CUDA: fix logic for clearing padding with -ngl 0 #13320
Merged
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
github-actions
bot
added
Nvidia GPU
slaren
reviewed
May 5, 2025
ggml/src/ggml-cuda/mmq.cu
Outdated
| const size_t size_data = ggml_nbytes(src0); | ||
| const size_t size_alloc = ggml_backend_buffer_get_alloc_size(src0->buffer, src0); | ||
| if (size_alloc > size_data) { | ||
| CUDA_CHECK(cudaMemsetAsync((char *) src0->data + size_data, 0, size_alloc - size_data)); |
There was a problem hiding this comment.
This is missing the stream parameter.
ggml/src/ggml-cuda/mmvq.cu
Outdated
| const size_t size_data = ggml_nbytes(src0); | ||
| const size_t size_alloc = ggml_backend_buffer_get_alloc_size(src0->buffer, src0); | ||
| if (size_alloc > size_data) { | ||
| CUDA_CHECK(cudaMemsetAsync((char *) src0->data + size_data, 0, size_alloc - size_data)); |
JohannesGaessler
force-pushed
the
cuda-fix-deepseek-partial-offload
branch
from
4634789 to
108dfde
Compare
JohannesGaessler
force-pushed
the
cuda-fix-deepseek-partial-offload
branch
from
108dfde to
ac78a42
Compare
slaren
approved these changes
May 5, 2025
|
@JohannesGaessler |
gabe-l-hart
added a commit
to gabe-l-hart/llama.cpp
that referenced
this pull request
May 6, 2025
* origin/master: (27 commits) llama : fix build_ffn without gate (ggml-org#13336) CUDA: fix bad asserts for partial offload (ggml-org#13337) convert : qwen2/3moe : set yarn metadata if present (ggml-org#13331) CUDA: fix --split-mode row for MMQ (ggml-org#13323) gguf-py : avoid requiring pyside6 for other scripts (ggml-org#13036) CUDA: fix logic for clearing padding with -ngl 0 (ggml-org#13320) sampling : Integrate Top-nσ into main sampling chain (and add it to the server) (ggml-org#13264) server : Webui - change setText command from parent window to also send the message. (ggml-org#13309) mtmd : rename llava directory to mtmd (ggml-org#13311) clip : fix confused naming ffn_up and ffn_down (ggml-org#13290) convert : bailingmoe : set yarn metadata if present (ggml-org#13312) SYCL: Disable mul_mat kernels for noncontiguous tensor b (ggml-org#13308) mtmd : add C public API (ggml-org#13184) rpc : use backend registry, support dl backends (ggml-org#13304) ggml : activate s390x simd for Q3_K (ggml-org#13301) llava/mtmd : fixes to fully support dl backends (ggml-org#13303) llama : build windows releases with dl backends (ggml-org#13220) CUDA: fix race condition in MMQ stream-k fixup (ggml-org#13299) CUDA: fix race condition in MMQ ids_dst (ggml-org#13294) vulkan: Additional type support for unary, binary, and copy (ggml-org#13266) ...
Nexesenex
added a commit
to Nexesenex/croco.cpp
that referenced
this pull request
May 9, 2025
)" This reverts commit 9070365.
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.
Fixes #13305 .
The problem is that in
ggml_cuda_op_mul_matthe padding of temporary compute buffers is cleared just prior to kernel execution and when I wrote code to directly launch MMVQ and MMQ I forgot to add the corresponding functionality; this PR simply adds it. The changes in MMVQ should not actually be invoked unless the minimum number of tokens for evaluating the model on the GPU is lowered.Prior to #13199 the logic for clearing the padding was already wrong, but in a different way. Instead of not clearing padding when it ought to be, because slices of
src0were passed toggml_cuda_op_mul_matthat would cause valid tensor data to be cleared. I did not touch that logic, now thatMUL_MAT_IDcan be handled in a single kernel launch it should never be invoked (I also added an assert).