Skip to content

Commit 0f3b945

Browse files
committed
Add SGLANG_RADIX_FORCE_MISS=1 to all MTP configs
1 parent fb00c84 commit 0f3b945

6 files changed

Lines changed: 12 additions & 0 deletions

benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/8k1k/disagg-low-latency-1p1d-tp4-tp4-mtp.yaml

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -31,6 +31,7 @@ backend:
3131

3232
prefill_environment:
3333
PYTHONUNBUFFERED: "1"
34+
SGLANG_RADIX_FORCE_MISS: "1"
3435
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
3536
SGLANG_DEFAULT_THINKING: "1"
3637
SGLANG_DSV4_REASONING_EFFORT: "max"
@@ -45,6 +46,7 @@ backend:
4546

4647
decode_environment:
4748
PYTHONUNBUFFERED: "1"
49+
SGLANG_RADIX_FORCE_MISS: "1"
4850
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
4951
SGLANG_DEFAULT_THINKING: "1"
5052
SGLANG_DSV4_REASONING_EFFORT: "max"

benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/8k1k/disagg-low-latency-1p6d-dep4-tp4-mtp.yaml

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -31,6 +31,7 @@ backend:
3131

3232
prefill_environment:
3333
PYTHONUNBUFFERED: "1"
34+
SGLANG_RADIX_FORCE_MISS: "1"
3435
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
3536
SGLANG_DEFAULT_THINKING: "1"
3637
SGLANG_DSV4_REASONING_EFFORT: "max"
@@ -51,6 +52,7 @@ backend:
5152

5253
decode_environment:
5354
PYTHONUNBUFFERED: "1"
55+
SGLANG_RADIX_FORCE_MISS: "1"
5456
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
5557
SGLANG_DEFAULT_THINKING: "1"
5658
SGLANG_DSV4_REASONING_EFFORT: "max"

benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/8k1k/disagg-mid-curve-1p1d-dep4-dep16-mtp.yaml

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -33,6 +33,7 @@ backend:
3333

3434
prefill_environment:
3535
PYTHONUNBUFFERED: "1"
36+
SGLANG_RADIX_FORCE_MISS: "1"
3637
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
3738
SGLANG_DEFAULT_THINKING: "1"
3839
SGLANG_DSV4_REASONING_EFFORT: "max"
@@ -53,6 +54,7 @@ backend:
5354

5455
decode_environment:
5556
PYTHONUNBUFFERED: "1"
57+
SGLANG_RADIX_FORCE_MISS: "1"
5658
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
5759
SGLANG_DEFAULT_THINKING: "1"
5860
SGLANG_DSV4_REASONING_EFFORT: "max"

benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/8k1k/disagg-mid-curve-1p1d-dep4-dep8-mtp.yaml

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -33,6 +33,7 @@ backend:
3333

3434
prefill_environment:
3535
PYTHONUNBUFFERED: "1"
36+
SGLANG_RADIX_FORCE_MISS: "1"
3637
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
3738
SGLANG_DEFAULT_THINKING: "1"
3839
SGLANG_DSV4_REASONING_EFFORT: "max"
@@ -53,6 +54,7 @@ backend:
5354

5455
decode_environment:
5556
PYTHONUNBUFFERED: "1"
57+
SGLANG_RADIX_FORCE_MISS: "1"
5658
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
5759
SGLANG_DEFAULT_THINKING: "1"
5860
SGLANG_DSV4_REASONING_EFFORT: "max"

benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/8k1k/disagg-mid-curve-2p1d-dep4-dep8-mtp.yaml

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -33,6 +33,7 @@ backend:
3333

3434
prefill_environment:
3535
PYTHONUNBUFFERED: "1"
36+
SGLANG_RADIX_FORCE_MISS: "1"
3637
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
3738
SGLANG_DEFAULT_THINKING: "1"
3839
SGLANG_DSV4_REASONING_EFFORT: "max"
@@ -53,6 +54,7 @@ backend:
5354

5455
decode_environment:
5556
PYTHONUNBUFFERED: "1"
57+
SGLANG_RADIX_FORCE_MISS: "1"
5658
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
5759
SGLANG_DEFAULT_THINKING: "1"
5860
SGLANG_DSV4_REASONING_EFFORT: "max"

benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/8k1k/disagg-mid-curve-4p1d-dep4-dep8-mtp.yaml

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -33,6 +33,7 @@ backend:
3333

3434
prefill_environment:
3535
PYTHONUNBUFFERED: "1"
36+
SGLANG_RADIX_FORCE_MISS: "1"
3637
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
3738
SGLANG_DEFAULT_THINKING: "1"
3839
SGLANG_DSV4_REASONING_EFFORT: "max"
@@ -53,6 +54,7 @@ backend:
5354

5455
decode_environment:
5556
PYTHONUNBUFFERED: "1"
57+
SGLANG_RADIX_FORCE_MISS: "1"
5658
SGLANG_JIT_DEEPGEMM_FAST_WARMUP: "1"
5759
SGLANG_DEFAULT_THINKING: "1"
5860
SGLANG_DSV4_REASONING_EFFORT: "max"

0 commit comments

Comments
 (0)