Feature/sg 646 rename logs

src/super_gradients/common/auto_logging/auto_logger.py

@@ -32,7 +32,7 @@ def _setup_default_logging(self, log_level: str = None) -> None:
         # Therefore the log file will have the parent PID to being able to discriminate the logs corresponding to a single run.
         timestamp = time.strftime("%Y_%m_%d_%H_%M_%S", time.localtime())
         self._setup_logging(
-            filename=os.path.expanduser(f"~/sg_logs/sg_logs_{os.getppid()}_{timestamp}.log"),
+            filename=os.path.expanduser(f"~/sg_logs/logs_{os.getppid()}_{timestamp}.log"),
             copy_already_logged_messages=False,
             filemode="w",
             log_level=log_level,

src/super_gradients/common/data_interface/adnn_model_repository_data_interface.py

@@ -26,7 +26,7 @@ def __init__(self, data_connection_location: str = "local", data_connection_cred
         """
         super().__init__()
         self.tb_events_file_prefix = "events.out.tfevents"
-        self.log_file_prefix = "log_"
+        self.log_file_prefix = "experiment_logs_"
         self.latest_checkpoint_filename = "ckpt_latest.pth"
         self.best_checkpoint_filename = "ckpt_best.pth"
 

src/super_gradients/common/sg_loggers/base_sg_logger.py

@@ -22,6 +22,10 @@
 
 logger = get_logger(__name__)
 
+EXPERIMENT_LOGS_PREFIX = "experiment_logs"
+LOGGER_LOGS_PREFIX = "logs"
+CONSOLE_LOGS_PREFIX = "console"
+
 
 class BaseSGLogger(AbstractSGLogger):
     def __init__(
@@ -120,16 +124,22 @@ def _make_dir(self):
     @multi_process_safe
     def _init_log_file(self):
         time_string = time.strftime("%b%d_%H_%M_%S", time.localtime())
-        # There are two log files, since the regular log_file_path used for `manual` logging of configs/other info
-        self.log_file_path = f"{self._local_dir}/log_{time_string}.txt"
-        self.log_full_file_path = f"{self._local_dir}/sg_logs_{time_string}.txt"
-        self.console_sink_path = f"{self._local_dir}/console_{time_string}.txt"
-        AutoLoggerConfig.setup_logging(filename=self.log_full_file_path, copy_already_logged_messages=True)
+
+        # Where the experiment related info will be saved (config and training/validation results per epoch_
+        self.experiment_log_path = f"{self._local_dir}/{EXPERIMENT_LOGS_PREFIX}_{time_string}.txt"
+
+        # Where the logger.log will be saved
+        self.logs_path = f"{self._local_dir}/{LOGGER_LOGS_PREFIX}_{time_string}.txt"
+
+        # Where the console prints/logs will be saved
+        self.console_sink_path = f"{self._local_dir}/{CONSOLE_LOGS_PREFIX}_{time_string}.txt"
+
+        AutoLoggerConfig.setup_logging(filename=self.logs_path, copy_already_logged_messages=True)
         ConsoleSink.set_location(filename=self.console_sink_path)
 
     @multi_process_safe
     def _write_to_log_file(self, lines: list):
-        with open(self.log_file_path, "a" if os.path.exists(self.log_file_path) else "w") as log_file:
+        with open(self.experiment_log_path, "a" if os.path.exists(self.experiment_log_path) else "w") as log_file:
             for line in lines:
                 log_file.write(line + "\n")
 
@@ -243,7 +253,7 @@ def upload(self):
             self.model_checkpoints_data_interface.save_remote_tensorboard_event_files(self.experiment_name, self._local_dir)
 
         if self.save_logs_remote:
-            log_file_name = self.log_file_path.split("/")[-1]
+            log_file_name = self.experiment_log_path.split("/")[-1]
             self.model_checkpoints_data_interface.save_remote_checkpoints_file(self.experiment_name, self._local_dir, log_file_name)
 
     @multi_process_safe

src/super_gradients/common/sg_loggers/clearml_sg_logger.py

@@ -212,8 +212,8 @@ def upload(self):
             self.task.upload_artifact(name=name, artifact_object=self._get_tensorboard_file_name())
 
         if self.save_logs:
-            name = self.log_file_path.split("/")[-1]
-            self.task.upload_artifact(name=name, artifact_object=self.log_file_path)
+            name = self.experiment_log_path.split("/")[-1]
+            self.task.upload_artifact(name=name, artifact_object=self.experiment_log_path)
 
     @multi_process_safe
     def add_checkpoint(self, tag: str, state_dict: dict, global_step: int = 0):

src/super_gradients/common/sg_loggers/deci_platform_sg_logger.py

@@ -2,15 +2,14 @@
 from typing import Optional
 
 from super_gradients.common.abstractions.abstract_logger import get_logger
-from super_gradients.common.sg_loggers.base_sg_logger import BaseSGLogger
+from super_gradients.common.sg_loggers.base_sg_logger import BaseSGLogger, EXPERIMENT_LOGS_PREFIX, LOGGER_LOGS_PREFIX, CONSOLE_LOGS_PREFIX
 from super_gradients.common.environment.ddp_utils import multi_process_safe
 from super_gradients.common.plugins.deci_client import DeciClient
 
 logger = get_logger(__name__)
 
 
 TENSORBOARD_EVENTS_PREFIX = "events.out.tfevents"
-LOGS_PREFIX = "log_"
 
 
 class DeciPlatformSGLogger(BaseSGLogger):
@@ -74,7 +73,10 @@ def upload(self):
             raise ValueError("Provided directory does not exist")
 
         self._upload_latest_file_starting_with(start_with=TENSORBOARD_EVENTS_PREFIX)
-        self._upload_latest_file_starting_with(start_with=LOGS_PREFIX)
+        self._upload_latest_file_starting_with(start_with=EXPERIMENT_LOGS_PREFIX)
+        self._upload_latest_file_starting_with(start_with=LOGGER_LOGS_PREFIX)
+        self._upload_latest_file_starting_with(start_with=CONSOLE_LOGS_PREFIX)
+        self._upload_folder_files(folder_name=".hydra")
 
     @multi_process_safe
     def _upload_latest_file_starting_with(self, start_with: str):
@@ -91,3 +93,19 @@ def _upload_latest_file_starting_with(self, start_with: str):
         most_recent_file_path = max(files_path, key=os.path.getctime)
         self.platform_client.save_experiment_file(file_path=most_recent_file_path)
         logger.info(f"File saved to Deci platform: {most_recent_file_path}")
+
+    @multi_process_safe
+    def _upload_folder_files(self, folder_name: str):
+        """
+        Upload all the files of a given folder.
+
+        :param folder_name: Name of the folder that contains the files to upload
+        """
+        folder_path = os.path.join(self.checkpoints_dir_path, folder_name)
+
+        if not os.path.exists(folder_path):
+            return
+
+        for file in os.listdir(folder_path):
+            self.platform_client.save_experiment_file(file_path=f"{folder_path}/{file}")
+            logger.info(f"File saved to Deci platform: {folder_path}/{file}")

src/super_gradients/common/sg_loggers/wandb_sg_logger.py

@@ -223,7 +223,7 @@ def upload(self):
             wandb.save(glob_str=self._get_tensorboard_file_name(), base_path=self._local_dir, policy="now")
 
         if self.save_logs_wandb:
-            wandb.save(glob_str=self.log_file_path, base_path=self._local_dir, policy="now")
+            wandb.save(glob_str=self.experiment_log_path, base_path=self._local_dir, policy="now")
 
     @multi_process_safe
     def add_checkpoint(self, tag: str, state_dict: dict, global_step: int = 0):