-
Notifications
You must be signed in to change notification settings - Fork 6.7k
Closed
Description
System information
- OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
- Ray installed from (source or binary):
- Ray version: 0.7.0.dev2
- Python version:
- Exact command to reproduce:
Describe the problem
Passing restore=$HOME/exp/.../checkpoint-100
into tune.run()
, the checkpoint will not be loaded and ray will start a brand new experiment.
Source code / logs
In tune.run()
function, an experiment are automatically initialized as:
experiment = Experiment(
name, run_or_experiment, stop, config, resources_per_trial,
num_samples, local_dir, upload_dir, trial_name_creator, loggers,
sync_function, checkpoint_freq, checkpoint_at_end, export_formats,
max_failures, restore)
However, the arguments passed into the Experiment.__init__()
have wrong order as:
def __init__(self,
name,
run,
stop=None,
config=None,
resources_per_trial=None,
num_samples=1,
local_dir=None,
upload_dir=None,
trial_name_creator=None,
loggers=None,
sync_function=None,
checkpoint_freq=0,
checkpoint_at_end=False,
keep_checkpoints_num=None,
checkpoint_score_attr=None,
export_formats=None,
max_failures=3,
restore=None,
repeat=None,
trial_resources=None,
custom_loggers=None):
You can see that this is because the introduction of two new arguments:
keep_checkpoints_num=None,
checkpoint_score_attr=None,
I believe this is caused by the #4490 . I will try to fix this problem.
Metadata
Metadata
Assignees
Labels
No labels