[CE]add 21b mooncake yaml (#7033)

* [CE]add 21b cpu cache ,glm mtp,glm for rl config

* [CE]add 21b tp2 yaml

* [CE]add 21b mooncake yaml

* add fastdeploy benchmark,paddletest-155

* [CE] adjust vl wint4 config

* [CE]add glm mtp with updatemodel config

* [CE]fix

* fix

* test

* test

* test

---------

Co-authored-by: xiegegege <>

benchmarks/yaml/GLM45-air-32k-bf16-mtp-updatemodel.yaml

@@ -0,0 +1,10 @@
+max_model_len: 32768
+max_num_seqs: 128
+tensor_parallel_size: 4
+graph_optimization_config:
+  use_cudagraph: True
+  draft_model_use_cudagraph: True
+load_choices: "default_v1"
+dynamic_load_weight: True
+load_strategy: ipc_snapshot
+shutdown_comm_group_if_worker_idle: False

benchmarks/yaml/eb45-21b-a3b-32k-bf16-tp2-mooncake.yaml

@@ -0,0 +1,5 @@
+max_model_len: 131072
+max_num_seqs: 256
+tensor_parallel_size: 2
+kvcache_storage_backend: "mooncake"
+enable_output_caching: True

benchmarks/yaml/eb45-vl-32k-wint4-a800-tp8.yaml

@@ -7,3 +7,4 @@ tensor_parallel_size: 8
 quantization: wint4
 limit_mm_per_prompt: '{"image": 100, "video": 100}'
 reasoning_parser: ernie-45-vl
+max_num_batched_tokens: 4096