Revert "drop ascend scheduler" (#4580)

Reverts #4498
- vLLM version: v0.11.2
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.2

Author: MengqingCao

.github/workflows/_e2e_test.yaml

@@ -91,8 +91,10 @@ jobs:
           pytest -sv tests/e2e/singlecard/test_completion_with_prompt_embeds.py
           pytest -sv tests/e2e/singlecard/test_aclgraph.py
           pytest -sv tests/e2e/singlecard/test_aclgraph_mem.py
+          pytest -sv tests/e2e/singlecard/test_ascend_scheduler.py
           pytest -sv tests/e2e/singlecard/test_bge_model.py
           pytest -sv tests/e2e/singlecard/test_camem.py
+          pytest -sv tests/e2e/singlecard/test_chunked.py
           pytest -sv tests/e2e/singlecard/test_embedding.py
           # pytest -sv tests/e2e/singlecard/test_embedding_aclgraph.py
           pytest -sv tests/e2e/singlecard/test_guided_decoding.py

docs/source/tutorials/DeepSeek-V3.2-Exp.md

@@ -108,7 +108,7 @@ vllm serve vllm-ascend/DeepSeek-V3.2-Exp-W8A8 \
 --trust-remote-code \
 --no-enable-prefix-caching \
 --gpu-memory-utilization 0.92 \
---additional-config '{"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}'
+--additional-config '{"ascend_scheduler_config":{"enabled":true},"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}'
 ```
 
 ### Multi-node Deployment
@@ -160,7 +160,7 @@ vllm serve /root/.cache/Modelers_Park/DeepSeek-V3.2-Exp \
 --trust-remote-code \
 --no-enable-prefix-caching \
 --gpu-memory-utilization 0.9 \
---additional-config '{"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}'
+--additional-config '{"ascend_scheduler_config":{"enabled":true},"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}'
 ```
 
 **Node 1**
@@ -204,7 +204,7 @@ vllm serve /root/.cache/Modelers_Park/DeepSeek-V3.2-Exp \
 --trust-remote-code \
 --no-enable-prefix-caching \
 --gpu-memory-utilization 0.92 \
---additional-config '{"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}'
+--additional-config '{"ascend_scheduler_config":{"enabled":true},"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}'
 ```
 
 ::::
@@ -252,7 +252,7 @@ vllm serve vllm-ascend/DeepSeek-V3.2-Exp-W8A8 \
 --quantization ascend \
 --no-enable-prefix-caching \
 --gpu-memory-utilization 0.9 \
---additional-config '{"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}'
+--additional-config '{"ascend_scheduler_config":{"enabled":true},"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}'
 ```
 
 **Node 1**
@@ -299,7 +299,7 @@ vllm serve vllm-ascend/DeepSeek-V3.2-Exp-W8A8 \
 --quantization ascend \
 --no-enable-prefix-caching \
 --gpu-memory-utilization 0.92 \
---additional-config '{"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}'
+--additional-config '{"ascend_scheduler_config":{"enabled":true},"torchair_graph_config":{"enabled":true,"graph_batch_sizes":[16]}}'
 ```
 
 ::::

docs/source/tutorials/multi_node.md

@@ -137,7 +137,7 @@ vllm serve vllm-ascend/DeepSeek-V3.1-W8A8 \
 --trust-remote-code \
 --no-enable-prefix-caching \
 --gpu-memory-utilization 0.9 \
---additional-config '{"torchair_graph_config":{"enabled":true}}'
+--additional-config '{"ascend_scheduler_config":{"enabled":true},"torchair_graph_config":{"enabled":true}}'
 ```
 
 **Node 1**
@@ -182,7 +182,7 @@ vllm serve vllm-ascend/DeepSeek-V3.1-W8A8 \
 --trust-remote-code \
 --no-enable-prefix-caching \
 --gpu-memory-utilization 0.92 \
---additional-config '{"torchair_graph_config":{"enabled":true}}'
+--additional-config '{"ascend_scheduler_config":{"enabled":true},"torchair_graph_config":{"enabled":true}}'
 ```
 
 The deployment view looks like:

docs/source/tutorials/multi_node_kimi.md

@@ -93,7 +93,7 @@ vllm serve /home/cache/weights/Kimi-K2-Instruct-W8A8 \
 --trust-remote-code \
 --no-enable-prefix-caching \
 --gpu-memory-utilization 0.9 \
---additional-config '{"torchair_graph_config":{"enabled":true}}'
+--additional-config '{"ascend_scheduler_config":{"enabled":true},"torchair_graph_config":{"enabled":true}}'
 ```
 
 **Node 1**
@@ -137,7 +137,7 @@ vllm serve /home/cache/weights/Kimi-K2-Instruct-W8A8 \
 --trust-remote-code \
 --no-enable-prefix-caching \
 --gpu-memory-utilization 0.92 \
---additional-config '{"torchair_graph_config":{"enabled":true}}'
+--additional-config '{"ascend_scheduler_config":{"enabled":true},"torchair_graph_config":{"enabled":true}}'
 ```
 
 The deployment view looks like:

docs/source/tutorials/multi_npu_moge.md

@@ -158,6 +158,11 @@ if __name__ == "__main__":
             'torchair_graph_config': {
             'enabled': True,
             },
+            'ascend_scheduler_config':{
+            'enabled': True,
+            'enable_chunked_prefill' : False,
+            'chunked_prefill_enabled': False
+            },
             })
 
     outputs = llm.generate(prompts, sampling_params)

docs/source/user_guide/configuration/additional_config.md

@@ -27,6 +27,7 @@ The following table lists additional configuration options available in vLLM Asc
 | Name                                | Type | Default | Description                                                                                                                                   |
 |-------------------------------------|------|---------|-----------------------------------------------------------------------------------------------------------------------------------------------|
 | `torchair_graph_config`             | dict | `{}`    | Configuration options for torchair graph mode                                                                                                    |
+| `ascend_scheduler_config`           | dict | `{}`    | Configuration options for ascend scheduler                                                                                                       |
 | `weight_prefetch_config`            | dict | `{}`    | Configuration options for weight prefetch                                                                                                        |
 | `refresh`                           | bool | `false` | Whether to refresh global Ascend configuration content. This is usually used by rlhf or ut/e2e test case.                                      |
 | `expert_map_path`                   | str  | `None`  | When using expert load balancing for an MoE model, an expert map path needs to be passed in.                                                 |
@@ -60,6 +61,18 @@ The details of each configuration option are as follows:
 | `enable_kv_nz`| bool | `False` | Whether to enable KV Cache NZ layout. This option only takes effect on models using MLA (for example, DeepSeek). |
 | `enable_super_kernel` | bool | `False` | Whether to enable super kernel to fuse operators in deepseek moe layers. This option only takes effects on moe models using dynamic w8a8 quantization.|
 
+**ascend_scheduler_config**
+
+| Name | Type | Default | Description |
+| ---- | ---- | ------- | ----------- |
+| `enabled` | bool | `False` | Whether to enable ascend scheduler for V1 engine.|
+| `enable_pd_transfer` | bool | `False` | Whether to enable P-D transfer. When it is enabled, decode is started only when prefill of all requests is done. This option only takes effect on offline inference. |
+| `decode_max_num_seqs` | int | `0` | Whether to change max_num_seqs of decode phase when P-D transfer is enabled. This option only takes effect when enable_pd_transfer is True. |
+| `max_long_partial_prefills` | Union[int, float] | `float('inf')` | The maximum number of prompts longer than long_prefill_token_threshold that will be prefilled concurrently. |
+| `long_prefill_token_threshold` | Union[int, float] | `float('inf')` | a request is considered long if the prompt is longer than this number of tokens. |
+
+ascend_scheduler_config also supports the options from [vllm scheduler config](https://docs.vllm.ai/en/stable/api/vllm/config.html#vllm.config.SchedulerConfig). For example, you can add `enable_chunked_prefill: True` to ascend_scheduler_config as well.
+
 **weight_prefetch_config**
 
 | Name             | Type | Default                                                     | Description                        |
@@ -80,6 +93,12 @@ An example of additional configuration is as follows:
         "graph_batch_sizes_init": False,
         "enable_kv_nz": False
     },
+    "ascend_scheduler_config": {
+        "enabled": True,
+        "enable_chunked_prefill": True,
+        "max_long_partial_prefills": 1,
+        "long_prefill_token_threshold": 4096,
+    },
     "weight_prefetch_config": {
         "enabled": True,
         "prefetch_ratio": {

docs/source/user_guide/feature_guide/graph_mode.md

@@ -45,14 +45,14 @@ import os
 from vllm import LLM
 
 # TorchAirGraph only works without chunked-prefill now
-model = LLM(model="path/to/DeepSeek-R1-0528", additional_config={"torchair_graph_config": {"enabled": True}})
+model = LLM(model="path/to/DeepSeek-R1-0528", additional_config={"torchair_graph_config": {"enabled": True},"ascend_scheduler_config": {"enabled": True}})
 outputs = model.generate("Hello, how are you?")
 ```
 
 Online example:
 
 ```shell
-vllm serve path/to/DeepSeek-R1-0528 --additional-config='{"torchair_graph_config": {"enabled": true}}'
+vllm serve path/to/DeepSeek-R1-0528 --additional-config='{"torchair_graph_config": {"enabled": true},"ascend_scheduler_config": {"enabled": true}}'
 ```
 
 You can find more details about additional configuration [here](../configuration/additional_config.md).

examples/offline_inference_npu_long_seq.py

@@ -42,6 +42,7 @@
         enable_chunked_prefill=False,
         max_num_batched_tokens=2048,
         max_model_len=1024,
+        additional_config={"ascend_scheduler_config": {"enabled": False}},
         max_num_seqs=1,
         block_size=128,
         gpu_memory_utilization=0.9

examples/run_dp_server.sh

@@ -28,4 +28,4 @@ vllm serve Qwen/Qwen1.5-MoE-A2.7B  \
   --gpu-memory-utilization 0.9 \
   --trust-remote-code \
   --enforce-eager \
-  --additional-config '{"torchair_graph_config":{"enabled":false, "use_cached_graph":false}}'
+  --additional-config '{"ascend_scheduler_config":{"enabled":true},"torchair_graph_config":{"enabled":false, "use_cached_graph":false}}'

tests/e2e/310p/test_offline_inference_parallel_310p.py

@@ -24,12 +24,15 @@
 MODELS = [
     "IntervitensInc/pangu-pro-moe-model",
 ]
-# set additional config for torchair graph
+# set additional config for ascend scheduler and torchair graph
 ADDITIONAL_CONFIG = [{
     "additional_config": {
         "torchair_graph_config": {
             "enabled": True
         },
+        "ascend_scheduler_config": {
+            "enabled": True,
+        }
     }
 }]