log meta (#2086)

* log meta

* move collect_env to mmdet/utils

* fix import error

* fix isort

* add dash line after env info

.github/ISSUE_TEMPLATE/error-report.md

@@ -26,7 +26,7 @@ A placeholder for the command.
 
 **Environment**
 
-1. Please run `python tools/collect_env.py` to collect necessary environment infomation and paste it here.
+1. Please run `python mmdet/utils/collect_env.py` to collect necessary environment infomation and paste it here.
 2. You may add addition that may be helpful for locating the problem, such as
     - How you installed PyTorch [e.g., pip, conda, source]
     - Other environment variables that may be related (such as `$PATH`, `$LD_LIBRARY_PATH`, `$PYTHONPATH`, etc.)

mmdet/apis/train.py

@@ -86,7 +86,8 @@ def train_detector(model,
                    cfg,
                    distributed=False,
                    validate=False,
-                   timestamp=None):
+                   timestamp=None,
+                   meta=None):
     logger = get_root_logger(cfg.log_level)
 
     # start training
@@ -97,15 +98,17 @@ def train_detector(model,
             cfg,
             validate=validate,
             logger=logger,
-            timestamp=timestamp)
+            timestamp=timestamp,
+            meta=meta)
     else:
         _non_dist_train(
             model,
             dataset,
             cfg,
             validate=validate,
             logger=logger,
-            timestamp=timestamp)
+            timestamp=timestamp,
+            meta=meta)
 
 
 def build_optimizer(model, optimizer_cfg):
@@ -193,7 +196,8 @@ def _dist_train(model,
                 cfg,
                 validate=False,
                 logger=None,
-                timestamp=None):
+                timestamp=None,
+                meta=None):
     # prepare data loaders
     dataset = dataset if isinstance(dataset, (list, tuple)) else [dataset]
     data_loaders = [
@@ -210,7 +214,12 @@ def _dist_train(model,
     # build runner
     optimizer = build_optimizer(model, cfg.optimizer)
     runner = Runner(
-        model, batch_processor, optimizer, cfg.work_dir, logger=logger)
+        model,
+        batch_processor,
+        optimizer,
+        cfg.work_dir,
+        logger=logger,
+        meta=meta)
     # an ugly walkaround to make the .log and .log.json filenames the same
     runner.timestamp = timestamp
 
@@ -244,7 +253,8 @@ def _non_dist_train(model,
                     cfg,
                     validate=False,
                     logger=None,
-                    timestamp=None):
+                    timestamp=None,
+                    meta=None):
     if validate:
         raise NotImplementedError('Built-in validation is not implemented '
                                   'yet in not-distributed training. Use '
@@ -267,7 +277,12 @@ def _non_dist_train(model,
     # build runner
     optimizer = build_optimizer(model, cfg.optimizer)
     runner = Runner(
-        model, batch_processor, optimizer, cfg.work_dir, logger=logger)
+        model,
+        batch_processor,
+        optimizer,
+        cfg.work_dir,
+        logger=logger,
+        meta=meta)
     # an ugly walkaround to make the .log and .log.json filenames the same
     runner.timestamp = timestamp
     # fp16 setting

mmdet/utils/__init__.py

@@ -1,8 +1,9 @@
+from .collect_env import collect_env
 from .flops_counter import get_model_complexity_info
 from .logger import get_root_logger, print_log
 from .registry import Registry, build_from_cfg
 
 __all__ = [
     'Registry', 'build_from_cfg', 'get_model_complexity_info',
-    'get_root_logger', 'print_log'
+    'get_root_logger', 'print_log', 'collect_env'
 ]

tools/collect_env.py → mmdet/utils/collect_env.py

@@ -9,7 +9,6 @@
 import torchvision
 
 import mmdet
-from mmdet.ops import get_compiler_version, get_compiling_cuda_version
 
 
 def collect_env():
@@ -53,12 +52,12 @@ def collect_env():
 
     env_info['MMCV'] = mmcv.__version__
     env_info['MMDetection'] = mmdet.__version__
+    from mmdet.ops import get_compiler_version, get_compiling_cuda_version
     env_info['MMDetection Compiler'] = get_compiler_version()
     env_info['MMDetection CUDA Compiler'] = get_compiling_cuda_version()
-
-    for name, val in env_info.items():
-        print('{}: {}'.format(name, val))
+    return env_info
 
 
 if __name__ == "__main__":
-    collect_env()
+    for name, val in collect_env().items():
+        print('{}: {}'.format(name, val))

tools/train.py

@@ -13,7 +13,7 @@
 from mmdet.apis import set_random_seed, train_detector
 from mmdet.datasets import build_dataset
 from mmdet.models import build_detector
-from mmdet.utils import get_root_logger
+from mmdet.utils import collect_env, get_root_logger
 
 
 def parse_args():
@@ -86,9 +86,20 @@ def main():
     log_file = osp.join(cfg.work_dir, '{}.log'.format(timestamp))
     logger = get_root_logger(log_file=log_file, log_level=cfg.log_level)
 
+    # init the meta dict to record some important information such as
+    # environment info and seed, which will be logged
+    meta = dict()
+    # log env info
+    env_info_dict = collect_env()
+    env_info = '\n'.join([('{}: {}'.format(k, v))
+                          for k, v in env_info_dict.items()])
+    dash_line = '-' * 60 + '\n'
+    logger.info('Environment info:\n' + dash_line + env_info + '\n' +
+                dash_line)
+    meta['env_info'] = env_info
+
     # log some basic info
     logger.info('Distributed training: {}'.format(distributed))
-    logger.info('MMDetection Version: {}'.format(__version__))
     logger.info('Config:\n{}'.format(cfg.text))
 
     # set random seeds
@@ -97,6 +108,7 @@ def main():
             args.seed, args.deterministic))
         set_random_seed(args.seed, deterministic=args.deterministic)
     cfg.seed = args.seed
+    meta['seed'] = args.seed
 
     model = build_detector(
         cfg.model, train_cfg=cfg.train_cfg, test_cfg=cfg.test_cfg)
@@ -119,7 +131,8 @@ def main():
         cfg,
         distributed=distributed,
         validate=args.validate,
-        timestamp=timestamp)
+        timestamp=timestamp,
+        meta=meta)
 
 
 if __name__ == '__main__':