Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
Menu
Open sidebar
OpenDAS
vllm_cscc
Commits
081057de
Commit
081057de
authored
Apr 29, 2025
by
zhuwenwen
Browse files
Merge tag 'v0.8.5' into v0.8.5-ori
parents
7cf5d5c4
ba41cc90
Changes
554
Hide whitespace changes
Inline
Side-by-side
Showing
20 changed files
with
2230 additions
and
18 deletions
+2230
-18
vllm/model_executor/guided_decoding/guidance_decoding.py
vllm/model_executor/guided_decoding/guidance_decoding.py
+13
-2
vllm/model_executor/guided_decoding/guided_fields.py
vllm/model_executor/guided_decoding/guided_fields.py
+6
-5
vllm/model_executor/guided_decoding/utils.py
vllm/model_executor/guided_decoding/utils.py
+1
-9
vllm/model_executor/layers/activation.py
vllm/model_executor/layers/activation.py
+1
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=1024,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json
...=1024,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json
+164
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=192,device_name=NVIDIA_H100_80GB_HBM3.json
...onfigs/E=128,N=192,device_name=NVIDIA_H100_80GB_HBM3.json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=192,device_name=NVIDIA_H20.json
...fused_moe/configs/E=128,N=192,device_name=NVIDIA_H20.json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=192,device_name=NVIDIA_H200.json
...used_moe/configs/E=128,N=192,device_name=NVIDIA_H200.json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=384,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128,128].json
...name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128,128].json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=384,device_name=NVIDIA_H20.json
...fused_moe/configs/E=128,N=384,device_name=NVIDIA_H20.json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=384,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128,128].json
...ame=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128,128].json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=384,device_name=NVIDIA_H200.json
...used_moe/configs/E=128,N=384,device_name=NVIDIA_H200.json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=512,device_name=NVIDIA_H100_80GB_HBM3.json
...onfigs/E=128,N=512,device_name=NVIDIA_H100_80GB_HBM3.json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128,128].json
...name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128,128].json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H20.json
...fused_moe/configs/E=128,N=768,device_name=NVIDIA_H20.json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128,128].json
...ame=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128,128].json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H200.json
...used_moe/configs/E=128,N=768,device_name=NVIDIA_H200.json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=128,N=96,device_name=NVIDIA_H20.json
.../fused_moe/configs/E=128,N=96,device_name=NVIDIA_H20.json
+146
-0
vllm/model_executor/layers/fused_moe/configs/E=256,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128,128].json
...name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128,128].json
+146
-0
vllm/model_executor/layers/fused_moe/configs/README
vllm/model_executor/layers/fused_moe/configs/README
+1
-2
No files found.
Too many changes to show.
To preserve performance only
554 of 554+
files are displayed.
Plain diff
Email patch
vllm/model_executor/guided_decoding/guidance_decoding.py
View file @
081057de
# SPDX-License-Identifier: Apache-2.0
import
json
from
re
import
escape
as
regex_escape
import
llguidance
...
...
@@ -7,6 +8,8 @@ from transformers import PreTrainedTokenizerBase
from
vllm.model_executor.guided_decoding.guidance_logits_processors
import
(
GuidanceLogitsProcessor
)
from
vllm.sampling_params
import
GuidedDecodingParams
from
vllm.v1.structured_output.backend_guidance
import
(
process_for_additional_properties
)
def
get_local_guidance_guided_decoding_logits_processor
(
...
...
@@ -20,9 +23,17 @@ def get_local_guidance_guided_decoding_logits_processor(
grm
=
""
any_whitespace
=
'disable-any-whitespace'
not
in
\
guided_params
.
backend_options
()
if
guided_params
.
json
:
if
(
guide_json
:
=
guided_params
.
json
)
is
not
None
:
# Optionally set additionalProperties to False at the top-level
# By default, other backends do not allow additional top-level
# properties, so this makes guidance more similar to other backends
if
'no-additional-properties'
in
guided_params
.
backend_options
():
if
not
isinstance
(
guide_json
,
str
):
guide_json
=
json
.
dumps
(
guide_json
)
guide_json
=
process_for_additional_properties
(
guide_json
)
grm
=
llguidance
.
LLMatcher
.
grammar_from_json_schema
(
guide
d_params
.
json
,
guide
_
json
,
overrides
=
{
"whitespace_pattern"
:
guided_params
.
whitespace_pattern
},
defaults
=
{
"whitespace_flexible"
:
any_whitespace
,
...
...
vllm/model_executor/guided_decoding/guided_fields.py
View file @
081057de
...
...
@@ -27,14 +27,15 @@ class GuidedDecodingRequest:
guided_decoding_backend
:
Optional
[
str
]
=
None
guided_whitespace_pattern
:
Optional
[
str
]
=
None
guided_json_object
:
Optional
[
bool
]
=
None
structural_tag
:
Optional
[
str
]
=
None
def
__post_init__
(
self
):
"""Validate that some fields are mutually exclusive."""
guide_count
=
sum
(
[
self
.
guided_json
is
not
None
,
self
.
guided_regex
is
not
None
,
self
.
guided_choice
is
not
None
,
self
.
guided_grammar
is
not
None
,
self
.
guided_json_object
is
not
None
]
)
guide_count
=
sum
(
x
is
not
None
for
x
in
(
self
.
guided_json
,
self
.
guided_regex
,
self
.
guided_choice
,
self
.
guided_grammar
,
self
.
guided_json_object
,
self
.
structural_tag
)
)
if
guide_count
>
1
:
raise
ValueError
(
"You can only use one kind of guided decoding but multiple are "
...
...
vllm/model_executor/guided_decoding/utils.py
View file @
081057de
...
...
@@ -10,16 +10,8 @@ def has_xgrammar_unsupported_json_features(schema: dict) -> bool:
if
not
isinstance
(
obj
,
dict
):
return
False
# Check for pattern restrictions
if
"pattern"
in
obj
:
return
True
# Check for numeric ranges
if
obj
.
get
(
"type"
)
in
(
"integer"
,
"number"
)
and
any
(
key
in
obj
for
key
in
[
"minimum"
,
"maximum"
,
"exclusiveMinimum"
,
"exclusiveMaximum"
,
"multipleOf"
]):
if
obj
.
get
(
"type"
)
in
(
"integer"
,
"number"
)
and
(
"multipleOf"
in
obj
):
return
True
# Check for array unsupported keywords
...
...
vllm/model_executor/layers/activation.py
View file @
081057de
...
...
@@ -354,6 +354,7 @@ def get_act_fn(act_fn_name: str) -> nn.Module:
_ACTIVATION_AND_MUL_REGISTRY
=
LazyDict
({
"gelu"
:
lambda
:
GeluAndMul
(),
"silu"
:
lambda
:
SiluAndMul
(),
"gelu_and_mul"
:
lambda
:
GeluAndMul
(),
})
...
...
vllm/model_executor/layers/fused_moe/configs/E=128,N=1024,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
16
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
2
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
16
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
16
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
2
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
2
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
2
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
4
,
"num_warps"
:
4
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"256"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"512"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
4
,
"num_warps"
:
4
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"1024"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
4
,
"num_warps"
:
8
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"1536"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
2
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
4
,
"num_warps"
:
8
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
2
,
"waves_per_eu"
:
0
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=192,device_name=NVIDIA_H100_80GB_HBM3.json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
32
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"2048"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"3072"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=192,device_name=NVIDIA_H20.json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=192,device_name=NVIDIA_H200.json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
32
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"256"
:
{
"BLOCK_SIZE_M"
:
32
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"1536"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
3
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=384,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128,128].json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=384,device_name=NVIDIA_H20.json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=384,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128,128].json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
}
}
\ No newline at end of file
vllm/model_executor/layers/fused_moe/configs/E=128,N=384,device_name=NVIDIA_H200.json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
32
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"1536"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"2048"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"3072"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"4096"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
4
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=512,device_name=NVIDIA_H100_80GB_HBM3.json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
32
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
32
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"256"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
32
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
32
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128,128].json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H20.json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128,128].json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
}
}
\ No newline at end of file
vllm/model_executor/layers/fused_moe/configs/E=128,N=768,device_name=NVIDIA_H200.json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
256
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
32
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"1536"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"2048"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"3072"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"4096"
:
{
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
8
,
"num_stages"
:
4
}
}
vllm/model_executor/layers/fused_moe/configs/E=128,N=96,device_name=NVIDIA_H20.json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
vllm/model_executor/layers/fused_moe/configs/E=256,N=128,device_name=NVIDIA_H20,dtype=fp8_w8a8,block_shape=[128,128].json
0 → 100644
View file @
081057de
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"4"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"16"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
}
}
vllm/model_executor/layers/fused_moe/configs/README
View file @
081057de
...
...
@@ -9,5 +9,4 @@ The example configurations provided are for the Mixtral model for TP2 on H100
and TP4 on A100. Mixtral has intermediate size N = 14336, i.e. for TP2 we have
N = 7168 and for TP4 we have N = 3584.
Please feel free to tune the configurations using scripts in `benchmarks/kernels/benchmark_moe.py`
Some of the configurations files are copied from the SGLang repository. Thank you!
See `benchmark/kernels/benchmark_moe.py` on how to generate these config files.
Prev
1
…
19
20
21
22
23
24
25
26
27
28
Next
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment
