[Doc] Remove redundant async-scheduling config. (vllm-project#13093)

### What this PR does / why we need it?
This PR removes the redundant `--async-scheduling` configuration option
from various model deployment tutorials and translation files.

### Does this PR introduce _any_ user-facing change?
No. This is a documentation cleanup only.

### How was this patch tested?
Documentation changes only, verified manually.
- vLLM version: v0.25.1
- vLLM main:
vllm-project/vllm@d02df74

Signed-off-by: zhiyu-wa <1959864813@qq.com>

Author: zhiyu-wa

docs/source/_templates/Model-Deployment-Tutorial-Template.zh.md

@@ -275,7 +275,6 @@ lm_eval \
 | --------- | --------- | --------- | --------- | --------- |
 | FlashComm_v1 | 大并发、张量并行(TP)场景 | `export VLLM_ASCEND_ENABLE_FLASHCOMM1=1` | 将传统Allreduce分解为Reduce-Scatter和All-Gather，减少RMSNorm计算维度 | 阈值保护：仅当实际token数超过阈值时生效，避免小并发场景性能倒退|
 | Matmul-ReduceScatter融合 | 大型分布式环境 | 启用FlashComm_v1后自动开启 | 将矩阵乘法与Reduce-Scatter操作融合，实现流水线并行处理 | 同FlashComm_v1，有阈值保护 |
-| 异步调度 | 大规模模型、高并发场景 | `--async-scheduling` | 非阻塞任务调度，提升并发处理能力 | 与FullGraph优化协同使用 |
 
 ## 10 FAQ
 

docs/source/locale/zh_CN/LC_MESSAGES/tutorials/models/DeepSeek-V4-Flash.po

@@ -169,14 +169,12 @@ msgid ""
 "- `--no-enable-prefix-caching` indicates that prefix caching is disabled. To enable it, remove this option.\n"
 "- `--speculative-config` configures the MTP (Multi-Token Prediction) speculative decoding to accelerate inference.\n"
 "- `--compilation-config '{\"cudagraph_mode\":\"FULL_DECODE_ONLY\"}'` enables full ACL graph execution in the decode phase to reduce scheduling latency.\n"
-"- `--async-scheduling` enables asynchronous scheduling to overlap CPU scheduling with NPU computation.\n"
 "- `VLLM_ASCEND_ENABLE_FLASHCOMM1=1` enables the FlashComm communication optimization."
 msgstr ""
 "- `--max-model-len` 指定最大上下文长度，即单个请求的输入和输出 token 之和。请根据实际场景调整。\n"
 "- `--no-enable-prefix-caching` 表示禁用前缀缓存。如需启用，请移除该选项。\n"
 "- `--speculative-config` 配置 MTP（多 token 预测）推测解码以加速推理。\n"
 "- `--compilation-config '{\"cudagraph_mode\":\"FULL_DECODE_ONLY\"}'` 在解码阶段启用全 ACL 图执行，以减少调度延迟。\n"
-"- `--async-scheduling` 启用异步调度，使 CPU 调度与 NPU 计算重叠。\n"
 "- `VLLM_ASCEND_ENABLE_FLASHCOMM1=1` 启用 FlashComm 通信优化。"
 
 msgid ""

docs/source/locale/zh_CN/LC_MESSAGES/tutorials/models/DeepSeek-V4-Pro.po

@@ -182,7 +182,6 @@ msgid ""
 "- `--max-model-len` specifies the maximum context length. Adjust it according to your actual scenario.\n"
 "- `--speculative-config` configures the MTP (Multi-Token Prediction) speculative decoding to accelerate inference.\n"
 "- `--compilation-config '{\"cudagraph_mode\":\"FULL_DECODE_ONLY\"}'` enables full ACL graph execution in the decode phase to reduce scheduling latency.\n"
-"- `--async-scheduling` enables asynchronous scheduling to overlap CPU scheduling with NPU computation.\n"
 "- `VLLM_ASCEND_ENABLE_FLASHCOMM1=1` enables the FlashComm communication optimization."
 msgstr ""
 "- `--data-parallel-start-rank` 指定当前节点的数据并行起始 rank。每个节点必须设置为唯一值（例如，节点0 = 0，节点1 = 1）。\n"
@@ -191,7 +190,6 @@ msgstr ""
 "- `--max-model-len` 指定最大上下文长度。请根据实际场景进行调整。\n"
 "- `--speculative-config` 配置 MTP（多 Token 预测）推测解码以加速推理。\n"
 "- `--compilation-config '{\"cudagraph_mode\":\"FULL_DECODE_ONLY\"}'` 在解码阶段启用完整的 ACL 图执行，以减少调度延迟。\n"
-"- `--async-scheduling` 启用异步调度，使 CPU 调度与 NPU 计算重叠。\n"
 "- `VLLM_ASCEND_ENABLE_FLASHCOMM1=1` 启用 FlashComm 通信优化。"
 
 msgid ""

docs/source/locale/zh_CN/LC_MESSAGES/tutorials/models/Qwen3-VL-235B-A22B-Instruct.po

@@ -610,12 +610,6 @@ msgstr ""
 "| 前缀缓存 | `--enable-prefix-caching` | 改善重复前缀的工作负载。 | 首先验证 HBM 使用情况。对于 "
 "PD，建议从禁用前缀缓存开始。 |"
 
-msgid ""
-"| Asynchronous scheduling | `--async-scheduling` | Can improve high-"
-"concurrency throughput. | Disable and compare for latency-sensitive "
-"workloads. |"
-msgstr "| 异步调度 | `--async-scheduling` | 可提高高并发吞吐量。 | 对于延迟敏感的工作负载，请禁用并比较。 |"
-
 msgid ""
 "| PD disaggregation | `--kv-transfer-config` | Separates prefill and decode "
 "resources. | Ensure producer/consumer DP and TP sizes match the actual "

docs/source/locale/zh_CN/LC_MESSAGES/tutorials/models/Qwen3-VL-30B-A3B-Instruct.po

@@ -442,12 +442,6 @@ msgid ""
 msgstr ""
 "| 前缀缓存 | `--enable-prefix-caching` | 提升重复前缀工作负载的性能。 | 随机提示或唯一媒体可能无法受益。 |"
 
-msgid ""
-"| Asynchronous scheduling | `--async-scheduling` | Can improve high-"
-"concurrency throughput. | Disable and compare for latency-sensitive "
-"workloads. |"
-msgstr "| 异步调度 | `--async-scheduling` | 可提升高并发吞吐量。 | 对于延迟敏感型工作负载，请禁用并进行对比。 |"
-
 msgid ""
 "| Pipeline parallel validation | `--pipeline-parallel-size 2` | Provides "
 "another two-card validation layout. | Use shorter context and lower batch "

docs/source/locale/zh_CN/LC_MESSAGES/tutorials/models/Qwen3.5-27B-Qwen3.6-27B.po

@@ -224,10 +224,10 @@ msgstr "PD（Prefill-Decode）分离通过将 Prefill 和 Decode 运行在专用
 
 msgid ""
 "- **Prefill nodes** focus on high-throughput prompt processing, optimized for compute and communication (e.g., enabling FlashComm for Allreduce acceleration).\n"
-"- **Decode nodes** focus on low-latency token generation, optimized for memory bandwidth (e.g., enabling async-scheduling and full-decode aclgraph)."
+"- **Decode nodes** focus on low-latency token generation, optimized for memory bandwidth (e.g., enabling full-decode aclgraph)."
 msgstr ""
 "- **Prefill节点**专注于高吞吐量的提示处理，针对计算和通信进行了优化（例如，启用FlashComm以加速Allreduce）。\n"
-"- **Decode节点**专注于低延迟的令牌生成，针对内存带宽进行了优化（例如，启用异步调度和全解码aclgraph）。"
+"- **Decode节点**专注于低延迟的令牌生成，针对内存带宽进行了优化（例如，启用全解码aclgraph）。"
 
 msgid ""
 "For `Qwen3.5-27B-w8a8` and `Qwen3.6-27B-w8a8`, a typical **1P1D** "
@@ -299,12 +299,10 @@ msgstr ""
 msgid ""
 "- `VLLM_ASCEND_ENABLE_FLASHCOMM1=1`: enables the Allreduce communication optimization on prefill nodes, which reduces the communication overhead of long-context prefill.\n"
 "- `recompute_scheduler_enable: true`: enables the recomputation scheduler. When the KV Cache of the decode node is insufficient, requests will be sent to the prefill node to recompute the KV Cache. In the PD separation scenario, enable this configuration only on decode nodes.\n"
-"- `--async-scheduling` (on decode nodes): enables asynchronous scheduling, which can reduce TPOT for high-concurrency decode workloads.\n"
 "- `--compilation-config '{\"cudagraph_mode\":\"FULL_DECODE_ONLY\"}'` (on decode nodes): enables the full-decode aclgraph mode, which significantly reduces scheduling latency on the decode side."
 msgstr ""
 "- `VLLM_ASCEND_ENABLE_FLASHCOMM1=1`：在prefill节点上启用Allreduce通信优化，减少长上下文prefill的通信开销。\n"
 "- `recompute_scheduler_enable: true`：启用重计算调度器。当decode节点的KV缓存不足时，请求将被发送到prefill节点以重新计算KV缓存。在PD分离场景中，仅在decode节点上启用此配置。\n"
-"- `--async-scheduling`（在decode节点上）：启用异步调度，可以减少高并发解码工作负载的TPOT。\n"
 "- `--compilation-config '{\"cudagraph_mode\":\"FULL_DECODE_ONLY\"}'`（在decode节点上）：启用全解码aclgraph模式，显著降低解码端的调度延迟。"
 
 msgid "4. Run server for each node:"

docs/source/locale/zh_CN/LC_MESSAGES/tutorials/models/Qwen3.6-35B-A3B.po

@@ -328,13 +328,6 @@ msgid ""
 "shape."
 msgstr "**原因：** 异步调度减少了阻塞开销，但其收益取决于并发度、提示/输出长度和图捕获形状。"
 
-msgid ""
-"**Solution:** Use `--async-scheduling` for high-throughput serving. For low-"
-"latency serving, compare TTFT and TPOT with and without this option."
-msgstr ""
-"**解决方案：** 对于高吞吐量服务，使用 `--async-scheduling`。对于低延迟服务，请比较启用和未启用此选项时的 TTFT 和 "
-"TPOT。"
-
 msgid ""
 "| Scenario | Deployment Mode | Total NPUs | Weight Version | Key "
 "Considerations |"
@@ -405,13 +398,6 @@ msgstr ""
 "| 共享专家重叠 | `--additional-config '{\"multistream_overlap_shared_expert\": "
 "true}'` | 在MoE工作负载中重叠共享专家计算。 | 推荐用于吞吐量场景。 |"
 
-msgid ""
-"| Asynchronous scheduling | `--async-scheduling` | Improves high-concurrency"
-" throughput by using non-blocking scheduling. | Disable it and compare if "
-"the workload is latency-sensitive. |"
-msgstr ""
-"| 异步调度 | `--async-scheduling` | 通过非阻塞调度提高高并发吞吐量。 | 如果工作负载对延迟敏感，请禁用它并进行比较。 |"
-
 msgid ""
 "| Prefix caching | `--enable-prefix-caching` | Improves repeated-prefix "
 "workloads. | Monitor HBM usage for long-context workloads. |"
@@ -495,7 +481,6 @@ msgid ""
 "    - `--quantization ascend` enables Ascend quantization for the W8A8 model. Remove this option when deploying the BF16 model.\n"
 "    - `--compilation-config '{\"cudagraph_mode\":\"FULL_DECODE_ONLY\"}'` enables full decode ACLGraph replay to reduce dispatch overhead.\n"
 "    - `--additional-config` enables Ascend-specific optimizations. `enable_flashcomm1` enables FlashComm1, `multistream_overlap_shared_expert` overlaps shared expert computation, and `enable_cpu_binding` enables Ascend-native CPU binding.\n"
-"    - `--async-scheduling` enables asynchronous scheduling, which can improve high-concurrency throughput."
 msgstr ""
 "- `--data-parallel-size 1`和`--tensor-parallel-size 2`为默认的单节点服务示例设置DP和TP。\n"
 "    - `--enable-expert-parallel`为MoE层启用专家并行。不要在同一个MoE层中混合使用MoE张量并行和专家并行。\n"
@@ -507,7 +492,6 @@ msgstr ""
 "    - `--quantization ascend`为W8A8模型启用Ascend量化。部署BF16模型时请移除该选项。\n"
 "    - `--compilation-config '{\"cudagraph_mode\":\"FULL_DECODE_ONLY\"}'`启用全解码ACLGraph重放以减少调度开销。\n"
 "    - `--additional-config`启用Ascend特定优化。`enable_flashcomm1`启用FlashComm1，`multistream_overlap_shared_expert`重叠共享专家计算，`enable_cpu_binding`启用Ascend原生CPU绑定。\n"
-"    - `--async-scheduling`启用异步调度，可以提高高并发吞吐量。"
 
 msgid ""
 "- `--tensor-parallel-size 2` maps the model across two Atlas inference devices. Adjust it together with `ASCEND_RT_VISIBLE_DEVICES` according to the available devices and memory.\n"

docs/source/tutorials/models/DeepSeek-V4-Flash.md

@@ -174,7 +174,6 @@ Single-node deployment completes both Prefill and Decode within the same node. T
         --block-size 128 \
         --speculative-config '{"num_speculative_tokens": 1,"method": "mtp","enforce_eager": true}' \
         --compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY"}' \
-        --async-scheduling \
         --additional-config '
         {"ascend_compilation_config":{
             "enable_npugraph_ex":true,
@@ -255,7 +254,6 @@ Single-node deployment completes both Prefill and Decode within the same node. T
         --block-size 128 \
         --speculative-config '{"num_speculative_tokens": 1,"method": "mtp","enforce_eager": true}' \
         --compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY"}' \
-        --async-scheduling \
         --additional-config '
         {"ascend_compilation_config":{
             "enable_npugraph_ex":true,
@@ -271,7 +269,6 @@ Key Parameter Descriptions:
 - `--no-enable-prefix-caching` indicates that prefix caching is disabled. To enable it, remove this option.
 - `--speculative-config` configures the MTP (Multi-Token Prediction) speculative decoding to accelerate inference.
 - `--compilation-config '{"cudagraph_mode":"FULL_DECODE_ONLY"}'` enables full ACL graph execution in the decode phase to reduce scheduling latency.
-- `--async-scheduling` enables asynchronous scheduling to overlap CPU scheduling with NPU computation.
 - `VLLM_ASCEND_ENABLE_FLASHCOMM1=1` enables the FlashComm communication optimization.
 
 Common Issues Tip: If you encounter issues, please refer to the [Public FAQ](https://docs.vllm.ai/projects/ascend/en/latest/faqs.html) for troubleshooting.
@@ -541,7 +538,6 @@ Before you start, please:
             --max-model-len 1048576 \
             --max-num-batched-tokens 120 \
             --max-num-seqs 60 \
-            --async-scheduling \
             --block-size 128 \
             --no-disable-hybrid-kv-cache-manager \
             --no-enable-prefix-caching \
@@ -725,7 +721,6 @@ Before you start, please:
             --max-num-seqs 16 \
             --block-size 128 \
             --enforce-eager \
-            --async-scheduling \
             --no-disable-hybrid-kv-cache-manager \
             --enable-prefix-caching \
             --trust-remote-code \
@@ -801,7 +796,6 @@ Before you start, please:
             --max-model-len 135000 \
             --max-num-batched-tokens 60 \
             --max-num-seqs 30 \
-            --async-scheduling \
             --block-size 128 \
             --no-disable-hybrid-kv-cache-manager \
             --no-enable-prefix-caching \

docs/source/tutorials/models/DeepSeek-V4-Pro.md

@@ -188,7 +188,6 @@ The quantized model `DeepSeek-V4-Pro-w4a8-mtp` requires at least 2 Atlas 800 A3
       --tool-call-parser deepseek_v4 \
       --enable-auto-tool-choice \
       --reasoning-parser deepseek_v4 \
-      --async-scheduling \
       --safetensors-load-strategy 'prefetch' \
       --block-size 128 \
       --speculative-config '{
@@ -265,7 +264,6 @@ The quantized model `DeepSeek-V4-Pro-w4a8-mtp` requires at least 2 Atlas 800 A3
       --tool-call-parser deepseek_v4 \
       --enable-auto-tool-choice \
       --reasoning-parser deepseek_v4 \
-      --async-scheduling \
       --safetensors-load-strategy 'prefetch' \
       --block-size 128 \
       --headless \
@@ -336,7 +334,6 @@ The quantized model `DeepSeek-V4-Pro-w4a8-mtp` requires at least 2 Atlas 800 A3
       --port 8900 \
       --host 0.0.0.0 \
       --block-size 128 \
-      --async-scheduling \
       --compilation-config '{"cudagraph_mode":"FULL_DECODE_ONLY"}' \
       --tokenizer-mode deepseek_v4 \
       --tool-call-parser deepseek_v4 \
@@ -395,7 +392,6 @@ The quantized model `DeepSeek-V4-Pro-w4a8-mtp` requires at least 2 Atlas 800 A3
       --port 8900 \
       --host 0.0.0.0 \
       --block-size 128 \
-      --async-scheduling \
       --compilation-config '{"cudagraph_mode":"FULL_DECODE_ONLY"}' \
       --tokenizer-mode deepseek_v4 \
       --tool-call-parser deepseek_v4 \
@@ -419,7 +415,6 @@ Key Parameter Descriptions:
 - `--max-model-len` specifies the maximum context length. Adjust it according to your actual scenario.
 - `--speculative-config` configures the MTP (Multi-Token Prediction) speculative decoding to accelerate inference.
 - `--compilation-config '{"cudagraph_mode":"FULL_DECODE_ONLY"}'` enables full ACL graph execution in the decode phase to reduce scheduling latency.
-- `--async-scheduling` enables asynchronous scheduling to overlap CPU scheduling with NPU computation.
 - `VLLM_ASCEND_ENABLE_FLASHCOMM1=1` enables the FlashComm communication optimization.
 
 Common Issues Tip: If you encounter issues, please refer to the [Public FAQ](https://docs.vllm.ai/projects/ascend/en/latest/faqs.html) for troubleshooting.
@@ -755,7 +750,6 @@ Before you start, please:
             --max-model-len 131072 \
             --max-num-batched-tokens 120 \
             --max-num-seqs 60 \
-            --async-scheduling \
             --block-size 128 \
             --no-enable-prefix-caching \
             --tokenizer-mode deepseek_v4 \
@@ -1061,7 +1055,6 @@ Before you start, please:
             --max-model-len 133072 \
             --max-num-batched-tokens 120 \
             --max-num-seqs 60 \
-            --async-scheduling \
             --block-size 128 \
             --no-disable-hybrid-kv-cache-manager \
             --trust-remote-code \

docs/source/tutorials/models/GLM5.2.md

@@ -315,7 +315,6 @@ If you want to deploy multi-node environment, you need to verify multi-node comm
     --port 7000 \
     --safetensors-load-strategy 'prefetch' \
     --block-size 128 \
-    --async-scheduling \
     --additional-config '{"multistream_overlap_shared_expert": true}' \
     --compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY"}' \
     --speculative-config '{"num_speculative_tokens": 5, "method": "deepseek_mtp", "enforce_eager": true}'
@@ -372,7 +371,6 @@ If you want to deploy multi-node environment, you need to verify multi-node comm
     --port 7000 \
     --safetensors-load-strategy 'prefetch' \
     --block-size 128 \
-    --async-scheduling \
     --additional-config '{"multistream_overlap_shared_expert": true}' \
     --compilation-config '{"cudagraph_mode": "FULL_DECODE_ONLY"}' \
     --speculative-config '{"num_speculative_tokens": 5, "method": "deepseek_mtp", "enforce_eager": true}'
@@ -675,7 +673,6 @@ Before you start, please
             --trust-remote-code \
             --max-num-seqs 32 \
             --gpu-memory-utilization 0.92 \
-            --async-scheduling \
             --quantization ascend \
             --enable-auto-tool-choice \
             --tool-call-parser glm47 \
@@ -744,7 +741,6 @@ Before you start, please
             --trust-remote-code \
             --max-num-seqs 32 \
             --gpu-memory-utilization 0.92 \
-            --async-scheduling \
             --quantization ascend \
             --enable-auto-tool-choice \
             --tool-call-parser glm47 \
@@ -900,7 +896,6 @@ vllm serve /root/.cache/modelscope/hub/models/vllm-ascend/GLM-5.2-w4a8c8 \
     --seed 1024 \
     --enable-chunked-prefill \
     --served-model-name glm-5 \
-    --async-scheduling \
     --max-model-len 256000 \
     --max-num-batched-tokens 8192 \
     --trust-remote-code \
@@ -1004,7 +999,6 @@ vllm serve /root/.cache/modelscope/hub/models/vllm-ascend/GLM-5.2-w4a8c8 \
     --enable-prefix-caching \
     --seed 1024 \
     --served-model-name glm-5 \
-    --async-scheduling \
     --max-model-len 256000 \
     --max-num-batched-tokens 256 \
     --trust-remote-code \