Eagle3 - update docs, enforce limitations #3939
Open
+46
−17
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
Contributor
There was a problem hiding this comment.
Pull request overview
This PR adds Eagle3 support to the codebase by introducing sequential processing enforcement and configuration options. Eagle3 is a speculative decoding variant that requires stricter limitations than standard speculative decoding, including forced greedy sampling and single-request processing.
Changes:
- Added Eagle3-specific configuration fields and mutex-based sequential processing enforcement
- Created new
EAGLE3decoding method with enforced greedy sampling (disabling random sampling and beam search) - Updated documentation to clarify Eagle3 limitations and enforcement mechanisms
Reviewed changes
Copilot reviewed 9 out of 9 changed files in this pull request and generated 7 comments.
Show a summary per file
| File | Description |
|---|---|
| src/llm/servable.hpp | Added mutex and lock fields to enforce sequential processing for Eagle3 |
| src/llm/servable.cpp | Added Eagle3 mode detection and decoding method assignment |
| src/llm/llm_calculator.proto | Added draft_eagle3_mode configuration option and renumbered subsequent fields |
| src/llm/language_model/continuous_batching/servable_initializer.cpp | Set eagle3Mode property from node options |
| src/llm/io_processing/base_generation_config_builder.hpp | Added EAGLE3 enum value and documentation |
| src/llm/io_processing/base_generation_config_builder.cpp | Implemented Eagle3 configuration enforcement (greedy sampling only) |
| src/llm/http_llm_calculator.cc | Added lock acquisition/release logic for Eagle3 sequential processing |
| demos/continuous_batching/speculative_decoding/README.md | Updated documentation with enforcement details |
| demos/common/export_models/export_model.py | Renamed flag and updated template to enforce max_num_seqs=1 for Eagle3 |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
mzegla
requested review from
dkalinowski,
dtrawins,
michalkulakowski and
ngrozae
dkalinowski
reviewed
Jan 30, 2026
|
|
||
| namespace ovms { | ||
|
|
||
| // TODO: Monitor Eagle3 sampling support in GenAI and update this when Eagle3 supports more sampling strategies. |
Collaborator
Author
There was a problem hiding this comment.
Does not exist. As far as I know different sampling degrades performance, so no ticket has been created to support it.
dkalinowski
reviewed
Jan 30, 2026
mzegla
commented
Jan 30, 2026
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.
No description provided.