refactor: Speculative decoding buffers part 2 #5316
Conversation
Funatiq
changed the title
|
/bot run |
|
PR_Github #9356 [ run ] triggered by Bot |
|
PR_Github #9356 [ run ] completed with state |
Funatiq
force-pushed
the
dev/refactor_spec_decode_buffers_2
branch
from
ffee1b4 to
702261d
Compare
|
/bot run --only-multi-gpu-test |
|
PR_Github #9512 [ run ] triggered by Bot |
|
PR_Github #9512 [ run ] completed with state |
|
/bot run --only-multi-gpu-test --disable-fail-fast |
|
PR_Github #9541 [ run ] triggered by Bot |
|
PR_Github #9541 [ run ] completed with state |
|
/bot run |
Funatiq
requested review from
schetlur-nv and
dcampora
and removed request for
schetlur-nv
|
PR_Github #9550 [ run ] triggered by Bot |
|
PR_Github #9550 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #9556 [ run ] triggered by Bot |
|
PR_Github #9556 [ run ] completed with state |
|
/bot run |
|
PR_Github #9571 [ run ] triggered by Bot |
|
PR_Github #9571 [ run ] completed with state |
Funatiq
force-pushed
the
dev/refactor_spec_decode_buffers_2
branch
from
702261d to
2fe91ba
Compare
|
/bot run |
|
PR_Github #9659 [ run ] triggered by Bot |
|
PR_Github #9659 [ run ] completed with state |
|
PR_Github #9700 [ run ] triggered by Bot |
|
PR_Github #9700 [ run ] completed with state |
Funatiq
force-pushed
the
dev/refactor_spec_decode_buffers_2
branch
from
3071315 to
6793077
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #9720 [ run ] triggered by Bot |
|
PR_Github #9720 [ run ] completed with state |
|
/bot run |
|
PR_Github #9742 [ run ] triggered by Bot |
|
PR_Github #9742 [ run ] completed with state |
Signed-off-by: Robin Kobus <[email protected]>
- Introduced `setupMedusaLogits` method in `DecoderInputBuffers` to initialize logits for speculative decoding. - Updated `HandleContextLogits`, `HandleGenerationLogits` and `MakeDecodingBatchInputOutput` to utilize the new Medusa logits structure. - Removed draft logits handling from `DraftBuffers`, streamlining the management of logits. Signed-off-by: Robin Kobus <[email protected]>
- Removed indirect references to device buffers in favor of methods from `DecoderState`, streamlining buffer management. - Updated `DecoderStepAsyncSend` to utilize `DecoderState` for managing draft token buffers, enhancing clarity and maintainability. - Adjusted related functions and Python bindings to reflect these changes, ensuring compatibility with existing interfaces. These modifications improve the organization of the decoding process within the batch manager. Signed-off-by: Robin Kobus <[email protected]>
- Introduced `setupSpeculativeDecoding` method in `DecoderOutputBuffers` to streamline speculative decoding setup. - Removed `DraftBuffers` from `DecoderBuffers`, simplifying buffer management and enhancing clarity. - Updated `UpdateDecoderBuffers` to reflect changes in buffer handling, ensuring compatibility with the new structure. - Adjusted `TrtGptModelInflightBatching` to utilize the updated `DecoderOutputBuffers`. These changes improve the organization and maintainability of the decoding process within the batch manager. Signed-off-by: Robin Kobus <[email protected]>
Funatiq
force-pushed
the
dev/refactor_spec_decode_buffers_2
branch
from
6793077 to
98f99f2
Compare
|
/bot run |
|
PR_Github #9822 [ run ] triggered by Bot |
|
PR_Github #9822 [ run ] completed with state |
Signed-off-by: Robin Kobus <[email protected]>
Signed-off-by: Robin Kobus <[email protected]>
Signed-off-by: Robin Kobus <[email protected]>
Signed-off-by: Robin Kobus <[email protected]>
Signed-off-by: Robin Kobus <[email protected]>
Signed-off-by: Robin Kobus <[email protected]>
Signed-off-by: Robin Kobus <[email protected]>
Signed-off-by: Robin Kobus <[email protected]>
Signed-off-by: Robin Kobus <[email protected]>
Description
Remove
DraftBuffersclass and integrate components intoDecoderInputBuffersandDecoderOutputBuffers.Please see commit messages for details.
Test Coverage
GitHub Bot Help
/bot [-h] ['run', 'kill', 'skip', 'reuse-pipeline'] ...Provide a user friendly way for developers to interact with a Jenkins server.
Run
/bot [-h|--help]to print this help message.See details below for each supported subcommand.
run [--disable-fail-fast --skip-test --stage-list "A10-1, xxx" --gpu-type "A30, H100_PCIe" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-[Post-Merge]-1, xxx"]Launch build/test pipelines. All previously running jobs will be killed.
--disable-fail-fast(OPTIONAL) : Disable fail fast on build/tests/infra failures.--skip-test(OPTIONAL) : Skip all test stages, but still run build stages, package stages and sanity check stages. Note: Does NOT update GitHub check status.--stage-list "A10-1, xxx"(OPTIONAL) : Only run the specified test stages. Examples: "A10-1, xxx". Note: Does NOT update GitHub check status.--gpu-type "A30, H100_PCIe"(OPTIONAL) : Only run the test stages on the specified GPU types. Examples: "A30, H100_PCIe". Note: Does NOT update GitHub check status.--only-multi-gpu-test(OPTIONAL) : Only run the multi-GPU tests. Note: Does NOT update GitHub check status.--disable-multi-gpu-test(OPTIONAL) : Disable the multi-GPU tests. Note: Does NOT update GitHub check status.--add-multi-gpu-test(OPTIONAL) : Force run the multi-GPU tests. Will also run L0 pre-merge pipeline.--post-merge(OPTIONAL) : Run the L0 post-merge pipeline instead of the ordinary L0 pre-merge pipeline.--extra-stage "H100_PCIe-[Post-Merge]-1, xxx"(OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-[Post-Merge]-1, xxx".For guidance on mapping tests to stage names, see
docs/source/reference/ci-overview.md.kill
killKill all running builds associated with pull request.
skip
skip --comment COMMENTSkip testing for latest commit on pull request.
--comment "Reason for skipping build/test"is required. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.reuse-pipeline
reuse-pipelineReuse a previous pipeline to validate current commit. This action will also kill all currently running builds associated with the pull request. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.