Fix eval #5955
Closed
Fix eval #5955
Conversation
This file contains hidden or 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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/5955
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 4158d4c with merge base 36a5bc6 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D62198560 |
helunwencser
pushed a commit
to helunwencser/executorch
that referenced
this pull request
Oct 7, 2024
Summary: This PR fixes a bunch of issues in the eval pipeline: - Use the right token for `eot_token_id` - Do not add `bos` and `eos` during `tok_encode` based on this [discussion](https://fburl.com/code/uifmt746). - Update `executorch/examples/models/llama2/tokenizer/tiktoken.py` to be synced with llama 3.1's official [code](https://github.com/meta-llama/llama-models/blob/main/models/llama3/api/tokenizer.py). Majorly updated set of special tokens. - Update `--limit`'s default value to none based on `lm_eval`'s [doc](https://github.com/EleutherAI/lm-evaluation-harness/blob/main/docs/interface.md). It should only be used during test. - Update `--dtype-override`'s default value to none such that we don't override mode's dtype by default. For the context, we observed a gap between ExecuTorch's eval result and SpinQuant's eval result: 19 vs 14. After these fixes, ExecuTorch's eval result is closer to SpinQuant's eval. Differential Revision: D62198560
helunwencser
force-pushed
the
export-D62198560
branch
from
f105c39 to
bf05031
Compare
|
This pull request was exported from Phabricator. Differential Revision: D62198560 |
helunwencser
force-pushed
the
export-D62198560
branch
from
bf05031 to
004045b
Compare
helunwencser
pushed a commit
to helunwencser/executorch
that referenced
this pull request
Oct 8, 2024
Summary: This PR fixes a bunch of issues in the eval pipeline: - Use the right token for `eot_token_id` - Do not add `bos` and `eos` during `tok_encode` based on this [discussion](https://fburl.com/code/uifmt746). - Update `executorch/examples/models/llama2/tokenizer/tiktoken.py` to be synced with llama 3.1's official [code](https://github.com/meta-llama/llama-models/blob/main/models/llama3/api/tokenizer.py). Majorly updated set of special tokens. - Update `--limit`'s default value to none based on `lm_eval`'s [doc](https://github.com/EleutherAI/lm-evaluation-harness/blob/main/docs/interface.md). It should only be used during test. - Update `--dtype-override`'s default value to none such that we don't override mode's dtype by default. For the context, we observed a gap between ExecuTorch's eval result and SpinQuant's eval result: 19 vs 14. After these fixes, ExecuTorch's eval result is closer to SpinQuant's eval. Differential Revision: D62198560
|
This pull request was exported from Phabricator. Differential Revision: D62198560 |
helunwencser
pushed a commit
to helunwencser/executorch
that referenced
this pull request
Oct 8, 2024
Summary: This PR fixes a bunch of issues in the eval pipeline: - Use the right token for `eot_token_id` - Do not add `bos` and `eos` during `tok_encode` based on this [discussion](https://fburl.com/code/uifmt746). - Update `executorch/examples/models/llama2/tokenizer/tiktoken.py` to be synced with llama 3.1's official [code](https://github.com/meta-llama/llama-models/blob/main/models/llama3/api/tokenizer.py). Majorly updated set of special tokens. Differential Revision: D62198560
helunwencser
force-pushed
the
export-D62198560
branch
from
004045b to
1ecad67
Compare
|
This pull request was exported from Phabricator. Differential Revision: D62198560 |
mergennachin
approved these changes
Oct 8, 2024
Summary: This PR fixes a bunch of issues in the eval pipeline: - Use the right token for `eot_token_id` - Do not add `bos` and `eos` during `tok_encode` based on this [discussion](https://fburl.com/code/uifmt746). - Update `executorch/examples/models/llama2/tokenizer/tiktoken.py` to be synced with llama 3.1's official [code](https://github.com/meta-llama/llama-models/blob/main/models/llama3/api/tokenizer.py). Majorly updated set of special tokens. Reviewed By: mergennachin Differential Revision: D62198560
helunwencser
force-pushed
the
export-D62198560
branch
from
1ecad67 to
4158d4c
Compare
|
This pull request was exported from Phabricator. Differential Revision: D62198560 |
|
This pull request has been merged in 2027a14. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Summary:
This PR fixes a bunch of issues in the eval pipeline:
eot_token_idbosandeosduringtok_encodebased on this discussion.executorch/examples/models/llama2/tokenizer/tiktoken.pyto be synced with llama 3.1's official code. Majorly updated set of special tokens.Differential Revision: D62198560