huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-04-21 06:42:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

add softcap and slidingwindow

Browse Source 
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>
This commit is contained in:
Wang, Yi A 2025-04-07 22:42:19 -07:00
parent 102e29902a
commit d9e47b651c
1 changed files with 20 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
server/text_generation_server/layers/attention/ipex.py
View File

@ -8,6 +8,9 @@ from text_generation_server.models.globals import (
BLOCK_SIZE, BLOCK_SIZE,
) )
if ATTENTION == "flashdecoding-ipex":
SUPPORTS_WINDOWING = True
else:
SUPPORTS_WINDOWING = False SUPPORTS_WINDOWING = False
@ -25,8 +28,6 @@ def attention(
causal: bool = True, causal: bool = True,
softcap: Optional[float] = None, softcap: Optional[float] = None,
): ):
if softcap is not None:
raise NotImplementedError("softcap is not available in IPEX")
out = torch.empty_like(query) out = torch.empty_like(query)
kv_cache_dtype = "auto" kv_cache_dtype = "auto"
@ -37,6 +38,7 @@ def attention(
# We do not need to check window_size_left (not supported) here, so it is already checked ahead of time at model load. # We do not need to check window_size_left (not supported) here, so it is already checked ahead of time at model load.
if ATTENTION == "flashdecoding-ipex": if ATTENTION == "flashdecoding-ipex":
window_size_right = -1 if window_size_left == -1 else 0
ipex.llm.modules.PagedAttention.flash_attn_varlen_func( ipex.llm.modules.PagedAttention.flash_attn_varlen_func(
out, out,
query.contiguous() if query.device.type == "xpu" else query, query.contiguous() if query.device.type == "xpu" else query,
@ -50,11 +52,18 @@ def attention(
causal, causal,
block_tables, block_tables,
None, None,
window_size_left=window_size_left,
window_size_right=window_size_right,
kv_cache_dtype=kv_cache_dtype, kv_cache_dtype=kv_cache_dtype,
k_scale=kv_scales.key_scale_cpu, k_scale=kv_scales.key_scale_cpu,
v_scale=kv_scales.value_scale_cpu, v_scale=kv_scales.value_scale_cpu,
softcap=softcap,
) )
else: else:
if softcap is not None:
raise NotImplementedError(
"softcap is not available in IPEX paged attention"
)
ipex.llm.functional.varlen_attention( ipex.llm.functional.varlen_attention(
query.contiguous() if query.device.type == "xpu" else query, query.contiguous() if query.device.type == "xpu" else query,
key.contiguous() if key.device.type == "xpu" else key, key.contiguous() if key.device.type == "xpu" else key,
@ -88,17 +97,14 @@ def paged_attention(
softcap: Optional[float] = None, softcap: Optional[float] = None,
window_size_left: Optional[int] = -1, window_size_left: Optional[int] = -1,
): ):
if softcap is not None:
raise NotImplementedError("softcap is not available in IPEX")
out = torch.empty_like(query) out = torch.empty_like(query)
kv_cache_dtype = "auto" kv_cache_dtype = "auto"
if kv_cache.key.dtype == torch.float8_e5m2: if kv_cache.key.dtype == torch.float8_e5m2:
kv_cache_dtype = "fp8_e5m2" kv_cache_dtype = "fp8_e5m2"
if kv_cache.key.dtype == torch.float8_e4m3fn: if kv_cache.key.dtype == torch.float8_e4m3fn:
kv_cache_dtype = "fp8_e4m3" kv_cache_dtype = "fp8_e4m3"
if ATTENTION == "flashdecoding-ipex": if ATTENTION == "flashdecoding-ipex":
window_size_right = -1 if window_size_left == -1 else 0
ipex.llm.modules.PagedAttention.flash_attn_varlen_func( ipex.llm.modules.PagedAttention.flash_attn_varlen_func(
out, out,
query.contiguous() if query.device.type == "xpu" else query, query.contiguous() if query.device.type == "xpu" else query,
@ -112,12 +118,19 @@ def paged_attention(
True, True,
block_tables, block_tables,
None, None,
window_size_left=window_size_left,
window_size_right=window_size_right,
kv_cache_dtype=kv_cache_dtype, kv_cache_dtype=kv_cache_dtype,
k_scale=kv_scales.key_scale_cpu, k_scale=kv_scales.key_scale_cpu,
v_scale=kv_scales.value_scale_cpu, v_scale=kv_scales.value_scale_cpu,
softcap=softcap,
) )
else: else:
input_lengths = seqlen.input_lengths + seqlen.cache_lengths input_lengths = seqlen.input_lengths + seqlen.cache_lengths
if softcap is not None:
raise NotImplementedError(
"softcap is not available in IPEX paged attention"
)
ipex.llm.modules.PagedAttention.single_query_cached_kv_attention( ipex.llm.modules.PagedAttention.single_query_cached_kv_attention(
out, out,
query, query,
@ -130,9 +143,6 @@ def paged_attention(
BLOCK_SIZE, BLOCK_SIZE,
max_s, max_s,
None, None,
kv_cache_dtype=kv_cache_dtype,
k_scale=kv_scales.key_scale_cpu,
v_scale=kv_scales.value_scale_cpu,
) )
return out return out