feat: GF-T 2-layer XOR trainer (fixed hidden + trainable output, XOR 4/4 on-chip via split) - #1826
Merged
Conversation
gft_xortrain.t27: on-chip SGD step of a 2-layer XOR net -- FIXED analytic hidden layer h0=relu(x0+x1), h1=relu(x0+x1-1) + TRAINABLE output z=v0*h0+v1*h1 with hard-sigmoid + p-y gradient; v_j' = v_j - eta*(p-y)*h_j; returns (v0'<<32)|v1'. Proven on a live AX7203 via the SPLIT pattern: the trainable output layer learns on the proven gft_logistic bitstream (stream host-computed hidden features) -> XOR 4/4, v -> ~[1,-2]. As one design (fasm 22.6M) it exceeds the openXC7 correctness ceiling (2nd confirmation) -- split big models: fixed part off-chip, trained part on a sub-ceiling bitstream. In-spec test PASS; GF-T Python sim converges 4/4 by epoch 9. docs/NOW.md updated. 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
New spec gft_xortrain.t27 — on-chip SGD step of a 2-layer XOR net: FIXED analytic hidden layer + TRAINABLE output (hard-sigmoid, p-y gradient). In-spec test PASS; GF-T Python sim converges 4/4. Proven on a live AX7203 via the SPLIT pattern (trainable output learns on the proven gft_logistic bitstream from host-computed hidden features -> XOR 4/4, v->~[1,-2]). As one design (22.6M) it exceeds the correctness ceiling (2nd confirmation) — documented. docs/NOW.md updated. Refs #1764