fix(generation): fp4 byte-size and stale v1→v2 MLA warning#264
Open
Gpgabriel25 wants to merge 1 commit into
Open
fix(generation): fp4 byte-size and stale v1→v2 MLA warning#264Gpgabriel25 wants to merge 1 commit into
Gpgabriel25 wants to merge 1 commit into
Conversation
718ce5c to
806f8ee
Compare
Owner
|
Hi @Gpgabriel25 and thanks for contributing to easydel All of ur PRs are solid and good but it will take time to merge them since i want to keep git worktree the same before i merge branch 'vnext' |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Changes
fp4-safe byte-size calculation
_create_mixed_standard_ragged_page_cache_configsand_create_mixed_standard_unified_attention_cache_configsboth usedjnp.finfo(kvdtype).bits // 8— the same bug fixed in the cachinglayer (see PR #XX). Replaced with
jnp.dtype(kvdtype).itemsizesofp4 / integer KV dtypes work correctly through the generation mixin too.
Correct
v1→v2in mixed-MLA log warningThe warning that fires when mixed MLA/non-MLA layers are detected was
logging
multi_latent_ragged_page_attention_v1as the target mechanism.The actual target is
v2. Fixed the string literal.Tests
New:
test_esurge_compatible_model_logs_v2_for_mixed_mla— verifiesthat both the emitted warning and the resulting
attn_mechanism/decode_attn_mechanism/mla_attn_mechanismkwargs all referencemulti_latent_ragged_page_attention_v2. 1 passed.