Attempting to improve performance #10
Merged
+396
−39
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
- Removed the forced CPU backend assignment of kvaware_indices
- src/llama-sparse-topk.cpp: deleted the block that moved result to
backend_cpu. Now it stays where it’s produced.
- src/llama-model.cpp: removed both instances of
ggml_backend_sched_set_tensor_backend(sched, kvaware_indices,
backend_cpu) so we don’t bounce indices to host in MLA and MHA sparse
paths.
- Gate debug-only float32 cast of indices:
- src/llama-sparse-topk.cpp: only cast to F32 and log the f32 indices
when LLAMA_SPARSE_DEBUG is set. This cuts extra nodes/copies in
normal runs.
- Increase default Top-K token tile size:
- src/llama-sparse-topk.cpp: default TILE_T from 32 to 128, still
overridable via LLAMA_SPARSE_TOPK_TILE_T.
branches, so we avoid the extra backend hop to CPU after apply_sparse_attention_kvaware
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Outdated suggestions cannot be applied.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.
No description provided.