Remove awscli dependency

[dfilan]

Description

Fixes #560.

Testing

Ran all tests. Results: 3543 passed, 595 skipped, 4234 warnings.

[codecov]

Codecov Report

Merging #572 (ac944e3) into master (005c15f) will decrease coverage by 0.01%.
The diff coverage is 66.66%.

@@            Coverage Diff             @@
##           master     #572      +/-   ##
==========================================
- Coverage   97.43%   97.42%   -0.02%     
==========================================
  Files          88       88              
  Lines        8375     8378       +3     
==========================================
+ Hits         8160     8162       +2     
- Misses        215      216       +1     

Impacted Files	Coverage Δ
tests/test_experiments.py	100.00% <ø> (ø)
src/imitation/scripts/config/eval_policy.py	76.81% <66.66%> (-0.47%)	⬇️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[dfilan]

TODO get rid of scripts that download from aws

[dfilan]

Having a touch of trouble testing due to #576.

[dfilan]

Current status:

• have removed experiment scripts that critically rely on aws
• have retouched experiments/README.md accordingly
• have edited experiment config files to match with what's in HuggingFace
• experiment scripts don't currently work, because they don't like the format in which things are saved in HuggingFace

I turned the last bullet point into its own issue (#576), but think it should be addressed separately - IMO it kind of sucks that someone can't just clone this repository and run pip install .[dev] but instead has to comment out that bit of setup.py.

[dfilan]

BTW I kept imit_benchmark.sh because I think it makes sense as its own thing, even without the context of download_models.sh.

[AdamGleave]

LGTM apart from deleting one section that I think shouldn't have been.

[AdamGleave]

Why did you delete this? If it was because experiments/transfer_benchmark.sh doesn't exist I think it just got renamed to experiments/transfer_learn_benchmark.sh.

[dfilan]

I think maybe I deleted it because currently stuff doesn't save to data/reward_models. I guess I can reinstate it and leave fixing it for a general "make the experiment scripts work" PR.

[dfilan]

Have reinstated it.

[AdamGleave]

Ah yeah that is a problem but agree probably best to punt it to another issue. We should really overhaul the whole experiments/ directory -- I expect @taufeeque9 will end up doing this as part of his benchmarking work.