请问现在有支持eagle3投机解码的镜像吗,
minimax w8a8 两台c550大概能支撑多少并发呢,您这边知道吗
[Thu Jun 25 09:43:35 2026] MXCD.BCA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.BDA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.B2B00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.BAA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.B3A00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.BBA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.BDA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.B2B00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.B5C00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.B4D00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.B3A00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.BBA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.B5C00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.B4D00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.BCA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:35 2026] MXCD.BAA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.B4D00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.B2B00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.BBA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.B5C00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.BDA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.B2B00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.BBA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.B5C00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.BDA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.BAA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.BAA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.BCA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.BCA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.B4D00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.B3A00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:45 2026] MXCD.B3A00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.B4D00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.BDA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.BBA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.B3A00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.B5C00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.B2B00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.BCA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.BBA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.B5C00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.B3A00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.BDA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.B2B00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.B4D00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.BCA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.BAA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:43:55 2026] MXCD.BAA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.B3A00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.BCA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.BBA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.BAA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.B5C00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.B4D00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.BDA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.BCA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.BAA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.BBA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.BDA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.B4D00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.B2B00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.B2B00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.B3A00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:05 2026] MXCD.B5C00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.B4D00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.BDA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.B2B00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.BAA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.B2B00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.BDA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.B4D00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.BAA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.BCA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.B3A00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.B5C00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.B5C00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.BBA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.BBA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.BCA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:16 2026] MXCD.B3A00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.BCA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.B2B00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.BAA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.B4D00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.B3A00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.B5C00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.B3A00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.B2B00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.B4D00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.B5C00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.BAA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.BCA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.BDA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.BBA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.BDA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:26 2026] MXCD.BBA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.BBA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.B3A00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.B5C00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.B2B00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.B4D00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.BDA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.BCA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.BAA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.B5C00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.B2B00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.B3A00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.B4D00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.BCA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.BDA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.BBA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:44:36 2026] MXCD.BAA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.B2B00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.B2B00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.BAA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.BAA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.B3A00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.B3A00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.B5C00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.B4D00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.B5C00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.B4D00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.BCA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.BCA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.BBA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.BBA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.BDA00.D0.RINGBUF.ERROR wait_ret failed, -110
[Thu Jun 25 09:48:15 2026] MXCD.BDA00.D0.RINGBUF.ERROR type 0x0 create ringbuf failed, -110
[Thu Jun 25 10:14:20 2026] perf: interrupt took too long (2522 > 2500), lowering kernel.perf_event_max_sample_rate to 79250
[Thu Jun 25 14:28:32 2026] perf: interrupt took too long (3154 > 3152), lowering kernel.perf_event_max_sample_rate to 63250
[Fri Jun 26 00:44:23 2026] perf: interrupt took too long (3959 > 3942), lowering kernel.perf_event_max_sample_rate to 50500
一、软硬件信息
1.服务器厂家:
2.沐曦GPU型号:C550
3.操作系统内核版本:5.15.0-119-generic #129-Ubuntu
4.是否开启CPU虚拟化:是
5.mx-smi回显:=================== MetaX System Management Interface Log ===================
Timestamp : Fri Jun 26 10:28:46 2026
Attached GPUs : 8
+---------------------------------------------------------------------------------+
| MX-SMI 2.3.1 Kernel Mode Driver Version: 3.8.1 |
| MACA Version: 3.7.1.5 BIOS Version: 1.29.1.0 |
|------------------+-----------------+---------------------+----------------------|
| Board Name | GPU Persist-M | Bus-id | GPU-Util sGPU-M |
| Pwr:Usage/Cap | Temp Perf | Memory-Usage | GPU-State |
|==================+=================+=====================+======================|
| 0 MetaX C550 | 0 Off | 0000:2b:00.0 | 0% Disabled |
| 56W / 450W | 35C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 1 MetaX C550 | 1 Off | 0000:3a:00.0 | 0% Disabled |
| 56W / 450W | 37C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 2 MetaX C550 | 2 Off | 0000:4d:00.0 | 0% Disabled |
| 55W / 450W | 37C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 3 MetaX C550 | 3 Off | 0000:5c:00.0 | 0% Disabled |
| 54W / 450W | 34C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 4 MetaX C550 | 4 Off | 0000:aa:00.0 | 0% Disabled |
| 55W / 450W | 34C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 5 MetaX C550 | 5 Off | 0000:ba:00.0 | 0% Disabled |
| 53W / 450W | 36C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 6 MetaX C550 | 6 Off | 0000:ca:00.0 | 0% Disabled |
| 55W / 450W | 36C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 7 MetaX C550 | 7 Off | 0000:da:00.0 | 0% Disabled |
| 55W / 450W | 34C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
+---------------------------------------------------------------------------------+
| Process: |
| GPU PID Process Name GPU Memory |
| Usage(MiB) |
|=================================================================================|
| no process found |
+---------------------------------------------------------------------------------+
6.docker info回显:root@metax-host-043:/model# docker info
bash: docker: command not found
root@metax-host-043:/model# exit
exit
root@metax-host-043:/data4/model# docker info
Client:
Version: 29.1.3
Context: default
Debug Mode: false
Plugins:
trust: Manage trust on Docker images (Docker Inc.)
Version: 29.1.3
Path: /usr/libexec/docker/cli-plugins/docker-trust
Server:
Containers: 3
Running: 1
Paused: 0
Stopped: 2
Images: 4
Server Version: 29.1.3
Storage Driver: overlayfs
driver-type: io.containerd.snapshotter.v1
Logging Driver: json-file
Cgroup Driver: systemd
Cgroup Version: 2
Plugins:
Volume: local
Network: bridge host ipvlan macvlan null overlay
Log: awslogs fluentd gcplogs gelf journald json-file local splunk syslog
CDI spec directories:
/etc/cdi
/var/run/cdi
Swarm: inactive
Runtimes: io.containerd.runc.v2 runc
Default Runtime: runc
Init Binary: docker-init
containerd version:
runc version:
init version:
Security Options:
apparmor
seccomp
Profile: builtin
cgroupns
Kernel Version: 5.15.0-119-generic
Operating System: Ubuntu 22.04.5 LTS
OSType: linux
Architecture: x86_64
CPUs: 160
Total Memory: 1008GiB
Name: metax-host-043
ID: 45020429-1961-4531-9c5e-ac9172ee32ec
Docker Root Dir: /data4/docker
Debug Mode: false
Experimental: false
Insecure Registries:
::1/128
127.0.0.0/8
Live Restore Enabled: false
Firewall Backend: iptables
root@metax-host-043:/data4/model#
7.镜像版本: cr.metax-tech.com/public-init/minimax_m3:sglang_0.5.12_maca.ai3.7.1.7-torch2.8-py310-ubuntu22.04-amd64_v0
8.启动容器命令:docker run -dit --name sglang-m3 \
--ipc=host \
--privileged=true \
--device=/dev/dri \
--device=/dev/mxcd \
--device=/dev/infiniband \
--device=/dev/mem \
--group-add video \
--network=host \
--shm-size 100gb \
--ulimit memlock=-1 \
--security-opt seccomp=unconfined \
--security-opt apparmor=unconfined \
-v /data4/model/:/model \
-v /data4/logs:/workspace/logs \
-v /usr/local/:/usr/local/ \
-e VLLM_ENGINE_READY_TIMEOUT_S=3600 \
cr.metax-tech.com/public-init/minimax_m3:sglang_0.5.12_maca.ai3.7.1.7-torch2.8-py310-ubuntu22.04-amd64_v0 \
/bin/bash
9.容器内执行命令:两台各执行一遍,ulimit -n 65536
export GLOO_SOCKET_IFNAME=inbond1
export MCCL_SOCKET_IFNAME=inbond1
export MACA_PATH=/opt/maca
export LD_LIBRARY_PATH=/opt/maca/lib:/opt/maca/ompi/lib:/opt/maca/ucx/lib
export TRITON_ENABLE_MACA_OPT_MOVE_DOT_OPERANDS_OUT_LOOP=1
export TRITON_ENABLE_MACA_CHAIN_DOT_OPT=1
export TRITON_DISABLE_MACA_OPT_MMA_PREFETCH=1
export TRITON_ENABLE_MACA_COMPILER_INT8_OPT=True
export MCCL_IB_HCA=roce0,roce1,roce2,roce3
export MCCL_CROSS_NIC=1
export FORCE_ACTIVE_WAIT=2
export MACA_SMALL_PAGESIZE_ENABLE=1
export PYTORCH_ENABLE_PG_HIGH_PRIORITY_STREAM=1
export MACA_VISIBLE_DEVICE=0,1,2,3,6,5,4,7
export TRITON_ENABLE_MAC_A_OPT_MOVE_DOT_OPERANDS_OUT_LOOP=1
export TRITON_ENABLE_MAC_A_CHAIN_DOT_OPT=1
export MACA_DIRECT_DISPATCH=1
export MACA_QUEUE_SCHEDULE_POLICY=1
export TORCH_CUDA_ARCH_LIST="8.0 8.6+PTX"
export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True
nohup sglang serve --model-path /model/MiniMax-M3-W8A8-offical --served-model-name minimax-m3 --tp 8 --enable-dp-attention --chunked-prefill-size 32768 --host 0.0.0.0 --port 18090 --enable-metrics --dist-init-addr 192.168.2.43:5000 --pipeline-parallel-size 2 --context-length 32768 --mem-fraction-static 0.9 --nnodes 2 --node-rank $(uname -n | awk -F'-' '{print substr($NF, length($NF), 1) - 3}') --enable-cache-report --trust-remote-code --attention-backend flashinfer --quantization compressed-tensors --reasoning-parser auto --tool-call-parser auto > /workspace/logs/sglang.log 2>&1 &
tail -fn 100 /workspace/logs/sglang.log
二、问题现象
进行500并发压测,进行到一半会报错,如下图,
我想问一下我是两台c550 sglang mimimax-m3, tp ep dp pp应该如何配置能到最大吞吐呢,有最佳实践吗
请问有量化的镜像吗,我想把bf16量化成w8a8,我是c550机器
麻烦问一下minimax m3现在可以部署吗,需要新镜像吗,我这边是4台c550,另外除了bf16,有量化版本吗,
麻烦问一下,哪里有deepseekv4pro 的部署教程呢,还有就是可以启动bf16版本的吗,我这边有4台c550,
看起来报错一样跟docker里
一、软硬件信息
1.服务器厂家:
2.沐曦GPU型号:MetaX C550
3.操作系统内核版本:Linux metax-host-013 5.15.0-119-generic #129-Ubuntu SMP Fri Aug 2 19:25:20 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux
4.是否开启CPU虚拟化:Virtualization features:
Virtualization: VT-x
5.mx-smi回显:root@metax-host-013:/home/metax/models# mx-smi
mx-smi version: 2.3.1
=================== MetaX System Management Interface Log ===================
Timestamp : Wed Jun 3 08:52:42 2026
Attached GPUs : 8
+---------------------------------------------------------------------------------+
| MX-SMI 2.3.1 Kernel Mode Driver Version: 3.3.12 |
| MACA Version: 3.7.2.0 BIOS Version: 1.29.1.0 |
|------------------+-----------------+---------------------+----------------------|
| Board Name | GPU Persist-M | Bus-id | GPU-Util sGPU-M |
| Pwr:Usage/Cap | Temp Perf | Memory-Usage | GPU-State |
|==================+=================+=====================+======================|
| 0 MetaX C550 | 0 Off | 0000:2b:00.0 | 0% Disabled |
| 54W / 450W | 33C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 1 MetaX C550 | 1 Off | 0000:3a:00.0 | 0% Disabled |
| 53W / 450W | 34C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 2 MetaX C550 | 2 Off | 0000:4d:00.0 | 0% Disabled |
| 57W / 450W | 36C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 3 MetaX C550 | 3 Off | 0000:5c:00.0 | 0% Disabled |
| 53W / 450W | 33C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 4 MetaX C550 | 4 Off | 0000:aa:00.0 | 0% Disabled |
| 53W / 450W | 33C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 5 MetaX C550 | 5 Off | 0000:ba:00.0 | 0% Disabled |
| 55W / 450W | 35C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 6 MetaX C550 | 6 Off | 0000:ca:00.0 | 0% Disabled |
| 52W / 450W | 34C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 7 MetaX C550 | 7 Off | 0000:da:00.0 | 0% Disabled |
| 54W / 450W | 34C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
+---------------------------------------------------------------------------------+
| Process: |
| GPU PID Process Name GPU Memory |
| Usage(MiB) |
|=================================================================================|
| no process found |
+---------------------------------------------------------------------------------+
End of Log
6.docker info回显:root@metax-host-013:/home/metax/models# docker info
Client:
Version: 29.1.3
Context: default
Debug Mode: false
Plugins:
trust: Manage trust on Docker images (Docker Inc.)
Version: 29.1.3
Path: /usr/libexec/docker/cli-plugins/docker-trust
Server:
Containers: 1
Running: 1
Paused: 0
Stopped: 0
Images: 1
Server Version: 29.1.3
Storage Driver: overlayfs
driver-type: io.containerd.snapshotter.v1
Logging Driver: json-file
Cgroup Driver: systemd
Cgroup Version: 2
Plugins:
Volume: local
Network: bridge host ipvlan macvlan null overlay
Log: awslogs fluentd gcplogs gelf journald json-file local splunk syslog
CDI spec directories:
/etc/cdi
/var/run/cdi
Swarm: inactive
Runtimes: io.containerd.runc.v2 runc
Default Runtime: runc
Init Binary: docker-init
containerd version:
runc version:
init version:
Security Options:
apparmor
seccomp
Profile: builtin
cgroupns
Kernel Version: 5.15.0-119-generic
Operating System: Ubuntu 22.04.5 LTS
OSType: linux
Architecture: x86_64
CPUs: 160
Total Memory: 1008GiB
Name: metax-host-013
ID: d1337572-eb0e-4bef-b8e6-e9c1e294c26e
Docker Root Dir: /home/metax/docker
Debug Mode: false
Experimental: false
Insecure Registries:
::1/128
127.0.0.0/8
Registry Mirrors:
docker.m.daocloud.io/
huecker.io/
dockerhub.timeweb.cloud/
Live Restore Enabled: false
Firewall Backend: iptables
7.镜像版本:mx-devops-acr-cn-shanghai.cr.volces.com/pub-registry1/ai-opentest/release/maca/vllm-metax:0.21.0-maca.ai3.7.1.2-dsv4-torch2.8-py310-ubuntu22.04-amd64
8.启动容器命令:
docker run -itd \
--privileged \
--network=host \
--ipc=host \
--security-opt seccomp=unconfined \
--security-opt apparmor=unconfined \
--ulimit memlock=-1 \
-v /pde_ai:/pde_ai \
--name deepseek-v4-flash \
-v /home/metax/:/home/metax/ \
mx-devops-acr-cn-shanghai.cr.volces.com/pub-registry1/ai-opentest/release/maca/vllm-metax:0.21.0-maca.ai3.7.1.2-dsv4-torch2.8-py310-ubuntu22.04-amd64 \
bash
9.容器内执行命令:
nohup env CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 vllm serve /home/metax/models/Qwen3.6-27B --trust-remote-code --gpu-memory-utilization 0.9 --served-model-name qwen3.6-27b --api-key abcdaasdas--enable-auto-tool-choice --tool-call-parser qwen3_coder --reasoning-parser qwen3 --tensor-parallel-size 8 --max-num-batched-tokens 8192 --default-chat-template-kwargs '{"enable_thinking": false}' --enable-prompt-tokens-details --enable-force-include-usage --enable-prefix-caching> /home/metax/logs/vllm_log.log 2>&1 &
二、问题现象
vllm持续重试不启动
我这边会一直起不来,持续重试,不知道是什么情况呢,
c500可以直接用vllm镜像启动bf16的模型吗,我从modelscope上下的原始模型,可以直接启动吗,qwen3.6-27b
一、软硬件信息
1.服务器厂家:
2.沐曦GPU型号:MetaX C550
3.操作系统内核版本:Linux metax-host-013 5.15.0-119-generic #129-Ubuntu SMP Fri Aug 2 19:25:20 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux
4.是否开启CPU虚拟化:Virtualization features:
Virtualization: VT-x
5.mx-smi回显:root@metax-host-013:/home/metax/models# mx-smi
mx-smi version: 2.3.1
=================== MetaX System Management Interface Log ===================
Timestamp : Wed Jun 3 08:52:42 2026
Attached GPUs : 8
+---------------------------------------------------------------------------------+
| MX-SMI 2.3.1 Kernel Mode Driver Version: 3.3.12 |
| MACA Version: 3.7.2.0 BIOS Version: 1.29.1.0 |
|------------------+-----------------+---------------------+----------------------|
| Board Name | GPU Persist-M | Bus-id | GPU-Util sGPU-M |
| Pwr:Usage/Cap | Temp Perf | Memory-Usage | GPU-State |
|==================+=================+=====================+======================|
| 0 MetaX C550 | 0 Off | 0000:2b:00.0 | 0% Disabled |
| 54W / 450W | 33C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 1 MetaX C550 | 1 Off | 0000:3a:00.0 | 0% Disabled |
| 53W / 450W | 34C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 2 MetaX C550 | 2 Off | 0000:4d:00.0 | 0% Disabled |
| 57W / 450W | 36C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 3 MetaX C550 | 3 Off | 0000:5c:00.0 | 0% Disabled |
| 53W / 450W | 33C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 4 MetaX C550 | 4 Off | 0000:aa:00.0 | 0% Disabled |
| 53W / 450W | 33C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 5 MetaX C550 | 5 Off | 0000:ba:00.0 | 0% Disabled |
| 55W / 450W | 35C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 6 MetaX C550 | 6 Off | 0000:ca:00.0 | 0% Disabled |
| 52W / 450W | 34C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
| 7 MetaX C550 | 7 Off | 0000:da:00.0 | 0% Disabled |
| 54W / 450W | 34C P0 | 858/65536 MiB | Available |
+------------------+-----------------+---------------------+----------------------+
+---------------------------------------------------------------------------------+
| Process: |
| GPU PID Process Name GPU Memory |
| Usage(MiB) |
|=================================================================================|
| no process found |
+---------------------------------------------------------------------------------+
End of Log
6.docker info回显:root@metax-host-013:/home/metax/models# docker info
Client:
Version: 29.1.3
Context: default
Debug Mode: false
Plugins:
trust: Manage trust on Docker images (Docker Inc.)
Version: 29.1.3
Path: /usr/libexec/docker/cli-plugins/docker-trust
Server:
Containers: 1
Running: 1
Paused: 0
Stopped: 0
Images: 1
Server Version: 29.1.3
Storage Driver: overlayfs
driver-type: io.containerd.snapshotter.v1
Logging Driver: json-file
Cgroup Driver: systemd
Cgroup Version: 2
Plugins:
Volume: local
Network: bridge host ipvlan macvlan null overlay
Log: awslogs fluentd gcplogs gelf journald json-file local splunk syslog
CDI spec directories:
/etc/cdi
/var/run/cdi
Swarm: inactive
Runtimes: io.containerd.runc.v2 runc
Default Runtime: runc
Init Binary: docker-init
containerd version:
runc version:
init version:
Security Options:
apparmor
seccomp
Profile: builtin
cgroupns
Kernel Version: 5.15.0-119-generic
Operating System: Ubuntu 22.04.5 LTS
OSType: linux
Architecture: x86_64
CPUs: 160
Total Memory: 1008GiB
Name: metax-host-013
ID: d1337572-eb0e-4bef-b8e6-e9c1e294c26e
Docker Root Dir: /home/metax/docker
Debug Mode: false
Experimental: false
Insecure Registries:
::1/128
127.0.0.0/8
Registry Mirrors:
docker.m.daocloud.io/
huecker.io/
dockerhub.timeweb.cloud/
Live Restore Enabled: false
Firewall Backend: iptables
7.镜像版本:cr.metax-tech.com/public-ai-release/maca/modelzoo.llm.vllm:1.0.3-maca.ai3.7.0.107-torch2.8-py312-ubuntu22.04-amd64
8.启动容器命令:
docker run -itd \
--privileged \
--network=host \
--ipc=host \
--security-opt seccomp=unconfined \
--security-opt apparmor=unconfined \
--ulimit memlock=-1 \
-v /pde_ai:/pde_ai \
--name deepseek-1-p-1 \
-v /home/metax/models/:/home/models/ \
cr.metax-tech.com/public-ai-release/maca/modelzoo.llm.vllm:1.0.3-maca.ai3.7.0.107-torch2.8-py312-ubuntu22.04-amd64 \
9.容器内执行命令:
export MACA_SMALL_PAGESIZE_ENABLE=1
export MACA_VLLM_ENABLE_MCTLASS_FUSED_MOE=1
export MACA_VLLM_ENABLE_MCTLASS_PYTHON_API=1
vllm serve /home/models/DeepSeek-V4-Flash-FlexSMQ-AWQ-W8A8/ --trust-remote-code \
--kv-cache-dtype bfloat16 --block-size 256 --gpu-memory-utilization 0.85 \
--tokenizer-mode deepseek_v4 --tool-call-parser deepseek_v4 \
--enable-auto-tool-choice --reasoning-parser deepseek_v4 \
--compilation-config '{"cudagraph_mode":"FULL_AND_PIECEWISE", "custom_ops":["all"]}' \
--max-num-seq 32 -tp 8 --speculative_config '{"method": "mtp", "num_speculative_tokens": 1}'
二、问题现象
算子重复注册,