feat: GF-T 2-neuron hidden layer on-chip trainer (proven on AX7203) - #1820
Merged
Conversation
gft_hidden2.t27: on-chip SGD step of a 2-neuron hidden layer y=relu(w0*x0)+relu(w1*x1) with fixed unit output weights. Two INDEPENDENT nonlinear units, each with its own relu gate; per-neuron gated gradient dw_j=e*relu'(z_j)*x_j; updates w_j-eta*dw_j; returns (w0'<<32)|w1'. Reuses the verified smul/sadd/neg/relu/relu_prime helpers. Proven on a live AX7203 (uart_hidden2.v): the two units gate INDEPENDENTLY -- killing one neuron's activation (its input <0) freezes only that neuron's weight while the other keeps learning (MOVED vs FROZEN across 3 phases). In-spec tests (both/onedead) PASS. 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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_hidden2.t27 — on-chip SGD step of a 2-neuron hidden layer y=relu(w0x0)+relu(w1x1) (fixed unit output weights). Two INDEPENDENT nonlinear units, each with its own relu gate; per-neuron gated gradient dw_j=e*relu'(z_j)x_j; updates w_j-etadw_j; returns (w0'<<32)|w1'. In-spec tests (both/onedead) PASS. Proven on a live AX7203 (uart_hidden2.v): the units gate INDEPENDENTLY — killing one neuron's activation freezes only its weight while the other keeps learning (MOVED vs FROZEN across 3 phases). docs/NOW.md updated. Refs #1764