Reduce vocabulary size in test model, fix bug in routing when overlapped #45
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
borzunov
reviewed
Aug 17, 2022
.github/workflows/run-tests.yaml
Outdated
| run: | | ||
| export HF_TAG=$(python -c "import os; print(os.environ.get('GITHUB_HEAD_REF') or os.environ.get('GITHUB_REF_NAME'))") | ||
| python -c "from huggingface_hub import delete_repo; delete_repo(token='$BLOOM_TESTING_WRITE_TOKEN', \ | ||
| name='test-bloomd-560m-$HF_TAG', organization='bloom-testing')" || true |
There was a problem hiding this comment.
Please consider moving that to a small Python script in the repo. Having a Bash command running Python, whose results are passed to another Bash command running Python, is not good.
There was a problem hiding this comment.
done, see tests/srcipts/remove_old_models.py
| sleep 60 # wait for server to download layers | ||
| sleep 10 # wait for initial servers to declare blocks, then let server decide which blocks to serve | ||
|
|
||
| python -m cli.run_server --converted_model_name_or_path $MODEL_NAME --block_indices 0:6 \ |
There was a problem hiding this comment.
Please consider running a loop instead, maybe using just --num_blocks without explicit --block_indices (I think using load balancing for all servers is a better test).
If it gets more complicated than just repeating something N times, please move it to a Python script.
There was a problem hiding this comment.
moved to #16
10 tasks
justheuristic
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
More than half of the test 560m model parameters is it's embedding layer (250_880 x 1024).
This adds 1-2 GB of RAM depending on the test scenario
This PR reduces this vocabulary size to save memory during conversion, keeping only the first 50k tokens
As a result,
Running CI with 4 servers uncovered a bug:
https://github.com/bigscience-workshop/petals/runs/7879399020?check_suite_focus=true
which is now also fixed
This PR also temporarily pins an older bnb version, pending @TimDettmers 's fix for cpuonly