Summary: 301 instances, 69 unique

Text	Count
# TODO: delete this logic if cloud service team can always pass model metadata	1
# TODO: refactor this when we migrate entirely to python 3	1
# TODO: Maybe implement some logic to make sure the park postion is always outside of the race track	1
# TODO: move this to before customer-specified so they can override	22
# TODO: use custom flag to indicate that this is in a pipeline rather than relying on the '*/*'	2
# TODO: Implement PBM: Position based model	1
# TODO: Fix the bot car dimension so that all collisions are detected	1
# TODO: delete upload_body_shell if body shell type is past in by cloud service	1
# TODO: flesh out docstrings	2
# TODO: replace agent for multi agent training	1
MIN_RESET_COUNT = 10000  # TODO: change when console passes float("inf")	1
# TODO: Randomize user selection	1
# TODO:  Selection should be based on frames in DDB table	1
# TODO: Include fault_code in the json schema to track faults - pending cloud team assistance	1
# TODO: Can add additional user attributes here.	1
# TODO: It might be theorically possible to have different kms keys for simtrace and mp4	1
# TODO: Write to stdin in chunks so that PIPE buffer never overflows	1
# TODO: this is wrong. fix it	1
#! TODO this needs to be removed after muti part is fixed, note we don't have	1
# TODO:	1
self.hyperparameters = ConfigurationList()  # TODO: move to shared	22
# DH TODO: check shape of self.num_actions. It should be int	1
)  # TODO: add a getter and setter	1
self.num_checkpoints_to_keep = 4  # TODO: make this a parameter	1
# TODO: Include fault_code in the json schema to track faults - pending cloud team assistance	1
# TODO check with Google if it's normal there is no dropout on the token classifier of SQuAD in the TF version	1
# TODO: Replace with get_item	1
# TODO: a model can only be put to pending, from pending state.	21
# TODO: Service team can't handle "version" key in Evaluation Metrics due to	1
# TODO: Check for errors in CLI args by polling the process	1
#! TODO each agent should have own config	1
#! TODO currently left and front camera use the same embedders, this is how it is wired up	1
# TODO: replace 'agent' for multiagent training	1
#! TODO each agent should have own s3 bucket	1
# TODO: Remove this whole class when nobody's using it any more.	22
# TODO: find more specific exception for resource not found	1
TODO: Deprecate and embed this class in ModelRecord.	21
# TODO: use ConfigList from Coach launcher, and share customization code.	22
# TODO: Find an atomic way to check if file is present else create	1
#! TODO decide whether we need lidar layers for different network types	1
# TODO change agent to specific agent name for multip agent case	1
# TODO replace agent with agent_0 and so on for multiagent case	1
# TODO: conditional check to verify model is in *ing state while updating...	42
# TODO: figure out what display name to use during wait	1
TODO: implemement logic here if we decide to log non fatal error metrics	1
# TODO: replace 'agent' with 'agent_0' for multi agent training and	1
#! TODO remove ignore_lock after refactoring this class	1
#! TODO decide if we want to have a deep-deep topology that differes from deep	1
# TODO: Refactor the flow to remove conditional checks for specific algorithms	1
# TODO: find better way to load checkpoints that were saved with a global network into the online network	1
# TODO: maybe clamping the value will provide a better customer experience.	1
# TODO: delete after car_color is a mandatory yaml key	1
# TODO: replace 'agent' with specific agent name for multi agent training	1
# TODO: Hard coding the frustum values for now. These values need to be loaded from SDF directly	1
# TODO: Remove below camera_offsets and camera_pitch variables if we can get Camera pose directly	1
# TODO: This code should be removed when the cloud service starts providing VIDEO_JOB_TYPE YAML parameter	1
#! TODO evaluate if this is the best way to reset the car	1
# TODO: replace 'agent' with name of each agent for multi-agent training	1
# TODO: Refactor the flow to remove conditional checks for specific algorithms	1
# TODO: add validation/instructions if multiple deployment	21
# TODO import method from starter-kit	1
# % TODO - refactor this module to be more modular based on the training algorithm and avoid if-else	1
torch.cuda.empty_cache()  # TODO check if it helps	1
# TODO: replace 'agent' with name of each agent	3
# TODO: Since we are not running Grand Prix in RoboMaker,	1
# TODO: a model eval_state can only be put to pending, from pending state	21
# TODO: Error handling in parsing the given example	3
# TODO: THIS CHECK IS VERY UGLY	1
# TODO: remove this after converting all samples.	22