From 11d77a60df41e3c9fb0ffec9795333699b1f699d Mon Sep 17 00:00:00 2001 From: amote-i <49533125+amote-i@users.noreply.github.com> Date: Sat, 25 Apr 2026 15:37:07 +0800 Subject: [PATCH] [NPU] [DOC] Update supported models and features of npu (#23564) --- .../ascend_npu_support_features.mdx | 224 ++++++++++++++++++ .../ascend-npus/ascend_npu_support_models.mdx | 99 ++++++++ 2 files changed, 323 insertions(+) diff --git a/docs_new/docs/hardware-platforms/ascend-npus/ascend_npu_support_features.mdx b/docs_new/docs/hardware-platforms/ascend-npus/ascend_npu_support_features.mdx index a8ab43b80..dfcbc451b 100644 --- a/docs_new/docs/hardware-platforms/ascend-npus/ascend_npu_support_features.mdx +++ b/docs_new/docs/hardware-platforms/ascend-npus/ascend_npu_support_features.mdx @@ -173,6 +173,58 @@ click [Server Arguments](../../advanced_features/server_arguments). +## SSL/TLS + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
ArgumentDefaultsOptionsServer supported
`--ssl-keyfile``None`Type: strA2, A3
`--ssl-certfile``None`Type: strA2, A3
`--ssl-keyfile-password``None`Type: strA2, A3
`--enable-ssl-refresh``False`bool flag
(set to enable)
A2, A3
`--enable-http2``False`bool flag
(set to enable)
A2, A3
+ ## Quantization and data type @@ -335,6 +387,18 @@ click [Server Arguments](../../advanced_features/server_arguments). + + + + + + + + + + + + @@ -495,6 +559,18 @@ click [Server Arguments](../../advanced_features/server_arguments). + + + + + + + + + + + + @@ -561,6 +637,18 @@ click [Server Arguments](../../advanced_features/server_arguments). + + + + + + + + + + + +
bool flag
(set to enable)
A2, A3
`--disable-priority-preemption``False`bool flag
(set to enable)
A2, A3
`--default-priority-value``None`Type: intA2, A3
`--schedule-low-priority-`
`values-first`
`False`bool flag (set to enable) A2, A3
--stream-response-default-include-usageFalsebool flag (set to enable)A2, A3
--enable-streaming-sessionFalsebool flag (set to enable)A2, A3
--random-seed Nonebool flag (set to enable) A2, A3
--use-rayFalsebool flag (set to enable)A2, A3
--custom-sigquit-handlerNoneOnly for engineA2, A3
@@ -626,6 +714,12 @@ click [Server Arguments](../../advanced_features/server_arguments). bool flag
(set to enable) A2, A3 + + `--enable-mfu-metrics` + `False` + bool flag
(set to enable) + A2, A3 + `--enable-metrics-for-`
`all-schedulers` `False` @@ -644,6 +738,12 @@ click [Server Arguments](../../advanced_features/server_arguments). List[str] A2, A3 + + `--extra-metric-labels` + `None` + Type: JSON/Dict + A2, A3 + `--bucket-time-to-`
`first-token` `None` @@ -826,6 +926,12 @@ click [Server Arguments](../../advanced_features/server_arguments). Type: str A2, A3 + + --file-storage-path + sglang_storage + Type: str + Unused reserved parameter + --enable-cache-report False @@ -1035,6 +1141,24 @@ click [Server Arguments](../../advanced_features/server_arguments). triton,<br/>csgmv,<br/>ascend,<br/>torch_native A2, A3 + + --experts-shared-outer-loras + None + Type: bool + A2, A3 + + + --lora-use-virtual-experts + False + bool flag
(set to enable) + A2, A3 + + + --lora-strict-loading + False + Type: bool + A2, A3 + `--max-lora-chunk-size` `16` @@ -1302,6 +1426,24 @@ click [Server Arguments](../../advanced_features/server_arguments). Type: int Experimental + + `--speculative-ngram-external-corpus-path` + `None` + Type: str + Experimental + + + `--speculative-ngram-external-sam-budget` + `0` + Type: int + Experimental + + + `--speculative-ngram-external-corpus-max-tokens` + `10000000` + Type: int + Experimental + @@ -1612,6 +1754,40 @@ click [Server Arguments](../../advanced_features/server_arguments). +## Diffusion LLM + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
ArgumentDefaultsOptionsServer supported
`--dllm-algorithm``None`Type: strA2, A3
`--dllm-algorithm-config``None`Type: strA2, A3
+ ## Offloading (must be used with `--disable-cuda-graph`) @@ -1821,6 +1997,12 @@ click [Server Arguments](../../advanced_features/server_arguments). + + + + + + @@ -1971,6 +2153,12 @@ click [Server Arguments](../../advanced_features/server_arguments). + + + + + + @@ -2043,12 +2231,36 @@ click [Server Arguments](../../advanced_features/server_arguments). + + + + + + + + + + + + + + + + + + + + + + + +
bool flag
(set to enable)
A2, A3
--enable-dp-attention-local-control-broadcast`False`bool flag
(set to enable)
A2, A3
--enable-dp-lm-head `False`bool flag
(set to enable)
A2, A3
--enforce-shared-experts-fusionFalsebool flag
(set to enable)
A2, A3
--disable-chunked-<br/>prefix-cache Truebool flag
(set to enable)
A2, A3
`--enable-prefill-context-parallel``False`bool flag
(set to enable)
A2, A3
`--prefill-cp-mode``in-seq-split`Type: strA2, A3
`--enable-fused-qk-`
`norm-rope`
`False` bool flag
(set to enable)
Special for GPU
`--enable-precise-embedding-interpolation``False`bool flag
(set to enable)
A2, A3
`--gc-threshold``None`List[int]A2, A3
@@ -2281,6 +2493,18 @@ click [Server Arguments](../../advanced_features/server_arguments). bool flag
(set to enable) A2, A3 + + `--weight-loader-prefetch-checkpoints` + `False` + bool flag
(set to enable) + A2, A3 + + + `--weight-loader-prefetch-num-threads` + `4` + Type: int + A2, A3 + `--remote-instance-weight-`
`loader-seed-instance-ip` `None` diff --git a/docs_new/docs/hardware-platforms/ascend-npus/ascend_npu_support_models.mdx b/docs_new/docs/hardware-platforms/ascend-npus/ascend_npu_support_models.mdx index 728346d74..d7ef03691 100644 --- a/docs_new/docs/hardware-platforms/ascend-npus/ascend_npu_support_models.mdx +++ b/docs_new/docs/hardware-platforms/ascend-npus/ascend_npu_support_models.mdx @@ -49,6 +49,12 @@ You are welcome to enable various models based on your business requirements. ✅ ✅ + + Qwen/Qwen3.5-397B-A17B + Qwen + ✅ + ✅ + Qwen/Qwen3-30B-A3B-Instruct-2507 Qwen @@ -265,6 +271,12 @@ You are welcome to enable various models based on your business requirements. ✅ ✅ + + moonshotai/Kimi-Linear-48B-A3B-Instruct + Kimi Linear (48B-A3B) + ✅ + ✅ + eigen-ai-labs/gpt-oss-120b-bf16 GPTOSS @@ -289,6 +301,12 @@ You are welcome to enable various models based on your business requirements. ✅ ✅ + + FLM/Tele-FLM + Tele FLM (52B-1T) + ✅ + ✅ + bigcode/starcoder2-7b StarCoder2 @@ -301,6 +319,18 @@ You are welcome to enable various models based on your business requirements. ✅ ✅ + + OrionStarAI/Orion-14B-Base + Orion (14B) + ✅ + ✅ + + + EleutherAI/gpt-j-6b + GPT-J (6B) + ✅ + ✅ + @@ -454,6 +484,63 @@ You are welcome to enable various models based on your business requirements. ✅ ✅ + + PaddlePaddle/ERNIE-4.5-VL-28B-A3B-PT + Ernie4.5-VL + ✅ + ✅ + + + Qwen/Qwen3-Omni-30B-A3B-Instruct + Qwen3-Omni + ✅ + ✅ + + + stepfun-ai/Step3-VL-10B + Step3-VL (10B) + ✅ + ✅ + + + + +## Diffusion language models + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
ModelsModel FamilyA2 SupportedA3 Supported
inclusionAI/LLaDA2.0-flashLLaDA2.0 (mini, flash)
JetLM/SDAR-8B-ChatSDAR (JetLM)
JetLM/SDAR-30B-A3B-ChatSDAR (JetLM)
@@ -589,5 +676,17 @@ You are welcome to enable various models based on your business requirements. ✅ ✅ + + Qwen/Qwen3-Reranker-8B + Qwen3-Reranker (decoder-only yes/no) + ✅ + ✅ + + + Qwen/Qwen3-VL-Reranker-2B + Qwen3-VL-Reranker (multimodal yes/no) + ✅ + ✅ +