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

remove some useless code (#12035)

Browse source
This commit is contained in:
Yishuo Wang 2024-09-06 17:51:08 +08:00 committed by GitHub
parent d2e1b9aaff
commit 6cedb601e4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 7 additions and 20 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
python/llm/src/ipex_llm/transformers/models/llama.py
View file

@ -47,8 +47,7 @@ from ipex_llm.transformers.models.utils import init_fp8_kv_cache, append_fp8_kv_
from ipex_llm.transformers.models.utils import is_enough_kv_cache_room_4_31, \ from ipex_llm.transformers.models.utils import is_enough_kv_cache_room_4_31, \
apply_rotary_pos_emb, is_enough_kv_cache_room_4_36 apply_rotary_pos_emb, is_enough_kv_cache_room_4_36
from ipex_llm.transformers.models.utils import apply_rotary_pos_emb_no_cache_xpu from ipex_llm.transformers.models.utils import apply_rotary_pos_emb_no_cache_xpu
from ipex_llm.transformers.models.utils import use_flash_attention, use_sdp, use_sdp_fp8, \ from ipex_llm.transformers.models.utils import use_flash_attention, use_sdp, use_sdp_causal
use_sdp_causal
from ipex_llm.transformers.models.utils import mlp_fusion_check, fp16_fusion_check from ipex_llm.transformers.models.utils import mlp_fusion_check, fp16_fusion_check
from ipex_llm.transformers.models.utils import use_decoding_fast_path, get_q_proj_or_qkv_proj from ipex_llm.transformers.models.utils import use_decoding_fast_path, get_q_proj_or_qkv_proj
from transformers.modeling_outputs import BaseModelOutputWithPast from transformers.modeling_outputs import BaseModelOutputWithPast
@ -599,7 +598,7 @@ def llama_attention_forward_4_31_quantized(
kv_seq_len = key_states.shape[-2] kv_seq_len = key_states.shape[-2]
past_key_value = (key_states, value_states) past_key_value = (key_states, value_states)
if not use_sdp_fp8(q_len, key_states.shape[2], query_states): if not use_sdp(q_len, key_states.shape[2], self.head_dim, query_states):
key_states, value_states = restore_fp8_kv_cache(key_states, value_states, key_states, value_states = restore_fp8_kv_cache(key_states, value_states,
query_states.dtype) query_states.dtype)
# repeat k/v heads if n_kv_heads < n_heads # repeat k/v heads if n_kv_heads < n_heads
@ -1282,7 +1281,7 @@ def llama_attention_forward_4_41_quantized(
key_states, value_states = past_key_value.update(key_states, value_states, key_states, value_states = past_key_value.update(key_states, value_states,
self.layer_idx, cache_kwargs) self.layer_idx, cache_kwargs)
kv_seq_len = key_states.shape[-2] kv_seq_len = key_states.shape[-2]
if not use_sdp_fp8(q_len, key_states.shape[2], query_states): if not use_sdp(q_len, key_states.shape[2], self.head_dim, query_states):
key_states, value_states = restore_fp8_kv_cache(key_states, value_states, key_states, value_states = restore_fp8_kv_cache(key_states, value_states,
query_states.dtype) query_states.dtype)
key_states = repeat_kv(key_states, self.num_key_value_groups)\ key_states = repeat_kv(key_states, self.num_key_value_groups)\
@ -1873,7 +1872,7 @@ def llama_attention_forward_4_38_quantized(
key_states, value_states = past_key_value.update(key_states, value_states, key_states, value_states = past_key_value.update(key_states, value_states,
self.layer_idx, cache_kwargs) self.layer_idx, cache_kwargs)
kv_seq_len = key_states.shape[-2] kv_seq_len = key_states.shape[-2]
if not use_sdp_fp8(q_len, key_states.shape[2], query_states): if not use_sdp(q_len, key_states.shape[2], self.head_dim, query_states):
key_states, value_states = restore_fp8_kv_cache(key_states, value_states, key_states, value_states = restore_fp8_kv_cache(key_states, value_states,
query_states.dtype) query_states.dtype)
key_states = repeat_kv(key_states, self.num_key_value_groups)\ key_states = repeat_kv(key_states, self.num_key_value_groups)\

8
python/llm/src/ipex_llm/transformers/models/mistral.py
View file

@ -51,9 +51,7 @@ from ipex_llm.transformers.models.utils import init_fp8_kv_cache, append_fp8_kv_
from ipex_llm.transformers.models.utils import apply_rotary_pos_emb from ipex_llm.transformers.models.utils import apply_rotary_pos_emb
from ipex_llm.transformers.models.utils import is_enough_kv_cache_room_4_31, \ from ipex_llm.transformers.models.utils import is_enough_kv_cache_room_4_31, \
is_enough_kv_cache_room_4_36 is_enough_kv_cache_room_4_36
from ipex_llm.transformers.low_bit_linear import SYM_INT4, FP8E5, IQ2_XXS from ipex_llm.transformers.models.utils import use_flash_attention, use_sdp, use_sdp_causal
from ipex_llm.transformers.models.utils import use_flash_attention, use_sdp, use_sdp_fp8, \
use_sdp_causal
from ipex_llm.transformers.models.utils import use_decoding_fast_path from ipex_llm.transformers.models.utils import use_decoding_fast_path
from ipex_llm.transformers.models.llama import llama_decoding_fast_path_qtype_check from ipex_llm.transformers.models.llama import llama_decoding_fast_path_qtype_check
from ipex_llm.transformers.models.llama import should_use_xetla_mm_qkv from ipex_llm.transformers.models.llama import should_use_xetla_mm_qkv
@ -409,7 +407,7 @@ def mistral_attention_forward_quantized(
kv_seq_len = key_states.shape[-2] kv_seq_len = key_states.shape[-2]
past_key_value = (key_states, value_states) past_key_value = (key_states, value_states)
if not use_sdp_fp8(q_len, key_states.shape[2], query_states): if not use_sdp(q_len, key_states.shape[2], self.head_dim, query_states):
key_states, value_states = restore_fp8_kv_cache(key_states, value_states, key_states, value_states = restore_fp8_kv_cache(key_states, value_states,
query_states.dtype) query_states.dtype)
attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) attn_weights = torch.matmul(query_states, key_states.transpose(2, 3))
@ -845,7 +843,7 @@ def mistral_attention_forward_4_36_quantized(
key_states, value_states = past_key_value.update(key_states, value_states, key_states, value_states = past_key_value.update(key_states, value_states,
self.layer_idx, cache_kwargs) self.layer_idx, cache_kwargs)
kv_seq_len = key_states.shape[-2] kv_seq_len = key_states.shape[-2]
if not use_sdp_fp8(q_len, key_states.shape[2], query_states): if not use_sdp(q_len, key_states.shape[2], self.head_dim, query_states):
key_states, value_states = restore_fp8_kv_cache(key_states, value_states, key_states, value_states = restore_fp8_kv_cache(key_states, value_states,
query_states.dtype) query_states.dtype)
attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) attn_weights = torch.matmul(query_states, key_states.transpose(2, 3))

10
python/llm/src/ipex_llm/transformers/models/utils.py
View file

@ -22,7 +22,6 @@ from ipex_llm.ggml.quantize import ggml_tensor_qtype
from ipex_llm.transformers.utils import get_ipex_version, get_xpu_device_type from ipex_llm.transformers.utils import get_ipex_version, get_xpu_device_type
from ipex_llm.transformers.low_bit_linear import SYM_INT4, SYM_INT8, FP8E5, IQ2_XXS, FP4, FP8E4,\ from ipex_llm.transformers.low_bit_linear import SYM_INT4, SYM_INT8, FP8E5, IQ2_XXS, FP4, FP8E4,\
FP6, ASYM_INT4 FP6, ASYM_INT4
from ipex_llm.transformers.convert import is_deepspeed_available
FP8_KV_ALLOC_LENGTH = 512 FP8_KV_ALLOC_LENGTH = 512
KV_CACHE_ALLOC_BLOCK_LENGTH = int(os.environ.get("KV_CACHE_ALLOC_BLOCK_LENGTH", 256)) KV_CACHE_ALLOC_BLOCK_LENGTH = int(os.environ.get("KV_CACHE_ALLOC_BLOCK_LENGTH", 256))
@ -335,15 +334,6 @@ def use_sdp(q_len, kv_len, head_dim, query_states):
) )
def use_sdp_fp8(q_len, kv_len, query_states):
return (
query_states.device.type == "xpu"
and query_states.dtype in [torch.float, torch.half] # fp32/fp16
and q_len != kv_len # next token
and q_len <= 32 # lookup
)
def use_sdp_causal(q_len, kv_len, head_dim, query_states, training): def use_sdp_causal(q_len, kv_len, head_dim, query_states, training):
return ( return (
q_len == kv_len # first token q_len == kv_len # first token