-
LV.1 #3 whcjb回复使用mm_run报同样错误。
mm_run --magicmind_model ./Peop_Vehi_NoVe_v1.0.0.onnx_b1_fp32.cn
==================== Parameter Information
--magicmind_model<Str> : ./Peop_Vehi_NoVe_v1.0.0.onnx_b1_fp32.cn
--input_dims<Vec<Vec<Int>>> :
--batch_size<Vec<Int>> :
--run_config<Str> :
--input_files<Vec<Str>> :
--output_path<Str> :
--plugin<Vec<Str>> :
--devices<Vec<Int>> : 0
--threads<Int> : 1
--bind_cluster<Bool> : false
--warmup<Int> : 200
--duration<Int> : 3000
--iterations<Int> : 100
--disable_data_copy<Bool> : false
--host_async<Bool> : false
--buffer_depth<Int> : 2
--infer_depth<Int> : 2
--kernel_capture<Bool> : false
--trace_path<Str> :
--debug_path<Str> :
--perf_path<Str> :
--trace_pmu<Bool> : false
--trace_time<Str> : host
--avg_runs<Vec<Int>> : 100,10
MagicMind: 269484032
CNRT: 60202
CNAPI: 20202
PID: 1967
2023-08-03 08:48:31.830296: WARNING: samples/cc/mm_run/run.cc:116] Buffer/Infer depth beyond 1 actually shows weak performance impact when disable data copy is on, but doubles the memory footprint.
2023-08-03 08:48:31.830301: WARNING: samples/cc/mm_run/run.cc:117] If user is doing memory usage observation, we recommand to set infer/buffer depth to 1.
2023-08-03 08:48:31.834759: [cnrtError] [1967] [Card : 0] Error occurred during calling 'cnInit' in CNDrv interface.
2023-08-03 08:48:31.834819: [cnrtError] [1967] [Card : 0] Return value is 100004, CN_ERROR_NO_DEVICE, means that "No useful mlu device"
2023-08-03 08:48:31.834831: [cnrtError] [1967] [Card : 0] Failed to initialize CNDrv.
2023-08-03 08:48:31.834843: [cnrtError] [1967] [Card : 0] Error occurred during calling 'cnDeviceGetCount' in CNDrv interface.
2023-08-03 08:48:31.834853: [cnrtError] [1967] [Card : 0] Return value is 100002, CN_ERROR_NOT_INITIALIZED, means that "initialization error"
2023-08-03 08:48:31.834863: [cnrtError] [1967] [Card : 0] get device count failed!
2023-08-03 08:48:31.834877: [cnrtError] [1967] [Card : 0] Failed to initialize CNRT 632015
2023-08-03 08:48:31.834893: [cnrtError] [1967] [Card : 0] Set the current device ordinal failed.
2023-08-03 08:48:31.834906: ERROR: samples/cc/mm_run/run.cc:134] CNRT ERROR raised.
[samples/cc/mm_run/run.cc:134] CNRT error, code=100006(invalid argument.) "__ret"
-
LV.1 #2 whcjb回复
docker内部可以使用cnmon
================CNMON 1.19.9================
Card 0
Product Name : MLU270-S4
SN : 122009100958
UUID : 58091009-2012-0000-0000-000000000000
Firmware : v1.1.3
Driver : v4.9.13
Utilization :
Image Codec 0-1 : 0 % 0 %
Image Codec 2-3 : 0 % 0 %
Image Codec 4-5 : 0 % 0 %
Video Codec 0-1 : 0 % 0 %
Video Codec 2-3 : 0 % 0 %
Video Codec 4-5 : 0 % 0 %
MLU Average : 0 %
MLU 0-3 : 0 % 0 % 0 % 0 %
MLU 4-7 : 0 % 0 % 0 % 0 %
MLU 8-11 : 0 % 0 % 0 % 0 %
MLU 12-15 : 0 % 0 % 0 % 0 %
idle : 16
busy : 0
Device CPU Chip : 25 %
Device CPU Core 0-1 : 0 % 100 %
Device CPU Core 2-3 : 0 % 0 %
CPU Sampling Interval : 100 ms
Codec Turbo : N/A
Fan Speed : 0 %
Temperature :
Board : 76 C
Cluster 0 : 77 C
Cluster 1 : 77 C
Cluster 2 : 77 C
Cluster 3 : 77 C
Frequency :
IPU : 1000 MHz
Physical Memory Usage :
Total : 16384 MiB
Used : 2900 MiB
Free : 13484 MiB
Channel Memory Usage :
Channel 0 : 725 MiB
Channel 1 : 725 MiB
Channel 2 : 725 MiB
Channel 3 : 725 MiB
DDR Data Widths : 256 bit
DDR BandWidth : 102 GB/s
DDR Transfer Rate : 3200 Mbps
Virtual Memory Usage :
Total : 1048576 MiB
Used : 10240 MiB
Free : 1038336 MiB
Fast Alloc Memory :
Total : 0 KB
Used : 0 KB
Free : 0 KB
Power :
Usage : 29 W
Cap : 70 W
Initialized : On
DDR ECC Err Count : 0
Corrected Err : 0
One Bit Err : 0
Mul One Bit Err : 0
Uncorrected Err : 0
Mul Bits Err : 0
Mul Mul Bits Err : 0
Cache :
Total : N/A
Hit : N/A
PCI :
Vendor ID : 0xcabc
Device ID : 0x270
Sub-Vendor ID : 0xcabc
Sub-System ID : 0x12
Domain ID : 0000
Bus num : 30
Device : 00
Function : 0
Physical Slot : Unknown
Max Speed : 8 GT/s
Max Width : x16
Current Speed : 8 GT/s
Current Width : x8
Switch Link Speed : 8 GT/s
NUMA node id : 0
Bandwidth : N/A
PCIe throughput : N/A
Local memory : N/A
Chassis : N/A
Retired Pages : N/A
Row-Remapping : N/A
-
请登录后评论