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
7863e436
Unverified
Commit
7863e436
authored
Dec 28, 2024
by
Yineng Zhang
Committed by
GitHub
Dec 28, 2024
Browse files
add configs for block fp8 related kernels (#2628)
Co-authored-by:
HandH1998
<
1335248067@qq.com
>
parent
333e3bfd
Changes
37
Hide whitespace changes
Inline
Side-by-side
Showing
17 changed files
with
2405 additions
and
16 deletions
+2405
-16
python/sglang/srt/layers/quantization/configs/N=4608,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=4608,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
+146
-0
python/sglang/srt/layers/quantization/configs/N=576,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=576,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
+146
-0
python/sglang/srt/layers/quantization/configs/N=7168,K=1024,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=7168,K=1024,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=7168,K=1152,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=7168,K=1152,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=7168,K=16384,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=7168,K=16384,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=7168,K=18432,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=7168,K=18432,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=7168,K=2048,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=7168,K=2048,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=7168,K=2304,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=7168,K=2304,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/fp8_kernel.py
python/sglang/srt/layers/quantization/fp8_kernel.py
+69
-16
No files found.
python/sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"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"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=1152,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=1152,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json
0 → 100644
View file @
7863e436
{
"1"
:
{
"BLOCK_SIZE_M"
:
16
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"8"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"16"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
16
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"24"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"32"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"48"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"64"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"96"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"128"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
32
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"256"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"512"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1024"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"1536"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"2048"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"3072"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
64
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
1
,
"num_warps"
:
4
,
"num_stages"
:
2
},
"4096"
:
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
128
,
"BLOCK_SIZE_K"
:
128
,
"GROUP_SIZE_M"
:
64
,
"num_warps"
:
4
,
"num_stages"
:
2
}
}
python/sglang/srt/layers/quantization/fp8_kernel.py
View file @
7863e436
...
...
@@ -12,12 +12,20 @@
# limitations under the License.
# ==============================================================================
from
typing
import
List
,
Tuple
import
functools
import
json
import
logging
import
os
from
typing
import
Any
,
Dict
,
List
,
Optional
,
Tuple
import
torch
import
triton
import
triton.language
as
tl
from
sglang.srt.utils
import
get_device_name
logger
=
logging
.
getLogger
(
__name__
)
@
triton
.
jit
def
_per_token_group_quant_fp8
(
...
...
@@ -205,6 +213,48 @@ def _w8a8_block_fp8_matmul(
tl
.
store
(
c_ptrs
,
c
,
mask
=
c_mask
)
@
functools
.
lru_cache
def
get_w8a8_block_fp8_configs
(
N
:
int
,
K
:
int
,
block_n
:
int
,
block_k
:
int
)
->
Optional
[
Dict
[
int
,
Any
]]:
"""
Return optimized configurations for the w8a8 block fp8 kernel.
The return value will be a dictionary that maps an irregular grid of
batch sizes to configurations of the w8a8 block fp8 kernel. To evaluate the
kernel on a given batch size bs, the closest batch size in the grid should
be picked and the associated configuration chosen to invoke the kernel.
"""
# First look up if an optimized configuration is available in the configs
# directory
device_name
=
get_device_name
().
replace
(
" "
,
"_"
)
json_file_name
=
f
"N=
{
N
}
,K=
{
K
}
,device_name=
{
device_name
}
,dtype=fp8_w8a8,block_shape=[
{
block_n
}
,
{
block_k
}
].json"
config_file_path
=
os
.
path
.
join
(
os
.
path
.
dirname
(
os
.
path
.
realpath
(
__file__
)),
"configs"
,
json_file_name
)
if
os
.
path
.
exists
(
config_file_path
):
with
open
(
config_file_path
)
as
f
:
logger
.
info
(
"Using configuration from %s for W8A8 Block FP8 kernel."
,
config_file_path
,
)
# If a configuration has been found, return it
return
{
int
(
key
):
val
for
key
,
val
in
json
.
load
(
f
).
items
()}
# If no optimized configuration is available, we will use the default
# configuration
logger
.
warning
(
(
"Using default W8A8 Block FP8 kernel config. Performance might be sub-optimal! "
"Config file not found at %s"
),
config_file_path
,
)
return
None
def
w8a8_block_fp8_matmul
(
A
:
torch
.
Tensor
,
B
:
torch
.
Tensor
,
...
...
@@ -245,17 +295,23 @@ def w8a8_block_fp8_matmul(
C_shape
=
A
.
shape
[:
-
1
]
+
(
N
,)
C
=
A
.
new_empty
(
C_shape
,
dtype
=
output_dtype
)
# TODO(HandH1998):
# BLOCK_SIZE_M, BLOCK_SIZE_K, BLOCK_SIZE_N can be optimized.
# BLOCK_SIZE_K must be divisable by block_k
# BLOCK_SIZE_N and BLOCK_SIZE_M has no requirements
BLOCK_SIZE_M
=
128
if
M
<
BLOCK_SIZE_M
:
BLOCK_SIZE_M
=
triton
.
next_power_of_2
(
M
)
BLOCK_SIZE_M
=
max
(
BLOCK_SIZE_M
,
16
)
BLOCK_SIZE_K
=
block_k
assert
block_k
%
BLOCK_SIZE_K
==
0
BLOCK_SIZE_N
=
block_n
configs
=
get_w8a8_block_fp8_configs
(
N
,
K
,
block_size
[
0
],
block_size
[
1
])
if
configs
:
# If an optimal configuration map has been found, look up the
# optimal config
config
=
configs
[
min
(
configs
.
keys
(),
key
=
lambda
x
:
abs
(
x
-
M
))]
else
:
# Default config
# Block-wise quant: BLOCK_SIZE_N must be divisable by block_size[0]
# BLOCK_SIZE_K must be divisable by block_size[1]
config
=
{
"BLOCK_SIZE_M"
:
64
,
"BLOCK_SIZE_N"
:
block_size
[
0
],
"BLOCK_SIZE_K"
:
block_size
[
1
],
"GROUP_SIZE_M"
:
32
,
"num_warps"
:
4
,
"num_stages"
:
2
,
}
def
grid
(
META
):
return
(
...
...
@@ -283,10 +339,7 @@ def w8a8_block_fp8_matmul(
As
.
stride
(
-
1
),
Bs
.
stride
(
1
),
Bs
.
stride
(
0
),
BLOCK_SIZE_M
=
BLOCK_SIZE_M
,
BLOCK_SIZE_N
=
BLOCK_SIZE_N
,
BLOCK_SIZE_K
=
BLOCK_SIZE_K
,
GROUP_SIZE_M
=
8
,
**
config
,
)
return
C
Prev
1
2
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
HandH1998 <1335248067@qq.com>HandH1998 <1335248067@qq.com>