Conversation
Serialize MOLT models, compute transform norms through rank-space Gram matrices, and make the trained layer and pre-activation penalty configurable. Add architecture-aware activation extraction, MOLT replacement evaluation, atomic timed checkpoints, and reliable data-generator teardown.
Record RTX 5090 sweeps, Gemma controls, and Qwen sparsity and rank variants. Queue new training only after the predecessor checkpoint completes, with process-identity checks and failed-launch cleanup.
Compare analytic MOLT Jacobians with autograd Jacobians using per-token flattened cosine similarity. Add evaluation tooling, regression tests, and saved Gemma layer-22 measurements.
Collect held-out gates and context-dependent logit projections, render feature pages, and serve token-level interventions with a conversation builder. Handle inference contention and cancellation, add Python and browser coverage, and default to the L0≈11 sparsity3e-5 checkpoint.
Preserve investigation scripts, numerical and causal checks, steering sweeps, reconstruction ablations, and rendered reports for the L0≈3 checkpoint. Include matched-prefix follow-up measurements and rank-cap diagnostics; omit runtime logs and tensor caches.
Add single-layer activation collection, training schedules, graceful worker cleanup, and resumable Qwen campaigns. Add dashboard explanation and scoring pipelines, layer explorers, sharing configuration, documentation, and regression tests. Ignore generated campaign state and make browser checks independent of live scoring progress. Validation: 72 Python tests and all three dashboard browser checks passed.
The original remote MOLT implementation was adapted locally in 82048f3 and developed further. Join its history while keeping the current local source tree unchanged, allowing a normal push to MOLT.
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.
No description provided.