search

xinyu1205 / recognize-anything

Open-source and strong foundation image recognition models.
https://recognize-anything.github.io/
Apache License 2.0
2.86k stars 274 forks source link

Error when run finetune.py #130

Open shuxjweb opened 11 months ago

shuxjweb commented 11 months ago

Hi, I encountered an error while running finetune.py, the program died, what is the reason?

python -m torch.distributed.run --nproc_per_node=8 finetune.py \ --model-type ram_plus \ --config ram/configs/finetune.yaml \ --checkpoint /models/RAM/ram_plus_swin_large_14m.pth \ --output-dir /logs/RAM/20231205_ramplus_coco_finetune

(p38t20) [root@ts-80e08ce490704c3aa7d3ca229319b5a9-launcher /recognize_anything]# sh start.sh WARNING:main:

Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.

2023-12-06 15:56:54.766918: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0. 2023-12-06 15:56:54.779575: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0. 2023-12-06 15:56:54.783312: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0. 2023-12-06 15:56:54.811370: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0. 2023-12-06 15:56:54.811370: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0. 2023-12-06 15:56:54.819133: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. 2023-12-06 15:56:54.831586: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. 2023-12-06 15:56:54.835100: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. 2023-12-06 15:56:54.847594: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0. 2023-12-06 15:56:54.847615: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0. 2023-12-06 15:56:54.847626: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable TF_ENABLE_ONEDNN_OPTS=0. 2023-12-06 15:56:54.862773: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. 2023-12-06 15:56:54.862790: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. 2023-12-06 15:56:54.898752: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. 2023-12-06 15:56:54.898803: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. 2023-12-06 15:56:54.899451: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. 2023-12-06 15:56:55.554178: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT 2023-12-06 15:56:55.574930: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT 2023-12-06 15:56:55.578458: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT 2023-12-06 15:56:55.600197: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT 2023-12-06 15:56:55.600411: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT 2023-12-06 15:56:55.633909: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT 2023-12-06 15:56:55.636835: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT 2023-12-06 15:56:55.639246: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT | distributed init (rank 1, word 8): env:// | distributed init (rank 4, word 8): env:// | distributed init (rank 2, word 8): env:// | distributed init (rank 5, word 8): env:// | distributed init (rank 0, word 8): env:// | distributed init (rank 3, word 8): env:// | distributed init (rank 7, word 8): env:// | distributed init (rank 6, word 8): env:// Creating dataset loading /data/img_txt/recognize-anything-dataset-14m/coco_train_rmcocodev_ram.json number of training samples: 547741 Creating model load from: /models/RAM/ram_plus_swin_large_14m.pth Creating pretrained CLIP model Creating RAM model

/models/RAM/ram_plus_swin_large_14m.pth

load checkpoint from /models/RAM/ram_plus_swin_large_14m.pth vit: swin_l Start training WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 46292 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 46293 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 46294 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 46295 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 46297 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 46298 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 46299 closing signal SIGTERM ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: -11) local_rank: 4 (pid: 46296) of binary: /root/miniconda3/envs/p38t20/bin/python Traceback (most recent call last): File "/root/miniconda3/envs/p38t20/lib/python3.8/runpy.py", line 194, in _run_module_as_main return _run_code(code, main_globals, None, File "/root/miniconda3/envs/p38t20/lib/python3.8/runpy.py", line 87, in _run_code exec(code, run_globals) File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/run.py", line 798, in main() File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/elastic/multiprocessing/errors/init.py", line 346, in wrapper return f(*args, **kwargs) File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/run.py", line 794, in main run(args) File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/run.py", line 785, in run elastic_launch( File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/launcher/api.py", line 134, in call return launch_agent(self._config, self._entrypoint, list(args)) File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/launcher/api.py", line 250, in launch_agent raise ChildFailedError( torch.distributed.elastic.multiprocessing.errors.ChildFailedError:

finetune.py FAILED

Failures:

---------------------------------------------------------- Root Cause (first observed failure): [0]: time : 2023-12-06_15:57:22 host : ts-80e08ce490704c3aa7d3ca229319b5a9-launcher rank : 4 (local_rank: 4) exitcode : -11 (pid: 46296) error_file: traceback : Signal 11 (SIGSEGV) received by PID 46296 ==========================================================
shuxjweb commented 11 months ago

I have conducted several experiments for running finetune.py:

  1. 1GPU, image_size=224, the code can run successful.
  2. 1GPU, image_size=384, the code can run successful.
  3. 8GPU, image_size=224, the code can run successful.

  4. 8GPU, image_size=384, the running kill be killed. The reported errors are shown as follows: """ Start training WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 17968 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 17970 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 17971 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 17972 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 17973 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 17974 closing signal SIGTERM WARNING:torch.distributed.elastic.multiprocessing.api:Sending process 17975 closing signal SIGTERM ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: -11) local_rank: 1 (pid: 17969) of binary: /root/miniconda3/envs/p38t20/bin/python Traceback (most recent call last): File "/root/miniconda3/envs/p38t20/lib/python3.8/runpy.py", line 194, in _run_module_as_main return _run_code(code, main_globals, None, File "/root/miniconda3/envs/p38t20/lib/python3.8/runpy.py", line 87, in _run_code exec(code, run_globals) File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/run.py", line 798, in main() File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/elastic/multiprocessing/errors/init.py", line 346, in wrapper return f(*args, **kwargs) File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/run.py", line 794, in main run(args) File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/run.py", line 785, in run elastic_launch( File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/launcher/api.py", line 134, in call return launch_agent(self._config, self._entrypoint, list(args)) File "/root/miniconda3/envs/p38t20/lib/python3.8/site-packages/torch/distributed/launcher/api.py", line 250, in launch_agent raise ChildFailedError( torch.distributed.elastic.multiprocessing.errors.ChildFailedError:

    finetune_fp16.py FAILED

    Failures:

    ---------------------------------------------------------- Root Cause (first observed failure): [0]: time : 2023-12-07_00:39:25 host : ts-80e08ce490704c3aa7d3ca229319b5a9-launcher rank : 1 (local_rank: 1) exitcode : -11 (pid: 17969) error_file: traceback : Signal 11 (SIGSEGV) received by PID 17969 ========================================================== """
xinyu1205 commented 11 months ago

This seems to be an environmental issue instead of a code issue, and I have never encountered this issue before.

shuxjweb commented 11 months ago

Thank you for your reply. I feel very strange, single-machine multi-gpu, the image size set to 224 can run normally, but change to 384 will report an error. I have checked the GPU memory, because the batchsize setting is small (2), when setting 384, the memory occupation is less than half, should not be caused by insufficient GPU memory.