Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
change
sglang
Commits
afa0341e
Unverified
Commit
afa0341e
authored
Dec 29, 2024
by
HandH1998
Committed by
GitHub
Dec 29, 2024
Browse files
Update Triton configs for block fp8 kernels (#2641)
parent
30828e71
Changes
43
Hide whitespace changes
Inline
Side-by-side
Showing
20 changed files
with
919 additions
and
483 deletions
+919
-483
benchmark/kernels/fused_moe_triton/tuning_fused_moe_triton.py
...hmark/kernels/fused_moe_triton/tuning_fused_moe_triton.py
+1
-2
python/sglang/srt/layers/moe/fused_moe_triton/configs/E=256,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
...H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+14
-14
python/sglang/srt/layers/moe/fused_moe_triton/configs/E=256,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
...me=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+11
-11
python/sglang/srt/layers/moe/fused_moe_triton/fused_moe.py
python/sglang/srt/layers/moe/fused_moe_triton/fused_moe.py
+1
-2
python/sglang/srt/layers/quantization/configs/N=1536,K=1536,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
...H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+146
-0
python/sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
...H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+33
-33
python/sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
...me=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+30
-30
python/sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
...H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+31
-31
python/sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
...me=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+35
-35
python/sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
...H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+30
-30
python/sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
...me=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+35
-35
python/sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
...H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+40
-40
python/sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
...me=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+33
-33
python/sglang/srt/layers/quantization/configs/N=256,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
...H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+146
-0
python/sglang/srt/layers/quantization/configs/N=3072,K=1536,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
...me=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+146
-0
python/sglang/srt/layers/quantization/configs/N=3072,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
...H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+35
-35
python/sglang/srt/layers/quantization/configs/N=3072,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
...me=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+36
-36
python/sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
...H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+40
-40
python/sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
...me=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
+35
-35
python/sglang/srt/layers/quantization/configs/N=36864,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
...H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
+41
-41
No files found.
benchmark/kernels/fused_moe_triton/tuning_fused_moe_triton.py
View file @
afa0341e
...
...
@@ -418,8 +418,7 @@ def main(args: argparse.Namespace):
search_space
=
[
config
for
config
in
search_space
if
block_n
%
config
[
"BLOCK_SIZE_N"
]
==
0
and
block_k
%
config
[
"BLOCK_SIZE_K"
]
==
0
if
block_k
%
config
[
"BLOCK_SIZE_K"
]
==
0
]
print
(
f
"Start tuning over
{
len
(
search_space
)
}
configurations..."
)
...
...
python/sglang/srt/layers/moe/fused_moe_triton/configs/E=256,N=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
...
...
@@ -11,7 +11,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -19,7 +19,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -27,7 +27,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -35,7 +35,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
6
4
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -43,7 +43,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -59,7 +59,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
6
4
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -67,7 +67,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -75,7 +75,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
6
4
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -99,7 +99,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -107,7 +107,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
6
4
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -123,7 +123,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -131,15 +131,15 @@
"BLOCK_SIZE_M"
:
128
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
}
...
...
python/sglang/srt/layers/moe/fused_moe_triton/configs/E=256,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
...
...
@@ -32,10 +32,10 @@
"num_stages"
:
3
},
"16"
:
{
"BLOCK_SIZE_M"
:
6
4
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"BLOCK_SIZE_M"
:
1
6
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -43,7 +43,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
6
4
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -51,7 +51,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -67,7 +67,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -75,7 +75,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -83,7 +83,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -99,7 +99,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
@@ -107,7 +107,7 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
3
},
...
...
python/sglang/srt/layers/moe/fused_moe_triton/fused_moe.py
View file @
afa0341e
...
...
@@ -475,8 +475,7 @@ def get_default_config(
"num_stages"
:
4
,
}
else
:
# Block-wise quant: BLOCK_SIZE_N must be divisable by block_shape[0]
# BLOCK_SIZE_K must be divisable by block_shape[1]
# Block-wise quant: BLOCK_SIZE_K must be divisable by block_shape[1]
config
=
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
block_shape
[
0
],
...
...
python/sglang/srt/layers/quantization/configs/N=1536,K=1536,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"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
}
}
python/sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -21,7 +21,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -29,47 +29,47 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -77,7 +77,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -85,15 +85,15 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -101,46 +101,46 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
6
4
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
...
...
@@ -5,7 +5,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -13,7 +13,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -21,39 +21,39 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -61,7 +61,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -69,31 +69,31 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -101,7 +101,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -109,38 +109,38 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -21,7 +21,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -29,7 +29,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -37,47 +37,47 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -85,31 +85,31 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
6
4
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -117,7 +117,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -125,7 +125,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -133,7 +133,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -141,6 +141,6 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -21,15 +21,15 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -37,95 +37,95 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
6
4
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
6
4
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -133,7 +133,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -141,6 +141,6 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
...
...
@@ -5,7 +5,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -13,7 +13,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -21,7 +21,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -29,7 +29,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -37,15 +37,15 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -53,15 +53,15 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
6
4
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -69,23 +69,23 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -93,7 +93,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -101,46 +101,46 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
...
...
@@ -5,47 +5,47 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -53,39 +53,39 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -93,7 +93,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -101,46 +101,46 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
6
4
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
1
6
,
"BLOCK_SIZE_N"
:
6
4
,
"BLOCK_SIZE_M"
:
6
4
,
"BLOCK_SIZE_N"
:
25
6
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -45,15 +45,15 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -61,39 +61,39 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -101,7 +101,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -109,38 +109,38 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
1
6
,
"BLOCK_SIZE_N"
:
6
4
,
"BLOCK_SIZE_M"
:
6
4
,
"BLOCK_SIZE_N"
:
25
6
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_warps"
:
8
,
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_warps"
:
8
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -21,7 +21,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -29,31 +29,31 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -61,7 +61,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -69,15 +69,15 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
6
4
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -85,15 +85,15 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -101,7 +101,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -109,38 +109,38 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=256,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
5
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
}
}
python/sglang/srt/layers/quantization/configs/N=3072,K=1536,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
4
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"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"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"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"
:
16
,
"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
}
}
python/sglang/srt/layers/quantization/configs/N=3072,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
...
...
@@ -3,17 +3,17 @@
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -21,23 +21,23 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -45,15 +45,15 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -61,31 +61,31 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -93,7 +93,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -101,46 +101,46 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=3072,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
...
...
@@ -5,31 +5,31 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -37,55 +37,55 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
6
4
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
6
4
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
6
4
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
4
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -93,7 +93,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -101,46 +101,46 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
1
6
,
"BLOCK_SIZE_N"
:
6
4
,
"BLOCK_SIZE_M"
:
6
4
,
"BLOCK_SIZE_N"
:
25
6
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
1
6
,
"BLOCK_SIZE_N"
:
6
4
,
"BLOCK_SIZE_M"
:
6
4
,
"BLOCK_SIZE_N"
:
25
6
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"GROUP_SIZE_M"
:
6
4
,
"num_warps"
:
8
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
6
4
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
6
4
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -101,7 +101,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -109,7 +109,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -117,15 +117,15 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -133,14 +133,14 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
1
6
,
"BLOCK_SIZE_N"
:
6
4
,
"BLOCK_SIZE_M"
:
6
4
,
"BLOCK_SIZE_N"
:
25
6
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_warps"
:
8
,
"num_stages"
:
4
},
"2"
:
{
"BLOCK_SIZE_M"
:
1
6
,
"BLOCK_SIZE_N"
:
6
4
,
"BLOCK_SIZE_M"
:
6
4
,
"BLOCK_SIZE_N"
:
25
6
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_warps"
:
8
,
"num_stages"
:
4
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_N"
:
256
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
8
,
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -37,31 +37,31 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -69,31 +69,31 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -101,7 +101,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -109,7 +109,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -117,7 +117,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -125,7 +125,7 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -133,14 +133,14 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
python/sglang/srt/layers/quantization/configs/N=36864,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
View file @
afa0341e
{
"1"
:
{
"BLOCK_SIZE_M"
:
1
6
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_M"
:
6
4
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
6
4
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"2"
:
{
"BLOCK_SIZE_M"
:
1
6
,
"BLOCK_SIZE_M"
:
6
4
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -29,47 +29,47 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
5
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
1
6
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
...
...
@@ -77,70 +77,70 @@
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
6
,
"GROUP_SIZE_M"
:
6
4
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
"num_stages"
:
3
}
}
Prev
1
2
3
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