已合并
[feature] 示例与文档默认 KV Connector 切换为 MooncakeConnectorV1 #769
[feature] 示例与文档默认 KV Connector 切换为 MooncakeConnectorV1 #769
已合并
yuzechen创建于 19 天前
7 个文件变更+15-15
@@ -510,7 +510,7 @@ motor_engine_union_config字段用于**PD混部场景**,配置同一类 union
510 "enforce-eager": true,510 "enforce-eager": true,
511 "max_model_len": 2048,511 "max_model_len": 2048,
512 "kv_transfer_config": {512 "kv_transfer_config": {
513- "kv_connector": "MooncakeLayerwiseConnector",513+ "kv_connector": "MooncakeConnectorV1",
514 "kv_buffer_device": "npu",514 "kv_buffer_device": "npu",
515 "kv_parallel_size": 1,515 "kv_parallel_size": 1,
516 "kv_port": "30001",516 "kv_port": "30001",
@@ -630,7 +630,7 @@ motor_engine_prefill_config和motor_engine_decode_config字段用于**PD分离
630 "enforce-eager": true,630 "enforce-eager": true,
631 "max_model_len": 2048,631 "max_model_len": 2048,
632 "kv_transfer_config": {632 "kv_transfer_config": {
633- "kv_connector": "MooncakeLayerwiseConnector",633+ "kv_connector": "MooncakeConnectorV1",
634 "kv_buffer_device": "npu",634 "kv_buffer_device": "npu",
635 "kv_role": "kv_producer",635 "kv_role": "kv_producer",
636 "kv_parallel_size": 1,636 "kv_parallel_size": 1,
@@ -705,7 +705,7 @@ motor_engine_prefill_config和motor_engine_decode_config字段用于**PD分离
705 "enforce-eager": true,705 "enforce-eager": true,
706 "max_model_len": 2048,706 "max_model_len": 2048,
707 "kv_transfer_config": {707 "kv_transfer_config": {
708- "kv_connector": "MooncakeLayerwiseConnector",708+ "kv_connector": "MooncakeConnectorV1",
709 "kv_buffer_device": "npu",709 "kv_buffer_device": "npu",
710 "kv_role": "kv_producer",710 "kv_role": "kv_producer",
711 "kv_parallel_size": 1,711 "kv_parallel_size": 1,
@@ -120,7 +120,7 @@ MindIE Motor部署EPD分离只需修改user_config.json配置文件后,通过d
120 "ec_connector_extra_config": {"shared_storage_path": "/mnt/share/patch/ec_cache"}120 "ec_connector_extra_config": {"shared_storage_path": "/mnt/share/patch/ec_cache"}
121 },121 },
122 "kv_transfer_config": {122 "kv_transfer_config": {
123- "kv_connector": "MooncakeLayerwiseConnector",123+ "kv_connector": "MooncakeConnectorV1",
124 "kv_buffer_device": "npu",124 "kv_buffer_device": "npu",
125 "kv_role": "kv_producer",125 "kv_role": "kv_producer",
126 "kv_parallel_size": 1,126 "kv_parallel_size": 1,
@@ -149,7 +149,7 @@ MindIE Motor部署EPD分离只需修改user_config.json配置文件后,通过d
149 "no-enable-prefix-caching": true,149 "no-enable-prefix-caching": true,
150 "allowed-local-media-path": "/mnt/share/patch/media_path/",150 "allowed-local-media-path": "/mnt/share/patch/media_path/",
151 "kv_transfer_config": {151 "kv_transfer_config": {
152- "kv_connector": "MooncakeLayerwiseConnector",152+ "kv_connector": "MooncakeConnectorV1",
153 "kv_buffer_device": "npu",153 "kv_buffer_device": "npu",
154 "kv_role": "kv_consumer",154 "kv_role": "kv_consumer",
155 "kv_parallel_size": 1,155 "kv_parallel_size": 1,
@@ -94,7 +94,7 @@ MindIE Motor开启Tracing能力需修改env.json配置文件和user_config.json
94 "data_parallel_rpc_port": 9000,94 "data_parallel_rpc_port": 9000,
95 "otlp-traces-endpoint": "http://xx.xx.xx.xx:4318/v1/traces",95 "otlp-traces-endpoint": "http://xx.xx.xx.xx:4318/v1/traces",
96 "kv_transfer_config": {96 "kv_transfer_config": {
97- "kv_connector": "MooncakeLayerwiseConnector",97+ "kv_connector": "MooncakeConnectorV1",
98 "kv_buffer_device": "npu",98 "kv_buffer_device": "npu",
99 "kv_role": "kv_producer",99 "kv_role": "kv_producer",
100 "kv_connector_extra_config": {100 "kv_connector_extra_config": {
@@ -116,7 +116,7 @@ MindIE Motor开启Tracing能力需修改env.json配置文件和user_config.json
116 "data_parallel_rpc_port": 9000,116 "data_parallel_rpc_port": 9000,
117 "otlp-traces-endpoint": "http://xx.xx.xx.xx:4318/v1/traces",117 "otlp-traces-endpoint": "http://xx.xx.xx.xx:4318/v1/traces",
118 "kv_transfer_config": {118 "kv_transfer_config": {
119- "kv_connector": "MooncakeLayerwiseConnector",119+ "kv_connector": "MooncakeConnectorV1",
120 "kv_buffer_device": "npu",120 "kv_buffer_device": "npu",
121 "kv_role": "kv_consumer",121 "kv_role": "kv_consumer",
122 "kv_connector_extra_config": {122 "kv_connector_extra_config": {
@@ -119,7 +119,7 @@ chmod -R 755 /mnt/weight
119 "enforce-eager": true,119 "enforce-eager": true,
120 "max_model_len": 2048,120 "max_model_len": 2048,
121 "kv_transfer_config": {121 "kv_transfer_config": {
122- "kv_connector": "MooncakeLayerwiseConnector",122+ "kv_connector": "MooncakeConnectorV1",
123 "kv_buffer_device": "npu",123 "kv_buffer_device": "npu",
124 "kv_role": "kv_producer",124 "kv_role": "kv_producer",
125 "kv_parallel_size": 1,125 "kv_parallel_size": 1,
@@ -144,7 +144,7 @@ chmod -R 755 /mnt/weight
144 "enable_expert_parallel": false,144 "enable_expert_parallel": false,
145 "max_model_len": 2048,145 "max_model_len": 2048,
146 "kv_transfer_config": {146 "kv_transfer_config": {
147- "kv_connector": "MooncakeLayerwiseConnector",147+ "kv_connector": "MooncakeConnectorV1",
148 "kv_buffer_device": "npu",148 "kv_buffer_device": "npu",
149 "kv_role": "kv_consumer",149 "kv_role": "kv_consumer",
150 "kv_parallel_size": 1,150 "kv_parallel_size": 1,
@@ -44,7 +44,7 @@
44 "enforce-eager": true,44 "enforce-eager": true,
45 "max_model_len": 2048,45 "max_model_len": 2048,
46 "kv_transfer_config": {46 "kv_transfer_config": {
47- "kv_connector": "MooncakeLayerwiseConnector",47+ "kv_connector": "MooncakeConnectorV1",
48 "kv_buffer_device": "npu",48 "kv_buffer_device": "npu",
49 "kv_role": "kv_producer",49 "kv_role": "kv_producer",
50 "kv_parallel_size": 1,50 "kv_parallel_size": 1,
@@ -76,7 +76,7 @@
76 "cudagraph_mode": "FULL_DECODE_ONLY"76 "cudagraph_mode": "FULL_DECODE_ONLY"
77 },77 },
78 "kv_transfer_config": {78 "kv_transfer_config": {
79- "kv_connector": "MooncakeLayerwiseConnector",79+ "kv_connector": "MooncakeConnectorV1",
80 "kv_buffer_device": "npu",80 "kv_buffer_device": "npu",
81 "kv_role": "kv_consumer",81 "kv_role": "kv_consumer",
82 "kv_parallel_size": 1,82 "kv_parallel_size": 1,
@@ -29,7 +29,7 @@
29 "enforce-eager": true,29 "enforce-eager": true,
30 "max_model_len": 2048,30 "max_model_len": 2048,
31 "kv_transfer_config": {31 "kv_transfer_config": {
32- "kv_connector": "MooncakeLayerwiseConnector",32+ "kv_connector": "MooncakeConnectorV1",
33 "kv_buffer_device": "npu",33 "kv_buffer_device": "npu",
34 "kv_role": "kv_producer",34 "kv_role": "kv_producer",
35 "kv_parallel_size": 1,35 "kv_parallel_size": 1,
@@ -54,7 +54,7 @@
54 "enable_expert_parallel": false,54 "enable_expert_parallel": false,
55 "max_model_len": 2048,55 "max_model_len": 2048,
56 "kv_transfer_config": {56 "kv_transfer_config": {
57- "kv_connector": "MooncakeLayerwiseConnector",57+ "kv_connector": "MooncakeConnectorV1",
58 "kv_buffer_device": "npu",58 "kv_buffer_device": "npu",
59 "kv_role": "kv_consumer",59 "kv_role": "kv_consumer",
60 "kv_parallel_size": 1,60 "kv_parallel_size": 1,
@@ -34,7 +34,7 @@
34 "enforce-eager": true,34 "enforce-eager": true,
35 "max_model_len": 2048,35 "max_model_len": 2048,
36 "kv_transfer_config": {36 "kv_transfer_config": {
37- "kv_connector": "MooncakeLayerwiseConnector",37+ "kv_connector": "MooncakeConnectorV1",
38 "kv_buffer_device": "npu",38 "kv_buffer_device": "npu",
39 "kv_role": "kv_producer",39 "kv_role": "kv_producer",
40 "kv_parallel_size": 1,40 "kv_parallel_size": 1,
@@ -61,7 +61,7 @@
61 "enable_expert_parallel": false,61 "enable_expert_parallel": false,
62 "max_model_len": 2048,62 "max_model_len": 2048,
63 "kv_transfer_config": {63 "kv_transfer_config": {
64- "kv_connector": "MooncakeLayerwiseConnector",64+ "kv_connector": "MooncakeConnectorV1",
65 "kv_buffer_device": "npu",65 "kv_buffer_device": "npu",
66 "kv_role": "kv_consumer",66 "kv_role": "kv_consumer",
67 "kv_parallel_size": 1,67 "kv_parallel_size": 1,