test(ci): gate that the generated trainer LEARNS -- XOR 4/4 + nonlinear held-out >=90%, incl. deep 3-layer (Refs #1764) - #1909
Merged
Conversation
… bit-exact (Refs #1764) The bit-exact gates prove generated RTL == model, but nothing CI-enforced that the algorithm actually learns non-trivial tasks. The generator's self-tests already assert it (XOR 4/4; (2,4,1) held-out 58/60; (2,4,2) argmax 56/60; deep [2,4,3,1] 3-layer 59/60, all >=90%) but only ran by hand. Added `python3 tools/gft_backprop_microcode.py` to emit-bitexact-gate.yml so the learning + generalization claims are enforced per PR. Strengthens the thesis that the method learns and scales beyond XOR in the model -- the only limit is the open silicon flow's placement marginality (now addressed by the Vivado closure kit), not the algorithm. Refs #1764 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Contributor
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The bit-exact gates prove the generated RTL equals the model, but nothing CI-enforced that the algorithm actually learns non-trivial tasks. The generator's self-tests already assert exactly that — but they only ran when the file was executed by hand.
Adds
python3 tools/gft_backprop_microcode.pyas a CI step inemit-bitexact-gate.yml, enforcing per pull request:(2,4,1)on a noisy nonlinear dataset, held-out 58/60;(2,4,2)multi-output one-hot classifier, held-out 56/60;[2,4,3,1](3-layer, 2 hidden) nonlinear task, held-out 59/60 — all ≥ 90%.This strengthens the project's thesis: the method learns and scales beyond XOR (multi-layer, held-out generalization) in the model — the only limit was the open silicon flow's placement marginality, now addressed by the Vivado closure kit, not the algorithm.
CI-config only; no code change. Refs #1764