Commit Graph

Select branches

Hide Pull Requests

master

Mono Color

• b534be9fad commit stale changes master jens 2026-07-26 21:23:50 +02:00
• 74371975ad [JayRnn] - reduced VOCAB - changed test text jens 2026-06-07 17:49:05 +02:00
• 0febd4e6b6 Rbm::Train(cd_bb): added do_gibbs_sample_hidden jens 2026-06-07 17:48:22 +02:00
• 2d3361d5ce rework forward_step to use char-by-char state priming with SPACE token jens 2026-06-07 13:34:02 +02:00
• 51274ee1ae [jayRnn] - adjusted params jens 2026-06-06 12:49:09 +02:00
• 588f58883f - no "^" in vocabular - JayRnn: wrong but stable jens 2026-06-06 12:35:20 +02:00
• a41ddce6a8 scale up model, expand vocab, collect predicted text in forward pass jensandClaude Sonnet 4.6 2026-06-05 18:17:41 +02:00
• 5ac4608ba8 tune hyperparams and fix forward_step shift, re-enable training jensandClaude Sonnet 4.6 2026-06-05 17:47:30 +02:00
• f7ed8563d7 refactor RnnModel: multi-unit delay training and unified vc interface jensandClaude Sonnet 4.6 2026-06-05 16:29:36 +02:00
• b20ea4edb6 fixed wrong number of bits for binary encoding jens 2026-06-05 11:11:32 +02:00
• 49a99dc750 tune label fitter: epochs, momentum, encoding, decode threshold jensandClaude Sonnet 4.6 2026-06-05 10:03:55 +02:00
• 248ae4f199 tune hyperparams, add README, simplify training loop jensandClaude Sonnet 4.6 2026-06-05 09:34:33 +02:00
• b5a128fb61 fixed error for other WIN sizes jens 2026-06-04 23:19:18 +02:00
• 06ab8d0166 refactored forward_step jens 2026-06-04 19:30:58 +02:00
• 7f6818670e - changed sample text - changed model params jens 2026-06-04 16:43:41 +02:00
• 62dd450c9c increased difficulty jens 2026-06-04 11:15:17 +02:00
• 4bd516798a cleaned up jens 2026-06-04 10:36:50 +02:00
• bb6680c931 getting Rnn to work jens 2026-06-04 10:28:19 +02:00
• 851bad28d3 cleaned up jens 2026-06-03 23:01:10 +02:00
• 3a41014d50 improved context update for JayRnn jens 2026-06-03 22:58:18 +02:00
• 2701daf1c4 added JayRnn jens 2026-06-03 22:14:40 +02:00
• 6a75e0b33b added more Rnn helber jens 2026-06-03 22:14:34 +02:00
• 5c3734a2a4 extend model member detection by list[Entity] jens 2026-06-03 22:13:57 +02:00
• 27460c503d fixed type alias jens 2026-06-03 22:13:26 +02:00
• c62605de51 [update] README.rbm.md: add Q&A on image/video and code generation architectures jensandClaude Sonnet 4.6 2026-06-03 11:15:10 +02:00
• ae81e3e340 [add] README.rbm.md: Q&A on RBMs, VAEs, Transformers, and autodiff jensandClaude Sonnet 4.6 2026-06-03 11:05:48 +02:00
• df4a9b2ac2 [add] Rnn.py: unrolled character-level RNN-RBM matching docs/Rnn.drawio.png jensandClaude Sonnet 4.6 2026-06-03 10:49:46 +02:00
• 3edc124a5c [refactor] move tests to tests/, add pytest functions and main() entrypoints jensandClaude Sonnet 4.6 2026-06-02 21:32:58 +02:00
• 829547d271 [notebook] update cifar notebook outputs and add N_HID constant jensandClaude Sonnet 4.6 2026-06-02 21:06:41 +02:00
• 0b4794de4e [fix] use np.asnumpy in SubImageCombine to support CuPy arrays jensandClaude Sonnet 4.6 2026-06-02 20:57:58 +02:00
• 1954064a54 [refactor] move rbm.model → model.model jensandClaude Sonnet 4.6 2026-06-02 08:55:16 +02:00
• b41274c7a6 [refactor] move rbm.torch → compat.torch jensandClaude Sonnet 4.6 2026-06-02 08:36:21 +02:00
• 3d235b4214 [refactor] move rbm.label → label.label jensandClaude Sonnet 4.6 2026-06-02 08:30:26 +02:00
• 807de19fa7 [fix] restore SubImageCombine name in cifar notebook (clobbered by bulk rename) jensandClaude Sonnet 4.6 2026-06-02 08:23:08 +02:00
• a07698497d [refactor] SubImage → SubImageExtract, add SubImageCombine jensandClaude Sonnet 4.6 2026-06-02 08:20:38 +02:00
• 2da2262368 [refactor] move rbm.stack* + rbm.rnn_helper → stack/ jensandClaude Sonnet 4.6 2026-06-02 08:10:03 +02:00
• b0d370a56b [refactor] move rbm.image → image.sub_image jensandClaude Sonnet 4.6 2026-06-02 08:01:51 +02:00
• 599ba4f6d5 [cifar_sub_image] run with STRIDE=6 (25 patches/image, ~33% overlap) jensandClaude Sonnet 4.6 2026-06-01 20:07:10 +02:00
• dddb937909 [cifar_sub_image] add full 32x32 reconstruction from sub-images with overlap blending jensandClaude Sonnet 4.6 2026-06-01 19:59:30 +02:00
• 00993e29f1 [rnn_helper] add shift functions, ch2idx/idx2ch, and tests jensandClaude Sonnet 4.6 2026-06-01 16:43:33 +02:00
• cce3943b27 [RNN-RBM] rename to RNN-RBM, add unrolled depth-2 mode, update README jensandClaude Sonnet 4.6 2026-05-31 23:22:00 +02:00
• 73ff6c2bb4 [context33] - momentum 0.9, epochs 500 jensandClaude Sonnet 4.6 2026-05-31 20:47:27 +02:00
• 4a915f165a [context33] - increase momentum 0.5→0.7; run outputs 99% recon, 88% next-step jensandClaude Sonnet 4.6 2026-05-31 20:06:20 +02:00
• 586803856f [context33] - add batch tiling (N_REPEAT=500) for GPU utilisation jensandClaude Sonnet 4.6 2026-05-31 20:03:03 +02:00
• bbc03d37e1 [context33] - 200-char sentence, auto-uppercase, adaptive tick spacing jensandClaude Sonnet 4.6 2026-05-31 16:15:01 +02:00
• 0beb7a5c1b [context33] - replace armadillo data with 20-char sentence, moby encoding jensandClaude Sonnet 4.6 2026-05-31 15:35:59 +02:00
• 5506fef0fa [context33] - fix row order; sequence now reads 1 5 6 9 … R U X jensandClaude Sonnet 4.6 2026-05-31 15:22:43 +02:00
• 2a6b00db88 [context33] - pyRBM port of context33.prj; predict next char from current context jensandClaude Sonnet 4.6 2026-05-31 15:12:14 +02:00
• 952045b9f1 [StackRnn] - temporal-shift padding, joint training, vocab reduction, TEMPORAL_DEPTH=64 jensandClaude Sonnet 4.6 2026-05-31 14:37:08 +02:00
• 3698c85ed4 [StackRnn] - add unrolled (own-weights) mode; each position gets its own RBM jensandClaude Sonnet 4.6 2026-05-31 11:52:01 +02:00
• 0b302a9c12 [moby_rnn] - add README with architecture ASCII art; consolidate constants jensandClaude Sonnet 4.6 2026-05-31 11:42:55 +02:00
• eb1f96f4ec [StackRnn] - add recurrent RBM stack with character-level LM notebook jensandClaude Sonnet 4.6 2026-05-31 11:27:07 +02:00
• 17429ddce8 [tests] - fix CuPy display in test_learn_encoded_labels; enable model.load in test_linear jensandClaude Sonnet 4.6 2026-05-30 23:45:42 +02:00
• 37c6e33948 [CheckpointStatus] - add checkpoint status handler; Model.train() accepts status param jensandClaude Sonnet 4.6 2026-05-30 22:42:01 +02:00
• 2f9fe1b66b [bugfix] - support large CPU-side datasets; fix numpy/CuPy interop jensandClaude Sonnet 4.6 2026-05-30 22:26:06 +02:00
• 8c10bda002 [README] - update faces_sub_image with training results and CLI args jensandClaude Sonnet 4.6 2026-05-30 21:15:02 +02:00
• 68b376657c [faces_sub_image] - add --mini_batch_size CLI arg jensandClaude Sonnet 4.6 2026-05-30 21:13:25 +02:00
• 9911caae06 [README] - document cd_jens removal, dbv simplification, Gibbs sampling fix jensandClaude Sonnet 4.6 2026-05-30 20:55:00 +02:00
• b5104ebc97 [bugfix] - remove dead cd_jens; fix cd_gaussian_binary dbv; fix cd_binary_gaussian Gibbs sampling jensandClaude Sonnet 4.6 2026-05-30 20:54:09 +02:00
• a6bfdf8eaf [README] - document train() shuffling and elif fixes jensandClaude Sonnet 4.6 2026-05-30 20:44:13 +02:00
• 003152518a [bugfix] - shuffle data each epoch; replace if chains with elif/else in train() jensandClaude Sonnet 4.6 2026-05-30 20:43:42 +02:00
• 96f271f160 [faces_sub_image] - add --num_epochs CLI arg jensandClaude Sonnet 4.6 2026-05-30 20:38:33 +02:00
• edeeb1773b [README] - update faces_sub_image entry; bump N_HID to 256 jensandClaude Sonnet 4.6 2026-05-30 20:18:46 +02:00
• 1d18983192 [README] - document L1/L2 batch-size scaling fix jensandClaude Sonnet 4.6 2026-05-30 19:59:34 +02:00
• d4d3d33931 [bugfix] - fix L1/L2 scaling: move regularisation out of grad_compute into state_adjust jensandClaude Sonnet 4.6 2026-05-30 19:58:44 +02:00
• 95b114003e [README] - document entity.py L1/L2 and image.py normalize fixes jensandClaude Sonnet 4.6 2026-05-30 18:28:45 +02:00
• abee870ada [bugfix] - fix L1 and L2 regularisation gradients in grad_compute jensandClaude Sonnet 4.6 2026-05-30 18:26:06 +02:00
• e0dcac5c82 [normalize] - simplify and harden normalize(); remove duplicate in load_image_patches jensandClaude Sonnet 4.6 2026-05-30 18:22:19 +02:00
• 07fc7dc607 [faces_sub_image] - add --grayscale CLI flag for single-channel mode jensandClaude Sonnet 4.6 2026-05-30 16:07:08 +02:00
• fc10fab627 [faces_sub_image] - add overlap patch extraction and boundary blending reconstruction jensandClaude Sonnet 4.6 2026-05-30 15:33:49 +02:00
• feaf96417a [faces_sub_image] - add --load_model and --do_train CLI args jensandClaude Sonnet 4.6 2026-05-30 15:25:22 +02:00
• 56f881d155 [README] - add faces_sub_image test entry jensandClaude Sonnet 4.6 2026-05-30 15:21:46 +02:00
• 441354088f [faces_sub_image] - add GB-RBM autoencoder test on Caltech WebFaces using SubImage ROIs jensandClaude Sonnet 4.6 2026-05-30 15:20:21 +02:00
• 7bbacc729b [README] - document all bug fixes in implementation notes jensandClaude Sonnet 4.6 2026-05-30 14:47:23 +02:00
• 7d48f835a6 [bugfix] - fix momentum reset, cd_gaussian_gaussian noise, rms_error, label CuPy compat jensandClaude Sonnet 4.6 2026-05-30 14:46:45 +02:00
• f0e98eb714 [README] - document cd_gaussian_binary bug fixes jensandClaude Sonnet 4.6 2026-05-30 14:37:27 +02:00
• 07e430f109 [cd_gaussian_binary] - fix negative phase: remove spurious noise, sample hidden states jensandClaude Sonnet 4.6 2026-05-30 14:36:18 +02:00
• 3c85705d85 [README] - add gaussian_autoencoder test entry jensandClaude Sonnet 4.6 2026-05-30 14:29:25 +02:00
• b032db3099 [gaussian_autoencoder] - add GB-RBM autoencoder test with Gaussian blob data jensandClaude Sonnet 4.6 2026-05-30 14:28:57 +02:00
• 86d39813cd [binary_autoencoder] - add BB-RBM autoencoder test and tests README jensandClaude Sonnet 4.6 2026-05-30 14:24:49 +02:00
• 95be9dd632 [conv2d] - add 2D convolution function and test jensandClaude Sonnet 4.6 2026-05-29 10:11:11 +02:00
• 759f4b5348 updated notebooks jens 2026-01-11 11:23:35 +01:00
• 92afc16137 [SubImage] - handle different image shapes jens 2026-01-11 11:21:42 +01:00
• 3f4cdc277f updated notebooks jens 2026-01-10 22:19:52 +01:00
• 51a06605dd [cifar_test_sub_image] - use L1-regulation jens 2026-01-10 22:18:45 +01:00
• 7220d41e02 [cifar-test] - found good L1-regulation parameter for sparse weights jens 2026-01-10 21:44:03 +01:00
• 432a596092 [mnist-test] - found good L1-regulation parameter for sparse weights jens 2026-01-10 19:58:39 +01:00
• be64b0f8ce refactored jens 2026-01-10 16:46:27 +01:00
• 1722a68b2a [entity] - refactored training parameters for grad_compute() - added l1-norm - improved tests jens 2026-01-10 15:21:40 +01:00
• 00fe5fd178 updated notebooks jens 2026-01-10 14:13:18 +01:00
• f7e3693f45 - added image module with sub image and normalize - improved tests jens 2026-01-10 14:12:29 +01:00
• b6a511e33c improved tests jens 2026-01-10 10:24:50 +01:00
• f1aae9b0f1 [train] - reset grad at start of epoch [state] - init(): set default parameter mu=0 jens 2026-01-10 10:19:30 +01:00
• 381ec8d0e3 improved tests jens 2026-01-09 15:28:53 +01:00
• 4c939cab8b cd_gaussian_binary: no sampling of training data jens 2026-01-09 15:28:06 +01:00
• 074ded1581 improved tests jens 2026-01-06 21:51:28 +01:00
• 9025f54434 simplified EntityParams.from_dict jens 2026-01-06 21:49:35 +01:00
• e7b287e632 - simplified TrainingParams.from_dict - fixed test_rbm jens 2026-01-06 21:48:09 +01:00
• 226453c7ae fixed status indention jens 2026-01-06 21:06:11 +01:00
• c879ff583b updated notebooks jens 2026-01-06 20:21:01 +01:00