Commit Graph

  • b68cd36708 http://talkchess.com/forum3/viewtopic.php?f=2&t=76736&p=885254#p885254 Tomasz Sobczyk 2021-02-28 12:51:45 +01:00
  • 0f3f5d85fb Introduce DistanceFromPV Stéphane Nicolet 2021-02-26 10:13:37 +01:00
  • 7c30091a92 Introduce ProbCut for check evasions Vizvezdenec 2021-02-20 22:48:08 +01:00
  • 6294db7514 Tune search parameters (with Unai Corzo) Vizvezdenec 2021-02-19 10:32:12 +03:00
  • a31007c9e7 Restore development version Stéphane Nicolet 2021-02-20 22:19:14 +01:00
  • 3597f1942e Stockfish 13 sf_13 Joost VandeVondele 2021-02-15 13:58:51 +01:00
  • 9f6d69c544 Update README.md Stéphane Nicolet 2021-02-16 16:19:37 +01:00
  • 40cb0f076a Small trivial clean-ups, February 2021 Lolligerhans 2021-01-31 13:46:05 +01:00
  • b46813f9b7 Update Top CPU Contributors Stéphane Nicolet 2021-02-15 23:58:03 +01:00
  • 76daa88cf8 PV-Nodes likely to fail low Vizvezdenec 2021-02-11 10:45:16 +03:00
  • 573f0e364f Better code for hash table generation mattginsberg 2021-02-11 22:29:28 +01:00
  • 550fed3343 Enable New Pass Manager for Clang. Gian-Carlo Pascutto 2021-02-09 17:38:58 +01:00
  • b15e3b3fa9 Disable ThinLTO when using Clang. Gian-Carlo Pascutto 2021-02-09 14:05:35 +01:00
  • 1f87a9eb6c Fixes FreeBSD compilation when using Clang Andy Pilate 2021-02-08 22:02:42 +01:00
  • 29ed22de8c Search Parameters Tuning bmc4 2021-02-04 21:50:22 -03:00
  • 5ebdc40f83 Pawns Tuning FauziAkram 2021-02-07 13:44:11 +02:00
  • 9f8058bd26 Simplify En Passant bmc4 2021-01-31 10:12:32 -03:00
  • 6617ad6e03 Tune ordering of moves at internal nodes bmc4 2021-01-31 00:05:39 -03:00
  • dd96095214 Simplify Chess 960 castling bmc4 2021-01-17 15:21:45 -03:00
  • 0db374777e Speed Up Perft Search bmc4 2021-01-19 09:21:59 -03:00
  • befbcffb4e Clean Up Castling in gives_check bmc4 2021-01-18 17:03:43 -03:00
  • 7d0a16e06d Avoid more expensive legality check bmc4 2021-01-16 01:44:11 -03:00
  • 70a818cbd6 Small cleanups Lolligerhans 2021-01-12 14:59:51 +01:00
  • 1188141aa7 Improve play for closed positions Stéphane Nicolet 2021-01-21 18:53:29 +01:00
  • b7f643fe39 Add .gitignore Rod Johnson 2021-01-04 22:59:55 +11:00
  • 329ef2a6aa Change lock type Krystian Kuzniarek 2020-12-31 17:00:39 +01:00
  • 74774c36e1 Fix wrong multipv depth range. Fixes #291 Tomasz Sobczyk 2021-01-25 11:54:12 +01:00
  • 77eeea407c Add penalty for doubled pawns in agile structure Lolligerhans 2021-01-12 14:30:25 +01:00
  • 6dddcecb09 Optimize generate_moves Tomasz Sobczyk 2021-01-10 03:30:40 -03:00
  • ee3f7b6b6e Bad Outpost Pawn Scale FauziAkram 2021-01-10 01:31:09 +02:00
  • 37c2b5685e Refine stat based reductions Vizvezdenec 2021-01-09 17:42:58 +03:00
  • 5f222f1d98 Rethink En Passant Evasion Capture BM123499 2021-01-08 14:03:26 -03:00
  • 0266e70297 Fix static_assert. Dieter Dobbelaere 2021-01-10 12:22:52 +01:00
  • 87586b3d0c Use correct chess terms + fix spelling. Dieter Dobbelaere 2021-01-09 16:46:06 +01:00
  • b1bb376c3c Small code cleanup in LMR Vizvezdenec 2021-01-10 20:24:37 +03:00
  • 303713b560 Affine transform robust implementation MaximMolchanov 2021-01-11 07:49:41 +02:00
  • 4d30438400 Remove Condition from Generate_Move Loop bmc4 2021-01-10 03:30:40 -03:00
  • c4d67d77c9 Update copyright years Joost VandeVondele 2021-01-08 17:04:23 +01:00
  • 2c1be0be8e Reorder conditions in LMR and pruning Vizvezdenec 2021-01-01 12:59:35 +03:00
  • 23c385ec36 Affine transform refactoring. MaximMolchanov 2021-01-06 05:29:32 +02:00
  • d21e421ad7 WeakUnopposed penalty for backwards on file A or H FauziAkram 2020-12-25 16:19:04 +02:00
  • 8ec97d161e Remove razoring Unai Corzo 2020-12-26 13:48:04 +01:00
  • 8985c210a1 Simplify away late irreversible move extension Unai Corzo 2020-12-27 21:19:19 +01:00
  • c57c71bf5c Assorted parameter tweak Unai Corzo 2020-12-30 15:38:25 +01:00
  • 4262461457 Tweak capture LMR. Stefan Geschwentner 2020-12-31 01:45:37 +01:00
  • 1f7e5d3861 Add thread sanitized run for instrumented_learn and fix races. Tomasz Sobczyk 2020-12-24 13:35:48 +01:00
  • 51deae8998 Do more LMR for captures Vizvezdenec 2020-12-25 01:11:09 +03:00
  • acf95c7c98 Accumulate clipping statistics to a 64 bit integer to prevent overflow for larger batch sizes. Tomasz Sobczyk 2020-12-24 13:30:04 +01:00
  • 1b560efabd Correctly handle the last batch of data in sfen_reader Tomasz Sobczyk 2020-12-24 17:06:12 +01:00
  • 6d28d97a91 Don't unload evalfile on set nnue false Tomasz Sobczyk 2020-12-24 17:01:56 +01:00
  • c1e69f450e Prevent q_ in loss calculation from reaching values that would produce NaN Tomasz Sobczyk 2020-12-24 13:25:48 +01:00
  • bb6188430d Add split_count argument to shuffle_binpack.py Joost VandeVondele 2020-12-24 14:34:26 +01:00
  • 4f6fdca31f Reduce the amount of sfens buffered for the validation step. Tomasz Sobczyk 2020-12-24 14:09:37 +01:00
  • 7636bcccd1 Correctly account for factors when computing the average absolute weight of the feature transformer. Tomasz Sobczyk 2020-12-24 13:49:44 +01:00
  • 2061be4730 smart_fen_skipping at gensfen_nonpv level Tomasz Sobczyk 2020-12-17 16:12:26 +01:00
  • 868b4e9421 add gensfen_nonpv docs Tomasz Sobczyk 2020-12-16 21:11:09 +01:00
  • 96b377a90a Add gensfen_nonpv Tomasz Sobczyk 2020-12-13 15:33:35 +01:00
  • 3f73c40412 More deterministic move accuracy validation. Tomasz Sobczyk 2020-12-21 21:50:30 +01:00
  • b50dcd7dde allow for repeated searches in rescoring Joost VandeVondele 2020-12-23 22:42:42 +01:00
  • b06ef36ae5 Correct Outflanking calculations in classical eval Moez Jellouli 2020-12-20 22:28:23 +01:00
  • 45b05328b6 Tweak the formulas for unsafeSquares FauziAkram 2020-12-20 17:50:34 +02:00
  • 8ca82646a9 Use plain nnue eval for validation loss calculation instead of first performing qsearch Tomasz Sobczyk 2020-12-21 20:35:23 +01:00
  • 6853b4aac2 Simple filtering for validation data. Tomasz Sobczyk 2020-12-21 20:57:51 +01:00
  • 50df3a7389 fix annoying warning Tomasz Sobczyk 2020-12-21 22:20:39 +01:00
  • 994eb5e183 rescore_fen -> rescore. Make it work on .bin and .binpack inputs. Tomasz Sobczyk 2020-12-21 00:01:34 +01:00
  • ffae19b5a1 Add docs for rescore_fen Tomasz Sobczyk 2020-12-20 23:06:19 +01:00
  • a9cfaa4d98 Add a tool for rescoring fens from an epd file with fixed depth search Tomasz Sobczyk 2020-12-13 23:20:32 +01:00
  • f56613ebf6 Add 'validation_count' option for 'learn' that specifies how many positions to use for validation Tomasz Sobczyk 2020-12-19 23:46:18 +01:00
  • 1f3b5b8b54 Simplify condition for assigning static-eval based bonus pb00067 2020-12-14 16:30:56 +01:00
  • 66a7a8a0cc Adjust definition of unsafeSquares FauziAkram 2020-12-16 15:35:39 +02:00
  • a88a38c3a9 Increase reduction in case of stable best move Vizvezdenec 2020-12-14 03:49:04 +03:00
  • 16adcb5374 Merge static history into main history, pb00067 2020-12-13 21:23:30 +01:00
  • d862ba4069 AVX512, AVX2 and SSSE3 speedups mstembera 2020-12-12 14:18:38 -08:00
  • a7378f3249 Make next_fen in opening_book a critical section Tomasz Sobczyk 2020-12-13 17:45:16 +01:00
  • 76fbc5e3d0 Make score sign flip optional Joost VandeVondele 2020-12-12 20:01:42 +01:00
  • f4b4430380 remove unnecessary makefile commands and fix blas on mac kennyfrc 2020-12-12 23:39:42 +08:00
  • d706ae62d7 New Imbalance Tables Tweak FauziAkram 2020-12-07 19:28:47 +02:00
  • c7f0a768cb Use arithmetic right shift for sign extension in MMX and SSE2 paths Fanael Linithien 2020-12-07 14:46:29 +01:00
  • 9c65e868f9 Enhance pgn_to_plain.py Joost VandeVondele 2020-12-08 23:57:54 +01:00
  • d99ba07b81 Fix incorrect enpassant flag for moves read from uci format in the binpack lib Tomasz Sobczyk 2020-12-08 23:44:26 +01:00
  • b49fd3ab30 Add -lstdc++fs to the link line of gcc Joost VandeVondele 2020-12-08 18:05:31 +01:00
  • ae045e2cd8 Merge pull request #258 from kennyfrc/stockfish-nnue-2020-08-30-macos nodchip 2020-12-09 08:39:36 +09:00
  • 055f907315 Merge branch 'master' into stockfish-nnue-2020-08-30-macos Kenn Costales 2020-12-08 22:49:11 +08:00
  • bb26ce5aa1 mac specific makefile with compilation instructions kennyfrc 2020-12-08 22:14:18 +08:00
  • 3a1bd1185f Add binpack coarse shuffle tool. Tomasz Sobczyk 2020-12-05 15:00:02 +01:00
  • 8630d03dd4 Add comments to uncommented parts of code Vizvezdenec 2020-12-01 15:02:35 +03:00
  • be7a03a957 Introduce static history Vizvezdenec 2020-12-05 04:00:41 +03:00
  • 28d6d7cb03 Avoid computing gradient for validation loss. Tomasz Sobczyk 2020-11-30 22:25:05 +01:00
  • fafb9557a8 Get train loss from update_parameters. Tomasz Sobczyk 2020-11-30 22:17:44 +01:00
  • 4eb0e77a2a Store references instead of copying the results of intermediate autograd computations. Tomasz Sobczyk 2020-11-30 21:43:24 +01:00
  • 6cd0b03098 Add some comments regarding the current state of autograd loss computation. Tomasz Sobczyk 2020-11-30 21:34:43 +01:00
  • 99cb869db3 Reintroduce use_wdl. Tomasz Sobczyk 2020-11-30 21:17:44 +01:00
  • cf6bc7ecaf Cleanup around get_loss Tomasz Sobczyk 2020-11-30 20:42:54 +01:00
  • 256c4b55ec Properly apply gradient norm clipping after it's scaled in the update_parameters. Tomasz Sobczyk 2020-11-30 20:39:46 +01:00
  • de675e3503 Reintroduce optional scaling of the teacher signal. Tomasz Sobczyk 2020-11-30 20:32:53 +01:00
  • 01ae7b1e2c Simplify passing constants that may vary between calls. Tomasz Sobczyk 2020-11-30 20:26:02 +01:00
  • cbd973fdaa Detect constant expressions in autograd and return 0 grad early. Tomasz Sobczyk 2020-11-30 16:50:51 +01:00
  • e975889132 Move cross_entropy calculation to a separate function. Tomasz Sobczyk 2020-11-30 15:21:39 +01:00
  • 891abf5511 Make the autograd loss expression chain thread_local. Tomasz Sobczyk 2020-11-30 15:14:26 +01:00
  • 8adf00ae6e Identify a single evalation chain by ID in autograd to prevent cache reuse for subsequent evaluations of the same expression tree. Tomasz Sobczyk 2020-11-30 14:01:31 +01:00