Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
Menu
Open sidebar
OpenDAS
aiter
Commits
1893a1e0
Commit
1893a1e0
authored
Apr 15, 2026
by
zhangxw
Browse files
[TRITON] Add missing config files for triton moe.
parent
85762c1a
Pipeline
#3506
failed with stages
in 0 seconds
Changes
4
Pipelines
1
Hide whitespace changes
Inline
Side-by-side
Showing
4 changed files
with
319 additions
and
0 deletions
+319
-0
aiter/ops/triton/configs/sage_attention/_attn_fwd-device=gfx936
...ops/triton/configs/sage_attention/_attn_fwd-device=gfx936
+0
-0
aiter/ops/triton/configs/sage_attention/_attn_fwd-device=gfx936:cu_80-dtype=f16_f16_f16_f32_f32_f16_f32.json
...evice=gfx936:cu_80-dtype=f16_f16_f16_f32_f32_f16_f32.json
+130
-0
aiter/ops/triton/configs/sage_attention/quant_per_block_int8_kernel-device=gfx936
.../sage_attention/quant_per_block_int8_kernel-device=gfx936
+0
-0
aiter/ops/triton/configs/sage_attention/quant_per_block_int8_kernel-device=gfx936:cu_80-dtype=f16_i8_f32.json
...ock_int8_kernel-device=gfx936:cu_80-dtype=f16_i8_f32.json
+189
-0
No files found.
aiter/ops/triton/configs/sage_attention/_attn_fwd-device=gfx936
deleted
100644 → 0
View file @
85762c1a
aiter/ops/triton/configs/sage_attention/_attn_fwd-device=gfx936:cu_80-dtype=f16_f16_f16_f32_f32_f16_f32.json
0 → 100644
View file @
1893a1e0
{
"key"
:
[
"qo_len"
,
"kv_len"
,
"H"
,
"num_kv_groups"
],
"config"
:
{
"(14040, 512, 40, 1)"
:
{
"STAGE"
:
1
,
"waves_per_eu"
:
2
,
"matrix_instr_nonkdim"
:
16
,
"sched_latency"
:
"none"
,
"kpack"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
,
"BLOCK_M"
:
64
,
"BLOCK_N"
:
64
},
"(4352, 4352, 24, 1)"
:
{
"STAGE"
:
1
,
"waves_per_eu"
:
2
,
"matrix_instr_nonkdim"
:
16
,
"sched_latency"
:
"mmac5-ds10"
,
"kpack"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
,
"BLOCK_M"
:
128
,
"BLOCK_N"
:
64
},
"(1280, 1280, 24, 1)"
:
{
"STAGE"
:
1
,
"waves_per_eu"
:
1
,
"matrix_instr_nonkdim"
:
16
,
"sched_latency"
:
"mmac5-ds10"
,
"kpack"
:
2
,
"num_warps"
:
8
,
"num_ctas"
:
1
,
"num_stages"
:
1
,
"BLOCK_M"
:
128
,
"BLOCK_N"
:
64
},
"(57600, 57600, 1, 1)"
:
{
"STAGE"
:
1
,
"waves_per_eu"
:
2
,
"matrix_instr_nonkdim"
:
16
,
"sched_latency"
:
"mmac5-ds10"
,
"kpack"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
,
"BLOCK_M"
:
128
,
"BLOCK_N"
:
64
},
"(14080, 14080, 1, 1)"
:
{
"STAGE"
:
1
,
"waves_per_eu"
:
2
,
"matrix_instr_nonkdim"
:
16
,
"sched_latency"
:
"none"
,
"kpack"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
,
"BLOCK_M"
:
64
,
"BLOCK_N"
:
64
},
"(19440, 19440, 1, 1)"
:
{
"STAGE"
:
1
,
"waves_per_eu"
:
2
,
"matrix_instr_nonkdim"
:
16
,
"sched_latency"
:
"none"
,
"kpack"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
,
"BLOCK_M"
:
128
,
"BLOCK_N"
:
64
},
"(2430, 512, 40, 1)"
:
{
"STAGE"
:
1
,
"waves_per_eu"
:
2
,
"matrix_instr_nonkdim"
:
16
,
"sched_latency"
:
"none"
,
"kpack"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
,
"BLOCK_M"
:
64
,
"BLOCK_N"
:
64
},
"(14040, 14040, 40, 1)"
:
{
"STAGE"
:
1
,
"waves_per_eu"
:
2
,
"matrix_instr_nonkdim"
:
16
,
"sched_latency"
:
"mmac5-ds10"
,
"kpack"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
,
"BLOCK_M"
:
128
,
"BLOCK_N"
:
64
},
"(3520, 512, 24, 1)"
:
{
"STAGE"
:
1
,
"waves_per_eu"
:
1
,
"matrix_instr_nonkdim"
:
16
,
"sched_latency"
:
"none"
,
"kpack"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
,
"BLOCK_M"
:
64
,
"BLOCK_N"
:
64
},
"(7200, 512, 40, 1)"
:
{
"STAGE"
:
1
,
"waves_per_eu"
:
2
,
"matrix_instr_nonkdim"
:
16
,
"sched_latency"
:
"none"
,
"kpack"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
,
"BLOCK_M"
:
64
,
"BLOCK_N"
:
64
}
}
}
\ No newline at end of file
aiter/ops/triton/configs/sage_attention/quant_per_block_int8_kernel-device=gfx936
deleted
100644 → 0
View file @
85762c1a
aiter/ops/triton/configs/sage_attention/quant_per_block_int8_kernel-device=gfx936:cu_80-dtype=f16_i8_f32.json
0 → 100644
View file @
1893a1e0
{
"key"
:
[
"L"
,
"C"
,
"BLK"
],
"config"
:
{
"(14040, 128, 128)"
:
{
"waves_per_eu"
:
1
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
2
},
"(512, 128, 64)"
:
{
"waves_per_eu"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
2
},
"(512, 128, 128)"
:
{
"waves_per_eu"
:
2
,
"num_warps"
:
8
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(4352, 128, 32)"
:
{
"waves_per_eu"
:
16
,
"num_warps"
:
1
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(4352, 128, 64)"
:
{
"waves_per_eu"
:
16
,
"num_warps"
:
2
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(1280, 128, 64)"
:
{
"waves_per_eu"
:
16
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
2
},
"(1280, 128, 32)"
:
{
"waves_per_eu"
:
8
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
3
},
"(1280, 128, 128)"
:
{
"waves_per_eu"
:
8
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
2
},
"(57600, 128, 128)"
:
{
"waves_per_eu"
:
4
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
3
},
"(57600, 128, 32)"
:
{
"waves_per_eu"
:
8
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
3
},
"(57600, 128, 64)"
:
{
"waves_per_eu"
:
16
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(14080, 128, 64)"
:
{
"waves_per_eu"
:
8
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(14080, 128, 32)"
:
{
"waves_per_eu"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(14080, 128, 128)"
:
{
"waves_per_eu"
:
8
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(14040, 128, 64)"
:
{
"waves_per_eu"
:
1
,
"num_warps"
:
1
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(512, 128, 32)"
:
{
"waves_per_eu"
:
4
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
2
},
"(19440, 128, 128)"
:
{
"waves_per_eu"
:
2
,
"num_warps"
:
8
,
"num_ctas"
:
1
,
"num_stages"
:
2
},
"(2430, 128, 32)"
:
{
"waves_per_eu"
:
8
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
2
},
"(4352, 128, 128)"
:
{
"waves_per_eu"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
3
},
"(19440, 128, 32)"
:
{
"waves_per_eu"
:
16
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(19440, 128, 64)"
:
{
"waves_per_eu"
:
4
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(14040, 128, 32)"
:
{
"waves_per_eu"
:
1
,
"num_warps"
:
1
,
"num_ctas"
:
1
,
"num_stages"
:
3
},
"(2430, 128, 64)"
:
{
"waves_per_eu"
:
16
,
"num_warps"
:
8
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(3520, 128, 32)"
:
{
"waves_per_eu"
:
4
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
2
},
"(7200, 128, 32)"
:
{
"waves_per_eu"
:
4
,
"num_warps"
:
1
,
"num_ctas"
:
1
,
"num_stages"
:
3
},
"(2430, 128, 128)"
:
{
"waves_per_eu"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(3520, 128, 64)"
:
{
"waves_per_eu"
:
1
,
"num_warps"
:
2
,
"num_ctas"
:
1
,
"num_stages"
:
1
},
"(3520, 128, 128)"
:
{
"waves_per_eu"
:
16
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
2
},
"(7200, 128, 64)"
:
{
"waves_per_eu"
:
16
,
"num_warps"
:
1
,
"num_ctas"
:
1
,
"num_stages"
:
3
},
"(7200, 128, 128)"
:
{
"waves_per_eu"
:
2
,
"num_warps"
:
4
,
"num_ctas"
:
1
,
"num_stages"
:
3
}
}
}
\ No newline at end of file
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