_key_value is not None:
|
||||
# update the number of seen tokens
|
||||
if self.layer_idx == 0:
|
||||
past_key_value.seen_tokens += key_states.shape[-2]
|
||||
|
||||
if len(past_key_value.key_cache) <= self.layer_idx:
|
||||
past_key_value.key_cache.append(key_states)
|
||||
past_key_value.value_cache.append(value_states)
|
||||
else:
|
||||
cache_k = past_key_value.key_cache[self.layer_idx]
|
||||
cache_v = past_key_value.value_cache[self.layer_idx]
|
||||
|
||||
if not enough_kv_room:
|
||||
# allocate new
|
||||
new_c_k, new_c_v = extend_kv_cache(bsz,
|
||||
self.num_key_value_heads, # Support GQA
|
||||
self.head_dim,
|
||||
cache_k.size(2),
|
||||
kv_seq_len + KV_CACHE_ALLOC_BLOCK_LENGTH,
|
||||
dtype=cache_k.dtype,
|
||||
device=device)
|
||||
|
||||
new_c_k[:] = cache_k
|
||||
new_c_v[:] = cache_v
|
||||
cache_k = new_c_k
|
||||
cache_v = new_c_v
|
||||
|
||||
key_states, value_states = append_kv_cache(cache_k,
|
||||
cache_v,
|
||||
key_states,
|
||||
value_states)
|
||||
|
||||
# update past_key_value
|
||||
past_key_value.key_cache[self.layer_idx] = key_states
|
||||
past_key_value.value_cache[self.layer_idx] = value_states
|
||||
|
||||
# repeat k/v heads if n_kv_heads < n_heads
|
||||
key_states = repeat_kv(key_states, self.num_key_value_groups)
|
||||
value_states = repeat_kv(value_states, self.num_key_value_groups)
|
||||
|
||||
if not self.training and not hidden_states.requires_grad and \
|
||||
use_flash_attention(query_states, key_states, attention_mask):
|
||||
attn_output = F.scaled_dot_product_attention(query_states.to(device, dtype=torch.float16),
|
||||
key_states.to(device, dtype=torch.float16),
|
||||
value_states.to(device, dtype=torch.float16),
|
||||
is_causal=True)
|
||||
attn_weights = None
|
||||
elif not self.training and not hidden_states.requires_grad and \
|
||||
use_sdp(q_len, key_states.shape[2], self.head_dim, query_states):
|
||||
import linear_q4_0
|
||||
attn_output = linear_q4_0.sdp(query_states, key_states, value_states, attention_mask)
|
||||
attn_output = attn_output.view(query_states.shape)
|
||||
attn_weights = None
|
||||
else:
|
||||
if should_split_qkv_tensor(query_states, bsz, self.num_heads,
|
||||
q_len, kv_seq_len, output_attentions):
|
||||
attn_output, attn_weights = native_sdp_split_qkv_tensor(query_states, key_states,
|
||||
value_states, attention_mask,
|
||||
bsz, q_len, kv_seq_len,
|
||||
self.head_dim, self.num_heads,
|
||||
self.attention_dropout,
|
||||
self.training)
|
||||
else:
|
||||
attn_weights = torch.matmul(query_states,
|
||||
key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
|
||||
|
||||
invalidInputError(attn_weights.size() == (bsz, self.num_heads, q_len, kv_seq_len),
|
||||
("Attention weights should be of size "
|
||||
f"{(bsz, self.num_heads, q_len, kv_seq_len)},"
|
||||
"but is {attn_weights.size()}"))
|
||||
|
||||
if attention_mask is not None:
|
||||
invalidInputError(attention_mask.size() == (bsz, 1, q_len, kv_seq_len),
|
||||
(f"Attention mask should be of size "
|
||||
f"{(bsz, 1, q_len, kv_seq_len)},"
|
||||
f" but is {attention_mask.size()}"))
|
||||
|
||||
attn_weights = attn_weights + attention_mask
|
||||
|
||||
if kv_seq_len >= 2048 or bsz >= 64:
|
||||
# for memory considerations, do not upcast attention to fp32
|
||||
# for long sequences or large batches
|
||||
attn_weights = nn.functional.softmax(attn_weights, dim=-1)
|
||||
else:
|
||||
# upcast attention to fp32
|
||||
attn_weights = nn.functional.softmax(attn_weights, dim=-1,
|
||||
dtype=torch.float32).to(query_states.dtype)
|
||||
attn_weights = nn.functional.dropout(attn_weights,
|
||||
p=self.attention_dropout,
|
||||
training=self.training)
|
||||
attn_output = torch.matmul(attn_weights, value_states)
|
||||
|
||||
invalidInputError(attn_output.size() == (bsz, self.num_heads, q_len, self.head_dim),
|
||||
"`attn_output` should be of size "
|
||||
f"{(bsz, self.num_heads, q_len, self.head_dim)},"
|
||||
f" but is {attn_output.size()}")
|
||||
|
||||
attn_output = attn_output.transpose(1, 2).contiguous()
|
||||
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
|
||||
|
||||
attn_output = self.o_proj(attn_output)
|
||||
|
||||
if not output_attentions:
|
||||
attn_weights = None
|
||||
|
||||
return attn_output.to(hidden_states.dtype), attn_weights, past_key_value
|
||||
|
||||
|
||||
def qwen2_sdpa_attention_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: bool = False,
|
||||
use_cache: bool = False,
|
||||
**kwargs,
|
||||
) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
|
||||
|
||||
use_fuse_rope = should_use_fuse_rope(self, hidden_states, position_ids)
|
||||
|
||||
if "padding_mask" in kwargs:
|
||||
warnings.warn(
|
||||
"Passing `padding_mask` is deprecated and will be removed in v4.37. "
|
||||
"Please make sure use `attention_mask` instead.`"
|
||||
)
|
||||
bsz, q_len, _ = hidden_states.size()
|
||||
device = hidden_states.device
|
||||
|
||||
enough_kv_room = is_enough_kv_cache_room_4_36(past_key_value, self.layer_idx)
|
||||
decoding_fast_path = use_decoding_fast_path(self.q_proj,
|
||||
use_fuse_rope,
|
||||
enough_kv_room,
|
||||
bsz * q_len)
|
||||
if decoding_fast_path:
|
||||
hidden_states = hidden_states.view(1, -1)
|
||||
cache_k = past_key_value.key_cache[self.layer_idx]
|
||||
cache_v = past_key_value.value_cache[self.layer_idx]
|
||||
kv_seq_len = cache_k.shape[-2]
|
||||
import linear_q4_0
|
||||
args = [hidden_states, self.q_proj.weight, self.k_proj.weight, self.v_proj.weight,
|
||||
self.q_proj.bias, self.k_proj.bias, self.v_proj.bias, position_ids, cache_k,
|
||||
cache_v, self.q_proj.weight.qtype, self.v_proj.weight.qtype, kv_seq_len,
|
||||
self.head_dim, self.rotary_emb.base]
|
||||
query_states, key_states, value_states = linear_q4_0.forward_qkv_bias(*args)
|
||||
kv_seq_len += 1
|
||||
if self.layer_idx == 0:
|
||||
past_key_value.seen_tokens = kv_seq_len
|
||||
past_key_value.key_cache[self.layer_idx] = key_states
|
||||
past_key_value.value_cache[self.layer_idx] = value_states
|
||||
|
||||
else:
|
||||
|
||||
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)
|
||||
|
||||
kv_seq_len = key_states.shape[-2]
|
||||
if past_key_value is not None:
|
||||
if self.layer_idx is None:
|
||||
invalidInputError(
|
||||
False,
|
||||
"The cache structure has changed since version v4.36. "
|
||||
f"If you are using {self.__class__.__name__} "
|
||||
"for auto-regressive decoding with k/v caching, "
|
||||
"please make sure to initialize the attention class with a layer index."
|
||||
)
|
||||
kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
|
||||
cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
|
||||
if use_fuse_rope:
|
||||
query_states, key_states = apply_rotary_pos_emb_cache_freq_xpu(query_states, key_states,
|
||||
sin, cos, "qwen2",
|
||||
position_ids)
|
||||
else:
|
||||
query_states, key_states = apply_rotary_pos_emb(query_states, key_states,
|
||||
cos, sin, position_ids)
|
||||
|
||||
if past_key_value is not None:
|
||||
# update the number of seen tokens
|
||||
if self.layer_idx == 0:
|
||||
past_key_value.seen_tokens += key_states.shape[-2]
|
||||
|
||||
if len(past_key_value.key_cache) <= self.layer_idx:
|
||||
past_key_value.key_cache.append(key_states)
|
||||
past_key_value.value_cache.append(value_states)
|
||||
else:
|
||||
cache_k = past_key_value.key_cache[self.layer_idx]
|
||||
cache_v = past_key_value.value_cache[self.layer_idx]
|
||||
|
||||
if not enough_kv_room:
|
||||
# allocate new
|
||||
new_c_k, new_c_v = extend_kv_cache(bsz,
|
||||
self.num_key_value_heads, # Support GQA
|
||||
self.head_dim,
|
||||
cache_k.size(2),
|
||||
kv_seq_len + KV_CACHE_ALLOC_BLOCK_LENGTH,
|
||||
dtype=cache_k.dtype,
|
||||
device=device)
|
||||
|
||||
new_c_k[:] = cache_k
|
||||
new_c_v[:] = cache_v
|
||||
cache_k = new_c_k
|
||||
cache_v = new_c_v
|
||||
|
||||
key_states, value_states = append_kv_cache(cache_k,
|
||||
cache_v,
|
||||
key_states,
|
||||
value_states)
|
||||
|
||||
# update past_key_value
|
||||
past_key_value.key_cache[self.layer_idx] = key_states
|
||||
past_key_value.value_cache[self.layer_idx] = value_states
|
||||
|
||||
# repeat k/v heads if n_kv_heads < n_heads
|
||||
key_states = repeat_kv(key_states, self.num_key_value_groups)
|
||||
value_states = repeat_kv(value_states, self.num_key_value_groups)
|
||||
|
||||
attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
|
||||
|
||||
invalidInputError(attn_weights.size() == (bsz, self.num_heads, q_len, kv_seq_len),
|
||||
("Attention weights should be of size "
|
||||
f"{(bsz, self.num_heads, q_len, kv_seq_len)},"
|
||||
"but is {attn_weights.size()}"))
|
||||
|
||||
if attention_mask is not None:
|
||||
invalidInputError(attention_mask.size() == (bsz, 1, q_len, kv_seq_len),
|
||||
(f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}"
|
||||
f" but is {attention_mask.size()}"))
|
||||
|
||||
attn_weights = attn_weights + attention_mask
|
||||
|
||||
from torch.nn.functional import scaled_dot_product_attention as sdpa
|
||||
if query_states.device.type == "cpu":
|
||||
attn_output = sdpa(query_states,
|
||||
key_states,
|
||||
value_states,
|
||||
attn_mask=attention_mask,
|
||||
dropout_p=self.attention_dropout if self.training else 0.0,
|
||||
is_causal=self.is_causal and attention_mask is None and q_len > 1)
|
||||
elif not self.training and not hidden_states.requires_grad and \
|
||||
use_flash_attention(query_states, key_states, attention_mask):
|
||||
attn_output = sdpa(query_states.to(device, dtype=torch.float16),
|
||||
key_states.to(device, dtype=torch.float16),
|
||||
value_states.to(device, dtype=torch.float16),
|
||||
is_causal=True).to(hidden_states.dtype)
|
||||
elif use_sdp(q_len, kv_seq_len, self.head_dim, query_states):
|
||||
import linear_q4_0
|
||||
if isinstance(past_key_value, DynamicFp8Cache):
|
||||
attn_output = linear_q4_0.sdp_fp8(query_states, key_states, value_states,
|
||||
attention_mask)
|
||||
else:
|
||||
attn_output = linear_q4_0.sdp(query_states, key_states, value_states, attention_mask)
|
||||
elif use_sdp_causal(q_len, kv_seq_len, self.head_dim, query_states, self.training):
|
||||
import linear_q4_0
|
||||
if isinstance(past_key_value, DynamicFp8Cache):
|
||||
attn_output = linear_q4_0.sdp_fp8_causal(query_states, key_states, value_states)
|
||||
else:
|
||||
attn_output = linear_q4_0.sdp_causal(query_states, key_states, value_states)
|
||||
else:
|
||||
if isinstance(past_key_value, DynamicFp8Cache):
|
||||
key_states, value_states = restore_fp8_kv_cache(key_states, value_states,
|
||||
query_states.dtype)
|
||||
# repeat k/v heads if n_kv_heads < n_heads
|
||||
key_states = repeat_kv(key_states, self.num_key_value_groups)
|
||||
value_states = repeat_kv(value_states, self.num_key_value_groups)
|
||||
|
||||
attn_weights = torch.matmul(query_states,
|
||||
key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
|
||||
if attention_mask is not None:
|
||||
attn_weights = attn_weights + attention_mask
|
||||
# upcast attention to fp32
|
||||
attn_weights = torch.nn.functional.softmax(attn_weights, dim=-1,
|
||||
dtype=torch.float32).to(query_states.dtype)
|
||||
attn_weights = torch.nn.functional.dropout(attn_weights, p=self.attention_dropout,
|
||||
training=self.training)
|
||||
attn_output = torch.matmul(attn_weights, value_states)
|
||||
|
||||
attn_output = attn_output.transpose(1, 2).contiguous()
|
||||
attn_output = attn_output.view(bsz, q_len, self.hidden_size)
|
||||
attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
|
||||
|
||||
attn_output = self.o_proj(attn_output)
|
||||
|
||||
return attn_output, None, past_key_value
|
||||
|
||||
|
||||
def native_sdp_split_qkv_tensor(query, key, value, attention_mask,
|
||||
bsz, q_len, kv_seq_len, head_dim, num_heads,
|
||||
attention_dropout, training):
|
||||
block_size = 8
|
||||
query_split = torch.split(query, block_size, dim=1)
|
||||
key_split = torch.split(key.transpose(2, 3), block_size, dim=1)
|
||||
value_split = torch.split(value, block_size, dim=1)
|
||||
attn_outputs = []
|
||||
for q, k, v in zip(query_split, key_split, value_split):
|
||||
attn_weights_split = torch.matmul(q, k) / math.sqrt(head_dim)
|
||||
block_actual_size = attn_weights_split.size(1)
|
||||
attn_weights_split_size = (bsz, block_actual_size, q_len, kv_seq_len)
|
||||
if attn_weights_split.size() != attn_weights_split_size:
|
||||
invalidInputError(False,
|
||||
f"Splitted attention weights should be of size "
|
||||
f"{attn_weights_split_size}, but is {attn_weights_split.size()}")
|
||||
|
||||
if attention_mask is not None:
|
||||
attn_mask_size = (bsz, 1, q_len, kv_seq_len)
|
||||
if attention_mask.size() != attn_mask_size:
|
||||
invalidInputError(False,
|
||||
f"Attention mask should be of size {attn_mask_size}, "
|
||||
f"but is {attention_mask.size()}")
|
||||
attn_weights_split = attn_weights_split + attention_mask
|
||||
attn_weights_split = nn.functional.softmax(attn_weights_split, dim=-1)
|
||||
attn_weights_split = nn.functional.dropout(attn_weights_split,
|
||||
p=attention_dropout,
|
||||
training=training)
|
||||
attn_outputs.append(torch.matmul(attn_weights_split, v))
|
||||
attn_output = torch.cat(attn_outputs, dim=1)
|
||||
return attn_output, None
|
||||
if not output_attentions:
|
||||
attn_weights = None
|
||||
return attn_output, attn_weights, past_key_value
|
||||
|
|
|
|||
Loading…
Reference in a new issue