deep-rl-class
deep-rl-class copied to clipboard
[FIX] Wrong Huggy path when pushing to Hub
Current version of pushing Huggy to Hub:
!mlagents-push-to-hf --run-id="HuggyTraining" --local-dir="./results/Huggy2" --repo-id="Laz4rz/hf-huggy-1-bonus" --commit-message="add Huggy agent"
yields:
[INFO] This function will create a model card and upload your HuggyTraining into HuggingFace Hub. This is a work in progress: If you encounter a bug, please send open an issue
Traceback (most recent call last):
File "/usr/local/bin/mlagents-push-to-hf", line 33, in <module>
sys.exit(load_entry_point('mlagents', 'console_scripts', 'mlagents-push-to-hf')())
File "/content/ml-agents/ml-agents/mlagents/utils/push_to_hf.py", line 205, in main
package_to_hub(
File "/content/ml-agents/ml-agents/mlagents/utils/push_to_hf.py", line 159, in package_to_hub
_generate_config(local_path, configfile_name)
File "/content/ml-agents/ml-agents/mlagents/utils/push_to_hf.py", line 29, in _generate_config
with open(os.path.join(local_dir, configfile_name)) as yaml_in:
FileNotFoundError: [Errno 2] No such file or directory: 'results/Huggy/configuration.yaml'
This only needs the path to be changed from "Huggy" to "Huggy2".
Hey there 👋 , the local dir depends on the run-id name.
There's no bug: for instance if you call your run-id = Huggytesttest the local dir will be --local-dir="./results/Huggytestest"
Hey,
yes of course it is run id dependent, but for now this is just not cohere. There is no reason the --run-id to be first Huggy2 and then Huggy when saving the model.
This of course is a very simple thing to fix while running the notebook. But there is simply no reason for these two not to be aligned. The point of the notebook is not to look for path correctness.
Ok I see what you mean sorry for the misunderstanding, I merge your PR.
Thanks again 🤗