Public slim logits parity from m1-demo-v1 - #2
Merged
Conversation
The m1-demo-v1 model_slim.pt matches this GPT (6/6/384/256, tied embed, 13,891,584 params) and can drive a forward. It is inference-only: no published PyTorch grads, so backward parity is not claimed. Architecture and schema gates always run; the 55 MB file is not committed and CI must not download it. Co-authored-by: RAFAELDCOELHO <RAFAELDCOELHO@users.noreply.github.com>
Freeze a 16-token forward from m1-demo-v1 model_slim.pt (PyTorch manual attn, float64) against this engine. Measured relative error 2.123856e-15. No private checkpoint, no optimizer/AdamW claim. CI runs without downloading the 55 MB weights. Co-authored-by: RAFAELDCOELHO <RAFAELDCOELHO@users.noreply.github.com>
cursor
Bot
changed the title
RAFAELDCOELHO
marked this pull request as ready for review
September 1, 2026 22:33
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Public proof from
m1-demo-v1model_slim.ptonly. No private checkpoint. No invented numbers.Forward / logits (allowed)
scripts/gen_slim_logits_fixture.pyloads the public slim, runs PersonaCore PyTorch (attn_impl=manual, float64) and this engine on the documented 16-token input[8184, 1, …, 15](not privateval.bin).fixtures/slim_logits_public.npz(~1.9 MB). Not the 55 MB.pt.1e-11bound /1e-12canary.M4 / AdamW (not allowed — honest skip)
{schema_version, model, model_config, git_sha, step, val_loss}. No optimizer.test_m4_adamw_from_public_slim_is_not_claimedis an explicit skip. Do not invent that proof.CI
.github/workflows/ci.ymlrunspytestwith no slim download.v1.0 proof table in the README is unchanged. LICENSE is not on this branch (it lives on PR #1).