docs: add DSpark backbone and long-context recipes - #740
Closed
Boreas618 wants to merge 3 commits into
Closed
Conversation
Contributor
|
Caution The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased. |
Sizing rules validated by the Inkling-Small v2 program: target-matched GQA geometry, uniform target taps, FFN width over depth, serve-ability constraints on the attention pattern, train-block = serve-block with block-size-adapted loss decay, and the constant-LR schedule.
Native-context stage 1 + YaRN weights-only warm-start stage 2, why the optimizer state must be dropped, the transformers-5.x dual-rope-schema checkpoint pitfall, and the bucketed-accept-length validation protocol.
Boreas618
force-pushed
the
agent/dspark-recipe-docs
branch
from
August 1, 2026 17:32
02087ad to
1e21895
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What changed
Why
The Inkling-Small v2 program exposed repeatable design and operational rules that are useful when adapting DSpark/DFlash to a new target. These rules were previously captured only in an old development branch.
Impact
Users get concrete, reviewable guidance for choosing a draft architecture and adapting it to long contexts without training every stage at the maximum sequence length.
Validation
git diff --check upstream/main...HEADmake -C docs html SPHINXOPTS='-W --keep-going'reads and renders both new pages successfully; the command exits nonzero on the pre-existingdocs/basic_usage/training.md:19header-level warning.