Fix proxy environment variables not used in v1.0 #3412
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Solve huggingface/transformers#41301 reported by @jerryzh168.
The root cause of this issue was when using a custom transport, the proxy env variables are not used by
httpx
even iftrust_env=True
. This is the case since encode/httpx#1122. We were using custom Transport in order to handleHF_HUB_OFFLINE
env variable (i.e. raise exception if attempt to make an HTTP call) and add custom request id header to help with debugging.The workaround in this PR is to use
event_hooks
instead. We are still catching offline mode + adding debug logs at runtime but don't use custom transport anymore. I've added a regression test to check thatHTTP_PROXY
/HTTPS_PROXY
are correctly read from env.Note: there is a drawback to this PR. We do not catch any
httpx.RequestError
s anymore to add the request id to the error message itself. This is not possible to do without a custom Transport. In practice it's not much of an issue because the request id is added back byhf_raise_for_status
. The only regression is if an error happens before the server respond (i.e. a readtimeout or connection error), in which case we won't have a request id in the error message. The benefits of this PR still outweigh this drawback IMO.