Intern-Decision: int8 Showdown snapshot - #22
Merged
Merged
Conversation
ensure(.showdown) now pins the int8 buckets of FluidInference/intern-decision-0.8b-showdown-coreml (revision 878e1404): same 90 ms per decision as fp16, about 0.85 GB in memory with one bucket in use instead of about 1.5 GB, 1.9 GB to download instead of 3.5 GB, 15-0 vs 14-1 against poke-env's max-power player over 15 battles. A re-pinned snapshot also removes packages and compiled models the new asset list no longer names, so an older precision left in the cache folder cannot be preferred by the loader. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
InternDecisionModelStore.ensure(.showdown)now pins the int8 buckets of FluidInference/intern-decision-0.8b-showdown-coreml at 878e1404.Measured on the Swift runtime, one bucket in use, 60 calls: 374 MB process footprint + 480 MB mapped weights (about 0.85 GB) vs 529 MB + 955 MB for fp16 (about 1.5 GB); p50 unchanged at 90 ms. In the harness, 15-0 vs poke-env's max-power player against 14-1 for fp16 (15 battles each). Download 1.9 GB instead of 3.5 GB.
A re-pinned snapshot also removes packages and
.mlmodelcfolders the new asset list no longer names, so an older precision left in the cache folder cannot be preferred by the loader.🤖 Generated with Claude Code