test+docs: gate a deeper [2,5,3,1] net learns; document the without-Vivado met-timing refinement (Refs #1764) - #1912
Closed
gHashTag wants to merge 2 commits into
Closed
Conversation
…o met-timing refinement (Refs #1764) Adds a deeper [2,5,3,1] (3-layer, 158-step) learning assertion to the generator self-test (held-out 59/60, deterministic), extending the CI-gated "the method scales beyond XOR" claim. Also documents a measured partial open-flow refinement for "without Vivado": constraining the internal clk net to a met period (create_clock 50ns on [get_nets clk]) and building without --timing-allow-fail closes timing deterministically -- several seeds come up bit-exact at ep0 and seed-search finds a fully-stable build, but it does not kill the lottery and more settle stays non-monotonic. So met-timing + seed-search is the best open recipe; deterministic no-seed-search still needs commercial P&R. Free Vivado supports xc7a200t but is x86-only. 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
… (Refs #1764) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Contributor
PR DashboardGenerated at: 2026-08-08 03:46:25 UTC
Summary
Seal Status
|
Contributor
|
📓 NotebookLM Notebook linked to this PR
This notebook contains session context, decisions, and artifacts for this work. |
Owner
Author
|
Superseded by the rebased branch (NOW.md conflict with #1911 resolved); no force-push. |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Stack (V2): adds a deeper
[2,5,3,1](3-layer, 158-step) learning assertion to the generator self-test — held-out 59/60 (98%), deterministic init — so the CI learning gate now enforces generalization on a deeper and wider net, extending the "the method scales beyond XOR" claim.Answering "can we do it without Vivado?": documents a measured partial open-flow refinement. Constrain the internal clock net to a period nextpnr actually meets (
create_clock -period 50 [get_nets clk], ~20 MHz < the ~21 MHz fmax) and build without--timing-allow-fail, so P&R closes timing deterministically. Result: several seeds come up bit-exact at ep0 (vs glitching from the start), and seed-search still finds a fully-stable build (a met-timing seed trained XOR 4/4 through all 60 epochs). It does not eliminate the lottery (some seeds diverge over epochs) and moresettlestays non-monotonic — so met-timing + seed-search is the best open recipe; deterministic no-seed-search closure still needs commercial P&R. Free Vivado (ML Standard) supports the xc7a200t but is x86 Linux/Windows only, not the macOS arm64 dev host.Refs #1764