vulkan: add Vulkan backend for GGML_OP_GATED_DELTA_NET#2
Closed
ProgenyAlpha wants to merge 1 commit into
Closed
Conversation
Adds compute shader and full ggml-vulkan.cpp wiring for the gated delta net recurrence op. Supports multi-token sequences, GQA stride-based broadcast, and inline sigmoid/exp transforms. Tested 7/7 on NVIDIA RTX 3080 Ti via test-backend-ops: - head_size 32/64/128, single/multi token, single/multi seq, GQA, permuted
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds Vulkan compute shader and ggml-vulkan.cpp wiring for the gated delta net recurrence op.
Changes
gated_delta_net.comp— new compute shader with dynamic head size, multi-token loop, GQA stride-based broadcast, inlineexp(g)andsigmoid(beta)transformsggml-vulkan.cpp— push constants, pipeline creation, dispatch function, build graph case, supports_opvulkan-shaders-gen.cpp— shader registrationTest Results
7/7 test-backend-ops cases pass on NVIDIA RTX 3080 Ti:
head_size=128, n_seq_tokens=1, n_seqs=1, v_repeat=1, permuted=0OKhead_size=64, n_seq_tokens=1, n_seqs=2, v_repeat=1, permuted=0OKhead_size=64, n_seq_tokens=4, n_seqs=1, v_repeat=1, permuted=0OKhead_size=64, n_seq_tokens=4, n_seqs=2, v_repeat=1, permuted=0OKhead_size=32, n_seq_tokens=4, n_seqs=2, v_repeat=2, permuted=0OKhead_size=64, n_seq_tokens=4, n_seqs=2, v_repeat=1, permuted=1OKhead_size=64, n_seq_tokens=4, n_seqs=1, v_repeat=1, permuted=1OK