ModelArtifactConfig creation refactor #206
Merged
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
sunggg
reviewed
Feb 13, 2024
| # ignore_eos=request.ignore_eos, | ||
| # use_beam_search=request.use_beam_search, | ||
| ) | ||
| sampling_params = SamplingParams() |
There was a problem hiding this comment.
Can we simply do
sampling_params = SamplingParams(
presence_penalty=request.presence_penalty,
....
)
Although we need to handle checks like if request.presence_penalty is not None:, I think this way we can make sure that these new values can go through the __post_init__.
And I think sampling_params.verify() can also stay within __post_init__ if this works.
There was a problem hiding this comment.
I tried that approach first, but I didn't want to repeat the default values specified in sampling_params.py here.
sunggg
approved these changes
Feb 13, 2024
Lunderberg
pushed a commit
to Lunderberg/mlc-llm
that referenced
this pull request
Feb 27, 2024
This PR brings the database merge feature, which automatically merges all database under the specified `--db-path` (which is `./log_db/` by default). For example, when there are two database under `./log_db/`: `./log_db/vicuna` and `./log_db/redpajama`, we won't need to manually specify one database to use. Instead, `build.py` will now load each of the database and merge them into one in memory. We are also backward compatible. Say if we have only one database which is `./log_db` itself, we are still able to load it in.
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Outdated suggestions cannot be applied.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.
As of #192,
StagingEngineneeds to havemodel_artifact_config, which is created in the worker process. Creating the config in the constructor again, viaget_model_artifact_config, is problematic for PT models since there is no build artifact for PT models (henceget_model_artifact_configcannot be used).This PR does a refactor around how
ModelArtifactConfigis constructed. Rather than creating it insidePagedCacheModel, we are now creating it at the beginning of the engine initialization. So the created config can be passed to all ofStagingEngine,StagingEngineWorker, andSyncEngine.Note that the same update needs to be applied to ollm in the next revision bump.
It also fixes the issue of
SamplingParamsnot being verified correctly due to how it is initialized.@sunggg @yelite