sglang 0.4.1.post2__py3-none-any.whl → 0.4.1.post3__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=14336,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=14336,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=1792,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=3072,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=3072,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=3072,device_name=NVIDIA_H100_80GB_HBM3.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=3584,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=1,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=1344,device_name=NVIDIA_A100-SXM4-40GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=1344,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=1344,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=14336,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=14336,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=1792,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=2688,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=2688,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=3072,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=3072,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=3200,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +130 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=3584,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json +218 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=6400,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +130 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=7168,device_name=NVIDIA_A100-SXM4-80GB,dtype=int8_w8a16.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=int8_w8a16.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=16,N=800,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +130 -0
- 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 +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=256,N=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=1280,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=1280,device_name=NVIDIA_A800-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=1280,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=2560,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=320,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_A800-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=64,N=640,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=14336,device_name=AMD_Instinct_MI300X.json +200 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=14336,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +138 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=AMD_Instinct_MI300X.json +200 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=NVIDIA_A100-SXM4-40GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=2048,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=2048,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=2048,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=AMD_Instinct_MI300X.json +200 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_A100-SXM4-40GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_GeForce_RTX_4090,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=3584,device_name=NVIDIA_L40S.json +173 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json +178 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=4096,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=7168,device_name=AMD_Instinct_MI300X.json +200 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=7168,device_name=NVIDIA_A100-SXM4-80GB.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=7168,device_name=NVIDIA_H100_80GB_HBM3.json +146 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=8192,device_name=AMD_Instinct_MI300X,dtype=fp8_w8a8.json +175 -0
- sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=8192,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8.json +146 -0
- sglang/srt/layers/quantization/configs/N=1536,K=1536,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=1536,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=2048,K=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=2304,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=24576,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=256,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=3072,K=1536,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=3072,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=3072,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=32768,K=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=36864,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=36864,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=4096,K=512,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=4096,K=512,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=4608,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=512,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=576,K=7168,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=1024,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=1024,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=1152,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=1152,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=128,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=16384,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=18432,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=2048,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=NVIDIA_H100_80GB_HBM3,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=2304,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/layers/quantization/configs/N=7168,K=256,device_name=NVIDIA_H200,dtype=fp8_w8a8,block_shape=[128, 128].json +146 -0
- sglang/srt/managers/detokenizer_manager.py +2 -0
- sglang/srt/managers/io_struct.py +11 -1
- sglang/srt/managers/scheduler.py +9 -1
- sglang/srt/managers/tokenizer_manager.py +7 -0
- sglang/srt/server_args.py +7 -0
- sglang/version.py +1 -1
- {sglang-0.4.1.post2.dist-info → sglang-0.4.1.post3.dist-info}/METADATA +1 -1
- sglang-0.4.1.post3.dist-info/RECORD +305 -0
- sglang-0.4.1.post2.dist-info/RECORD +0 -197
- {sglang-0.4.1.post2.dist-info → sglang-0.4.1.post3.dist-info}/LICENSE +0 -0
- {sglang-0.4.1.post2.dist-info → sglang-0.4.1.post3.dist-info}/WHEEL +0 -0
- {sglang-0.4.1.post2.dist-info → sglang-0.4.1.post3.dist-info}/top_level.txt +0 -0
@@ -0,0 +1,138 @@
|
|
1
|
+
{
|
2
|
+
"1": {
|
3
|
+
"BLOCK_SIZE_M": 64,
|
4
|
+
"BLOCK_SIZE_N": 64,
|
5
|
+
"BLOCK_SIZE_K": 128,
|
6
|
+
"GROUP_SIZE_M": 64,
|
7
|
+
"num_warps": 8,
|
8
|
+
"num_stages": 5
|
9
|
+
},
|
10
|
+
"2": {
|
11
|
+
"BLOCK_SIZE_M": 64,
|
12
|
+
"BLOCK_SIZE_N": 128,
|
13
|
+
"BLOCK_SIZE_K": 256,
|
14
|
+
"GROUP_SIZE_M": 64,
|
15
|
+
"num_warps": 4,
|
16
|
+
"num_stages": 3
|
17
|
+
},
|
18
|
+
"4": {
|
19
|
+
"BLOCK_SIZE_M": 64,
|
20
|
+
"BLOCK_SIZE_N": 128,
|
21
|
+
"BLOCK_SIZE_K": 256,
|
22
|
+
"GROUP_SIZE_M": 1,
|
23
|
+
"num_warps": 4,
|
24
|
+
"num_stages": 4
|
25
|
+
},
|
26
|
+
"8": {
|
27
|
+
"BLOCK_SIZE_M": 64,
|
28
|
+
"BLOCK_SIZE_N": 128,
|
29
|
+
"BLOCK_SIZE_K": 256,
|
30
|
+
"GROUP_SIZE_M": 32,
|
31
|
+
"num_warps": 4,
|
32
|
+
"num_stages": 4
|
33
|
+
},
|
34
|
+
"16": {
|
35
|
+
"BLOCK_SIZE_M": 64,
|
36
|
+
"BLOCK_SIZE_N": 64,
|
37
|
+
"BLOCK_SIZE_K": 128,
|
38
|
+
"GROUP_SIZE_M": 1,
|
39
|
+
"num_warps": 4,
|
40
|
+
"num_stages": 3
|
41
|
+
},
|
42
|
+
"24": {
|
43
|
+
"BLOCK_SIZE_M": 64,
|
44
|
+
"BLOCK_SIZE_N": 128,
|
45
|
+
"BLOCK_SIZE_K": 256,
|
46
|
+
"GROUP_SIZE_M": 1,
|
47
|
+
"num_warps": 4,
|
48
|
+
"num_stages": 4
|
49
|
+
},
|
50
|
+
"32": {
|
51
|
+
"BLOCK_SIZE_M": 64,
|
52
|
+
"BLOCK_SIZE_N": 128,
|
53
|
+
"BLOCK_SIZE_K": 256,
|
54
|
+
"GROUP_SIZE_M": 1,
|
55
|
+
"num_warps": 4,
|
56
|
+
"num_stages": 4
|
57
|
+
},
|
58
|
+
"48": {
|
59
|
+
"BLOCK_SIZE_M": 64,
|
60
|
+
"BLOCK_SIZE_N": 128,
|
61
|
+
"BLOCK_SIZE_K": 256,
|
62
|
+
"GROUP_SIZE_M": 1,
|
63
|
+
"num_warps": 4,
|
64
|
+
"num_stages": 4
|
65
|
+
},
|
66
|
+
"64": {
|
67
|
+
"BLOCK_SIZE_M": 64,
|
68
|
+
"BLOCK_SIZE_N": 128,
|
69
|
+
"BLOCK_SIZE_K": 256,
|
70
|
+
"GROUP_SIZE_M": 1,
|
71
|
+
"num_warps": 4,
|
72
|
+
"num_stages": 4
|
73
|
+
},
|
74
|
+
"96": {
|
75
|
+
"BLOCK_SIZE_M": 64,
|
76
|
+
"BLOCK_SIZE_N": 128,
|
77
|
+
"BLOCK_SIZE_K": 256,
|
78
|
+
"GROUP_SIZE_M": 1,
|
79
|
+
"num_warps": 4,
|
80
|
+
"num_stages": 2
|
81
|
+
},
|
82
|
+
"128": {
|
83
|
+
"BLOCK_SIZE_M": 64,
|
84
|
+
"BLOCK_SIZE_N": 128,
|
85
|
+
"BLOCK_SIZE_K": 256,
|
86
|
+
"GROUP_SIZE_M": 1,
|
87
|
+
"num_warps": 4,
|
88
|
+
"num_stages": 2
|
89
|
+
},
|
90
|
+
"256": {
|
91
|
+
"BLOCK_SIZE_M": 128,
|
92
|
+
"BLOCK_SIZE_N": 128,
|
93
|
+
"BLOCK_SIZE_K": 128,
|
94
|
+
"GROUP_SIZE_M": 1,
|
95
|
+
"num_warps": 8,
|
96
|
+
"num_stages": 3
|
97
|
+
},
|
98
|
+
"512": {
|
99
|
+
"BLOCK_SIZE_M": 128,
|
100
|
+
"BLOCK_SIZE_N": 256,
|
101
|
+
"BLOCK_SIZE_K": 128,
|
102
|
+
"GROUP_SIZE_M": 64,
|
103
|
+
"num_warps": 8,
|
104
|
+
"num_stages": 4
|
105
|
+
},
|
106
|
+
"1024": {
|
107
|
+
"BLOCK_SIZE_M": 128,
|
108
|
+
"BLOCK_SIZE_N": 256,
|
109
|
+
"BLOCK_SIZE_K": 128,
|
110
|
+
"GROUP_SIZE_M": 64,
|
111
|
+
"num_warps": 8,
|
112
|
+
"num_stages": 4
|
113
|
+
},
|
114
|
+
"1536": {
|
115
|
+
"BLOCK_SIZE_M": 128,
|
116
|
+
"BLOCK_SIZE_N": 256,
|
117
|
+
"BLOCK_SIZE_K": 128,
|
118
|
+
"GROUP_SIZE_M": 64,
|
119
|
+
"num_warps": 8,
|
120
|
+
"num_stages": 3
|
121
|
+
},
|
122
|
+
"2048": {
|
123
|
+
"BLOCK_SIZE_M": 128,
|
124
|
+
"BLOCK_SIZE_N": 256,
|
125
|
+
"BLOCK_SIZE_K": 128,
|
126
|
+
"GROUP_SIZE_M": 64,
|
127
|
+
"num_warps": 8,
|
128
|
+
"num_stages": 3
|
129
|
+
},
|
130
|
+
"3072": {
|
131
|
+
"BLOCK_SIZE_M": 128,
|
132
|
+
"BLOCK_SIZE_N": 256,
|
133
|
+
"BLOCK_SIZE_K": 128,
|
134
|
+
"GROUP_SIZE_M": 32,
|
135
|
+
"num_warps": 8,
|
136
|
+
"num_stages": 3
|
137
|
+
}
|
138
|
+
}
|
sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=AMD_Instinct_MI300X.json
ADDED
@@ -0,0 +1,200 @@
|
|
1
|
+
{
|
2
|
+
"1": {
|
3
|
+
"BLOCK_SIZE_M": 16,
|
4
|
+
"BLOCK_SIZE_N": 32,
|
5
|
+
"BLOCK_SIZE_K": 256,
|
6
|
+
"GROUP_SIZE_M": 1,
|
7
|
+
"num_warps": 2,
|
8
|
+
"num_stages": 0,
|
9
|
+
"waves_per_eu": 0,
|
10
|
+
"matrix_instr_nonkdim": 16,
|
11
|
+
"kpack": 2
|
12
|
+
},
|
13
|
+
"2": {
|
14
|
+
"BLOCK_SIZE_M": 16,
|
15
|
+
"BLOCK_SIZE_N": 64,
|
16
|
+
"BLOCK_SIZE_K": 128,
|
17
|
+
"GROUP_SIZE_M": 1,
|
18
|
+
"num_warps": 4,
|
19
|
+
"num_stages": 0,
|
20
|
+
"waves_per_eu": 0,
|
21
|
+
"matrix_instr_nonkdim": 16,
|
22
|
+
"kpack": 1
|
23
|
+
},
|
24
|
+
"4": {
|
25
|
+
"BLOCK_SIZE_M": 16,
|
26
|
+
"BLOCK_SIZE_N": 64,
|
27
|
+
"BLOCK_SIZE_K": 128,
|
28
|
+
"GROUP_SIZE_M": 1,
|
29
|
+
"num_warps": 4,
|
30
|
+
"num_stages": 0,
|
31
|
+
"waves_per_eu": 0,
|
32
|
+
"matrix_instr_nonkdim": 16,
|
33
|
+
"kpack": 2
|
34
|
+
},
|
35
|
+
"8": {
|
36
|
+
"BLOCK_SIZE_M": 16,
|
37
|
+
"BLOCK_SIZE_N": 16,
|
38
|
+
"BLOCK_SIZE_K": 256,
|
39
|
+
"GROUP_SIZE_M": 1,
|
40
|
+
"num_warps": 2,
|
41
|
+
"num_stages": 0,
|
42
|
+
"waves_per_eu": 0,
|
43
|
+
"matrix_instr_nonkdim": 16,
|
44
|
+
"kpack": 2
|
45
|
+
},
|
46
|
+
"16": {
|
47
|
+
"BLOCK_SIZE_M": 64,
|
48
|
+
"BLOCK_SIZE_N": 64,
|
49
|
+
"BLOCK_SIZE_K": 64,
|
50
|
+
"GROUP_SIZE_M": 1,
|
51
|
+
"num_warps": 8,
|
52
|
+
"num_stages": 0,
|
53
|
+
"waves_per_eu": 0,
|
54
|
+
"matrix_instr_nonkdim": 16,
|
55
|
+
"kpack": 1
|
56
|
+
},
|
57
|
+
"24": {
|
58
|
+
"BLOCK_SIZE_M": 16,
|
59
|
+
"BLOCK_SIZE_N": 64,
|
60
|
+
"BLOCK_SIZE_K": 64,
|
61
|
+
"GROUP_SIZE_M": 1,
|
62
|
+
"num_warps": 4,
|
63
|
+
"num_stages": 0,
|
64
|
+
"waves_per_eu": 0,
|
65
|
+
"matrix_instr_nonkdim": 16,
|
66
|
+
"kpack": 1
|
67
|
+
},
|
68
|
+
"32": {
|
69
|
+
"BLOCK_SIZE_M": 16,
|
70
|
+
"BLOCK_SIZE_N": 16,
|
71
|
+
"BLOCK_SIZE_K": 256,
|
72
|
+
"GROUP_SIZE_M": 4,
|
73
|
+
"num_warps": 2,
|
74
|
+
"num_stages": 0,
|
75
|
+
"waves_per_eu": 0,
|
76
|
+
"matrix_instr_nonkdim": 16,
|
77
|
+
"kpack": 2
|
78
|
+
},
|
79
|
+
"48": {
|
80
|
+
"BLOCK_SIZE_M": 16,
|
81
|
+
"BLOCK_SIZE_N": 128,
|
82
|
+
"BLOCK_SIZE_K": 64,
|
83
|
+
"GROUP_SIZE_M": 1,
|
84
|
+
"num_warps": 8,
|
85
|
+
"num_stages": 0,
|
86
|
+
"waves_per_eu": 0,
|
87
|
+
"matrix_instr_nonkdim": 16,
|
88
|
+
"kpack": 2
|
89
|
+
},
|
90
|
+
"64": {
|
91
|
+
"BLOCK_SIZE_M": 16,
|
92
|
+
"BLOCK_SIZE_N": 64,
|
93
|
+
"BLOCK_SIZE_K": 64,
|
94
|
+
"GROUP_SIZE_M": 1,
|
95
|
+
"num_warps": 2,
|
96
|
+
"num_stages": 0,
|
97
|
+
"waves_per_eu": 0,
|
98
|
+
"matrix_instr_nonkdim": 16,
|
99
|
+
"kpack": 1
|
100
|
+
},
|
101
|
+
"96": {
|
102
|
+
"BLOCK_SIZE_M": 32,
|
103
|
+
"BLOCK_SIZE_N": 32,
|
104
|
+
"BLOCK_SIZE_K": 256,
|
105
|
+
"GROUP_SIZE_M": 4,
|
106
|
+
"num_warps": 4,
|
107
|
+
"num_stages": 0,
|
108
|
+
"waves_per_eu": 0,
|
109
|
+
"matrix_instr_nonkdim": 16,
|
110
|
+
"kpack": 2
|
111
|
+
},
|
112
|
+
"128": {
|
113
|
+
"BLOCK_SIZE_M": 64,
|
114
|
+
"BLOCK_SIZE_N": 64,
|
115
|
+
"BLOCK_SIZE_K": 64,
|
116
|
+
"GROUP_SIZE_M": 4,
|
117
|
+
"num_warps": 8,
|
118
|
+
"num_stages": 0,
|
119
|
+
"waves_per_eu": 0,
|
120
|
+
"matrix_instr_nonkdim": 16,
|
121
|
+
"kpack": 1
|
122
|
+
},
|
123
|
+
"256": {
|
124
|
+
"BLOCK_SIZE_M": 128,
|
125
|
+
"BLOCK_SIZE_N": 128,
|
126
|
+
"BLOCK_SIZE_K": 64,
|
127
|
+
"GROUP_SIZE_M": 4,
|
128
|
+
"num_warps": 8,
|
129
|
+
"num_stages": 0,
|
130
|
+
"waves_per_eu": 0,
|
131
|
+
"matrix_instr_nonkdim": 16,
|
132
|
+
"kpack": 1
|
133
|
+
},
|
134
|
+
"512": {
|
135
|
+
"BLOCK_SIZE_M": 64,
|
136
|
+
"BLOCK_SIZE_N": 64,
|
137
|
+
"BLOCK_SIZE_K": 128,
|
138
|
+
"GROUP_SIZE_M": 1,
|
139
|
+
"num_warps": 8,
|
140
|
+
"num_stages": 0,
|
141
|
+
"waves_per_eu": 0,
|
142
|
+
"matrix_instr_nonkdim": 16,
|
143
|
+
"kpack": 2
|
144
|
+
},
|
145
|
+
"1024": {
|
146
|
+
"BLOCK_SIZE_M": 128,
|
147
|
+
"BLOCK_SIZE_N": 128,
|
148
|
+
"BLOCK_SIZE_K": 64,
|
149
|
+
"GROUP_SIZE_M": 1,
|
150
|
+
"num_warps": 8,
|
151
|
+
"num_stages": 0,
|
152
|
+
"waves_per_eu": 0,
|
153
|
+
"matrix_instr_nonkdim": 16,
|
154
|
+
"kpack": 1
|
155
|
+
},
|
156
|
+
"1536": {
|
157
|
+
"BLOCK_SIZE_M": 128,
|
158
|
+
"BLOCK_SIZE_N": 128,
|
159
|
+
"BLOCK_SIZE_K": 64,
|
160
|
+
"GROUP_SIZE_M": 1,
|
161
|
+
"num_warps": 8,
|
162
|
+
"num_stages": 0,
|
163
|
+
"waves_per_eu": 0,
|
164
|
+
"matrix_instr_nonkdim": 16,
|
165
|
+
"kpack": 1
|
166
|
+
},
|
167
|
+
"2048": {
|
168
|
+
"BLOCK_SIZE_M": 128,
|
169
|
+
"BLOCK_SIZE_N": 128,
|
170
|
+
"BLOCK_SIZE_K": 64,
|
171
|
+
"GROUP_SIZE_M": 1,
|
172
|
+
"num_warps": 8,
|
173
|
+
"num_stages": 0,
|
174
|
+
"waves_per_eu": 0,
|
175
|
+
"matrix_instr_nonkdim": 16,
|
176
|
+
"kpack": 2
|
177
|
+
},
|
178
|
+
"3072": {
|
179
|
+
"BLOCK_SIZE_M": 128,
|
180
|
+
"BLOCK_SIZE_N": 128,
|
181
|
+
"BLOCK_SIZE_K": 64,
|
182
|
+
"GROUP_SIZE_M": 1,
|
183
|
+
"num_warps": 8,
|
184
|
+
"num_stages": 0,
|
185
|
+
"waves_per_eu": 0,
|
186
|
+
"matrix_instr_nonkdim": 16,
|
187
|
+
"kpack": 1
|
188
|
+
},
|
189
|
+
"4096": {
|
190
|
+
"BLOCK_SIZE_M": 128,
|
191
|
+
"BLOCK_SIZE_N": 128,
|
192
|
+
"BLOCK_SIZE_K": 64,
|
193
|
+
"GROUP_SIZE_M": 1,
|
194
|
+
"num_warps": 8,
|
195
|
+
"num_stages": 0,
|
196
|
+
"waves_per_eu": 0,
|
197
|
+
"matrix_instr_nonkdim": 16,
|
198
|
+
"kpack": 1
|
199
|
+
}
|
200
|
+
}
|
sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=NVIDIA_A100-SXM4-40GB.json
ADDED
@@ -0,0 +1,146 @@
|
|
1
|
+
{
|
2
|
+
"1": {
|
3
|
+
"BLOCK_SIZE_M": 16,
|
4
|
+
"BLOCK_SIZE_N": 256,
|
5
|
+
"BLOCK_SIZE_K": 64,
|
6
|
+
"GROUP_SIZE_M": 1,
|
7
|
+
"num_warps": 8,
|
8
|
+
"num_stages": 4
|
9
|
+
},
|
10
|
+
"2": {
|
11
|
+
"BLOCK_SIZE_M": 16,
|
12
|
+
"BLOCK_SIZE_N": 32,
|
13
|
+
"BLOCK_SIZE_K": 64,
|
14
|
+
"GROUP_SIZE_M": 16,
|
15
|
+
"num_warps": 4,
|
16
|
+
"num_stages": 4
|
17
|
+
},
|
18
|
+
"4": {
|
19
|
+
"BLOCK_SIZE_M": 16,
|
20
|
+
"BLOCK_SIZE_N": 64,
|
21
|
+
"BLOCK_SIZE_K": 64,
|
22
|
+
"GROUP_SIZE_M": 64,
|
23
|
+
"num_warps": 8,
|
24
|
+
"num_stages": 4
|
25
|
+
},
|
26
|
+
"8": {
|
27
|
+
"BLOCK_SIZE_M": 16,
|
28
|
+
"BLOCK_SIZE_N": 32,
|
29
|
+
"BLOCK_SIZE_K": 128,
|
30
|
+
"GROUP_SIZE_M": 1,
|
31
|
+
"num_warps": 8,
|
32
|
+
"num_stages": 4
|
33
|
+
},
|
34
|
+
"16": {
|
35
|
+
"BLOCK_SIZE_M": 16,
|
36
|
+
"BLOCK_SIZE_N": 128,
|
37
|
+
"BLOCK_SIZE_K": 128,
|
38
|
+
"GROUP_SIZE_M": 16,
|
39
|
+
"num_warps": 8,
|
40
|
+
"num_stages": 4
|
41
|
+
},
|
42
|
+
"24": {
|
43
|
+
"BLOCK_SIZE_M": 32,
|
44
|
+
"BLOCK_SIZE_N": 128,
|
45
|
+
"BLOCK_SIZE_K": 64,
|
46
|
+
"GROUP_SIZE_M": 32,
|
47
|
+
"num_warps": 4,
|
48
|
+
"num_stages": 4
|
49
|
+
},
|
50
|
+
"32": {
|
51
|
+
"BLOCK_SIZE_M": 32,
|
52
|
+
"BLOCK_SIZE_N": 256,
|
53
|
+
"BLOCK_SIZE_K": 64,
|
54
|
+
"GROUP_SIZE_M": 64,
|
55
|
+
"num_warps": 4,
|
56
|
+
"num_stages": 4
|
57
|
+
},
|
58
|
+
"48": {
|
59
|
+
"BLOCK_SIZE_M": 16,
|
60
|
+
"BLOCK_SIZE_N": 32,
|
61
|
+
"BLOCK_SIZE_K": 64,
|
62
|
+
"GROUP_SIZE_M": 16,
|
63
|
+
"num_warps": 8,
|
64
|
+
"num_stages": 4
|
65
|
+
},
|
66
|
+
"64": {
|
67
|
+
"BLOCK_SIZE_M": 32,
|
68
|
+
"BLOCK_SIZE_N": 256,
|
69
|
+
"BLOCK_SIZE_K": 64,
|
70
|
+
"GROUP_SIZE_M": 64,
|
71
|
+
"num_warps": 4,
|
72
|
+
"num_stages": 4
|
73
|
+
},
|
74
|
+
"96": {
|
75
|
+
"BLOCK_SIZE_M": 32,
|
76
|
+
"BLOCK_SIZE_N": 128,
|
77
|
+
"BLOCK_SIZE_K": 128,
|
78
|
+
"GROUP_SIZE_M": 1,
|
79
|
+
"num_warps": 4,
|
80
|
+
"num_stages": 4
|
81
|
+
},
|
82
|
+
"128": {
|
83
|
+
"BLOCK_SIZE_M": 32,
|
84
|
+
"BLOCK_SIZE_N": 64,
|
85
|
+
"BLOCK_SIZE_K": 128,
|
86
|
+
"GROUP_SIZE_M": 32,
|
87
|
+
"num_warps": 8,
|
88
|
+
"num_stages": 4
|
89
|
+
},
|
90
|
+
"256": {
|
91
|
+
"BLOCK_SIZE_M": 64,
|
92
|
+
"BLOCK_SIZE_N": 256,
|
93
|
+
"BLOCK_SIZE_K": 64,
|
94
|
+
"GROUP_SIZE_M": 64,
|
95
|
+
"num_warps": 8,
|
96
|
+
"num_stages": 4
|
97
|
+
},
|
98
|
+
"512": {
|
99
|
+
"BLOCK_SIZE_M": 64,
|
100
|
+
"BLOCK_SIZE_N": 256,
|
101
|
+
"BLOCK_SIZE_K": 64,
|
102
|
+
"GROUP_SIZE_M": 32,
|
103
|
+
"num_warps": 8,
|
104
|
+
"num_stages": 4
|
105
|
+
},
|
106
|
+
"1024": {
|
107
|
+
"BLOCK_SIZE_M": 64,
|
108
|
+
"BLOCK_SIZE_N": 256,
|
109
|
+
"BLOCK_SIZE_K": 64,
|
110
|
+
"GROUP_SIZE_M": 64,
|
111
|
+
"num_warps": 8,
|
112
|
+
"num_stages": 4
|
113
|
+
},
|
114
|
+
"1536": {
|
115
|
+
"BLOCK_SIZE_M": 64,
|
116
|
+
"BLOCK_SIZE_N": 256,
|
117
|
+
"BLOCK_SIZE_K": 64,
|
118
|
+
"GROUP_SIZE_M": 32,
|
119
|
+
"num_warps": 8,
|
120
|
+
"num_stages": 4
|
121
|
+
},
|
122
|
+
"2048": {
|
123
|
+
"BLOCK_SIZE_M": 64,
|
124
|
+
"BLOCK_SIZE_N": 256,
|
125
|
+
"BLOCK_SIZE_K": 64,
|
126
|
+
"GROUP_SIZE_M": 16,
|
127
|
+
"num_warps": 8,
|
128
|
+
"num_stages": 4
|
129
|
+
},
|
130
|
+
"3072": {
|
131
|
+
"BLOCK_SIZE_M": 128,
|
132
|
+
"BLOCK_SIZE_N": 128,
|
133
|
+
"BLOCK_SIZE_K": 64,
|
134
|
+
"GROUP_SIZE_M": 16,
|
135
|
+
"num_warps": 8,
|
136
|
+
"num_stages": 4
|
137
|
+
},
|
138
|
+
"4096": {
|
139
|
+
"BLOCK_SIZE_M": 128,
|
140
|
+
"BLOCK_SIZE_N": 128,
|
141
|
+
"BLOCK_SIZE_K": 64,
|
142
|
+
"GROUP_SIZE_M": 16,
|
143
|
+
"num_warps": 8,
|
144
|
+
"num_stages": 4
|
145
|
+
}
|
146
|
+
}
|
sglang/srt/layers/moe/fused_moe_triton/configs/E=8,N=1792,device_name=NVIDIA_A100-SXM4-80GB.json
ADDED
@@ -0,0 +1,146 @@
|
|
1
|
+
{
|
2
|
+
"1": {
|
3
|
+
"BLOCK_SIZE_M": 16,
|
4
|
+
"BLOCK_SIZE_N": 32,
|
5
|
+
"BLOCK_SIZE_K": 128,
|
6
|
+
"GROUP_SIZE_M": 16,
|
7
|
+
"num_warps": 8,
|
8
|
+
"num_stages": 4
|
9
|
+
},
|
10
|
+
"2": {
|
11
|
+
"BLOCK_SIZE_M": 16,
|
12
|
+
"BLOCK_SIZE_N": 64,
|
13
|
+
"BLOCK_SIZE_K": 64,
|
14
|
+
"GROUP_SIZE_M": 16,
|
15
|
+
"num_warps": 4,
|
16
|
+
"num_stages": 4
|
17
|
+
},
|
18
|
+
"4": {
|
19
|
+
"BLOCK_SIZE_M": 16,
|
20
|
+
"BLOCK_SIZE_N": 128,
|
21
|
+
"BLOCK_SIZE_K": 128,
|
22
|
+
"GROUP_SIZE_M": 1,
|
23
|
+
"num_warps": 4,
|
24
|
+
"num_stages": 4
|
25
|
+
},
|
26
|
+
"8": {
|
27
|
+
"BLOCK_SIZE_M": 16,
|
28
|
+
"BLOCK_SIZE_N": 128,
|
29
|
+
"BLOCK_SIZE_K": 64,
|
30
|
+
"GROUP_SIZE_M": 32,
|
31
|
+
"num_warps": 4,
|
32
|
+
"num_stages": 4
|
33
|
+
},
|
34
|
+
"16": {
|
35
|
+
"BLOCK_SIZE_M": 16,
|
36
|
+
"BLOCK_SIZE_N": 256,
|
37
|
+
"BLOCK_SIZE_K": 64,
|
38
|
+
"GROUP_SIZE_M": 32,
|
39
|
+
"num_warps": 8,
|
40
|
+
"num_stages": 4
|
41
|
+
},
|
42
|
+
"24": {
|
43
|
+
"BLOCK_SIZE_M": 32,
|
44
|
+
"BLOCK_SIZE_N": 32,
|
45
|
+
"BLOCK_SIZE_K": 64,
|
46
|
+
"GROUP_SIZE_M": 32,
|
47
|
+
"num_warps": 8,
|
48
|
+
"num_stages": 4
|
49
|
+
},
|
50
|
+
"32": {
|
51
|
+
"BLOCK_SIZE_M": 16,
|
52
|
+
"BLOCK_SIZE_N": 32,
|
53
|
+
"BLOCK_SIZE_K": 256,
|
54
|
+
"GROUP_SIZE_M": 32,
|
55
|
+
"num_warps": 4,
|
56
|
+
"num_stages": 4
|
57
|
+
},
|
58
|
+
"48": {
|
59
|
+
"BLOCK_SIZE_M": 16,
|
60
|
+
"BLOCK_SIZE_N": 32,
|
61
|
+
"BLOCK_SIZE_K": 128,
|
62
|
+
"GROUP_SIZE_M": 32,
|
63
|
+
"num_warps": 4,
|
64
|
+
"num_stages": 4
|
65
|
+
},
|
66
|
+
"64": {
|
67
|
+
"BLOCK_SIZE_M": 32,
|
68
|
+
"BLOCK_SIZE_N": 32,
|
69
|
+
"BLOCK_SIZE_K": 256,
|
70
|
+
"GROUP_SIZE_M": 16,
|
71
|
+
"num_warps": 4,
|
72
|
+
"num_stages": 4
|
73
|
+
},
|
74
|
+
"96": {
|
75
|
+
"BLOCK_SIZE_M": 32,
|
76
|
+
"BLOCK_SIZE_N": 32,
|
77
|
+
"BLOCK_SIZE_K": 64,
|
78
|
+
"GROUP_SIZE_M": 64,
|
79
|
+
"num_warps": 4,
|
80
|
+
"num_stages": 4
|
81
|
+
},
|
82
|
+
"128": {
|
83
|
+
"BLOCK_SIZE_M": 16,
|
84
|
+
"BLOCK_SIZE_N": 128,
|
85
|
+
"BLOCK_SIZE_K": 128,
|
86
|
+
"GROUP_SIZE_M": 64,
|
87
|
+
"num_warps": 8,
|
88
|
+
"num_stages": 4
|
89
|
+
},
|
90
|
+
"256": {
|
91
|
+
"BLOCK_SIZE_M": 64,
|
92
|
+
"BLOCK_SIZE_N": 256,
|
93
|
+
"BLOCK_SIZE_K": 64,
|
94
|
+
"GROUP_SIZE_M": 32,
|
95
|
+
"num_warps": 4,
|
96
|
+
"num_stages": 4
|
97
|
+
},
|
98
|
+
"512": {
|
99
|
+
"BLOCK_SIZE_M": 64,
|
100
|
+
"BLOCK_SIZE_N": 256,
|
101
|
+
"BLOCK_SIZE_K": 64,
|
102
|
+
"GROUP_SIZE_M": 64,
|
103
|
+
"num_warps": 8,
|
104
|
+
"num_stages": 4
|
105
|
+
},
|
106
|
+
"1024": {
|
107
|
+
"BLOCK_SIZE_M": 64,
|
108
|
+
"BLOCK_SIZE_N": 256,
|
109
|
+
"BLOCK_SIZE_K": 64,
|
110
|
+
"GROUP_SIZE_M": 32,
|
111
|
+
"num_warps": 8,
|
112
|
+
"num_stages": 4
|
113
|
+
},
|
114
|
+
"1536": {
|
115
|
+
"BLOCK_SIZE_M": 64,
|
116
|
+
"BLOCK_SIZE_N": 256,
|
117
|
+
"BLOCK_SIZE_K": 64,
|
118
|
+
"GROUP_SIZE_M": 64,
|
119
|
+
"num_warps": 8,
|
120
|
+
"num_stages": 4
|
121
|
+
},
|
122
|
+
"2048": {
|
123
|
+
"BLOCK_SIZE_M": 64,
|
124
|
+
"BLOCK_SIZE_N": 256,
|
125
|
+
"BLOCK_SIZE_K": 64,
|
126
|
+
"GROUP_SIZE_M": 32,
|
127
|
+
"num_warps": 8,
|
128
|
+
"num_stages": 4
|
129
|
+
},
|
130
|
+
"3072": {
|
131
|
+
"BLOCK_SIZE_M": 128,
|
132
|
+
"BLOCK_SIZE_N": 128,
|
133
|
+
"BLOCK_SIZE_K": 64,
|
134
|
+
"GROUP_SIZE_M": 16,
|
135
|
+
"num_warps": 8,
|
136
|
+
"num_stages": 4
|
137
|
+
},
|
138
|
+
"4096": {
|
139
|
+
"BLOCK_SIZE_M": 128,
|
140
|
+
"BLOCK_SIZE_N": 128,
|
141
|
+
"BLOCK_SIZE_K": 64,
|
142
|
+
"GROUP_SIZE_M": 16,
|
143
|
+
"num_warps": 8,
|
144
|
+
"num_stages": 4
|
145
|
+
}
|
146
|
+
}
|