tacit

tacit synced and deleted reference refs/tags/refs/pull/428/merge at tacit/nanochat from mirror 2026-01-16 00:14:01 +00:00

tacit synced and deleted reference refs/tags/refs/pull/161/merge at tacit/nanochat from mirror 2026-01-16 00:14:01 +00:00

tacit synced commits to refs/pull/93/merge at tacit/nanochat from mirror 2026-01-15 16:03:50 +00:00

2d9acd1df8 Merge 7950813a41 into 6bb92403d5

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

7312ec9898 fix buggy midtrain and update all kwargs to be idiomatic. that is, argparse uses dashes variables use underscores. the underscores are just a remnant of the previous Configurator object. This is the right way

3b50b77ed3 fix base_loss to report correct loss by switching the dataloader to the new default

Compare 7 commits »

tacit synced commits to refs/pull/59/merge at tacit/nanochat from mirror 2026-01-15 07:54:01 +00:00

131235d4bb Merge 23393eae83 into 6bb92403d5

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

7312ec9898 fix buggy midtrain and update all kwargs to be idiomatic. that is, argparse uses dashes variables use underscores. the underscores are just a remnant of the previous Configurator object. This is the right way

3b50b77ed3 fix base_loss to report correct loss by switching the dataloader to the new default

Compare 10 commits »

tacit synced commits to refs/pull/438/merge at tacit/nanochat from mirror 2026-01-15 07:54:00 +00:00

f057b132e3 Merge b88cef6053 into 6bb92403d5

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

Compare 3 commits »

tacit synced commits to refs/pull/437/merge at tacit/nanochat from mirror 2026-01-15 07:54:00 +00:00

f00beb04da Merge 8cfa0451f4 into 6bb92403d5

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

Compare 3 commits »

tacit synced commits to refs/pull/436/merge at tacit/nanochat from mirror 2026-01-15 07:53:59 +00:00

1937f5a46d Merge 97364273e2 into 6bb92403d5

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

Compare 3 commits »

tacit synced commits to refs/pull/434/merge at tacit/nanochat from mirror 2026-01-15 07:53:58 +00:00

ca0c29af26 Merge 16d691d5f1 into 6bb92403d5

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

Compare 3 commits »

tacit synced commits to refs/pull/433/merge at tacit/nanochat from mirror 2026-01-15 07:53:57 +00:00

ffce7d4c10 Merge c0618a6b7e into 6bb92403d5

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

Compare 3 commits »

tacit synced commits to refs/pull/400/merge at tacit/nanochat from mirror 2026-01-15 07:53:56 +00:00

8c2445d702 Merge 32ce342c88 into 6bb92403d5

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

Compare 3 commits »

tacit synced commits to refs/pull/431/merge at tacit/nanochat from mirror 2026-01-15 07:53:56 +00:00

435ccafef8 Merge 1a9df65ee7 into 6bb92403d5

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

Compare 3 commits »

tacit synced commits to refs/pull/370/merge at tacit/nanochat from mirror 2026-01-15 07:53:55 +00:00

70bcc7fe69 Merge 9a9b12b1be into 6bb92403d5

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

Compare 3 commits »

tacit synced commits to master at tacit/nanochat from mirror 2026-01-15 07:53:54 +00:00

6bb92403d5 changes and optimizations to muon, making it more efficient and simpler/cleaner a bit

3142ca1a28 minor helpful message

Compare 2 commits »

tacit synced commits to refs/pull/436/head at tacit/nanochat from mirror 2026-01-14 23:43:41 +00:00

97364273e2 feat: restrict FA3 loading to Hopper+ GPUs (SM90+) to fix crashes on consumer hardware

d7fccbab82 fix: enforce (B, H, T, D) layout for SDPA fallback to support CPU strictness

Compare 2 commits »

tacit synced commits to refs/pull/436/merge at tacit/nanochat from mirror 2026-01-14 23:43:41 +00:00

7a0ff37b2f Merge 97364273e2 into 7312ec9898

97364273e2 feat: restrict FA3 loading to Hopper+ GPUs (SM90+) to fix crashes on consumer hardware

d7fccbab82 fix: enforce (B, H, T, D) layout for SDPA fallback to support CPU strictness

Compare 3 commits »

tacit synced commits to refs/pull/311/merge at tacit/nanochat from mirror 2026-01-14 23:43:41 +00:00

beb83a4af5 Merge 81a958350a into 7312ec9898

7312ec9898 fix buggy midtrain and update all kwargs to be idiomatic. that is, argparse uses dashes variables use underscores. the underscores are just a remnant of the previous Configurator object. This is the right way

3b50b77ed3 fix base_loss to report correct loss by switching the dataloader to the new default

f92efce169 add negative result about not allowing attention across BOS tokens. A lot more code complexity for basically no gain in performance

43c29dd9d5 Big DataLoader refactor: BOS-aligned dataloaders with epoch tracking for pre/mid-training

Compare 5 commits »

tacit synced commits to refs/pull/436/head at tacit/nanochat from mirror 2026-01-14 15:33:49 +00:00

68e66be05c fix: wrap FA3 import in try-except block to support both CUDA and MPS

tacit synced commits to refs/pull/85/merge at tacit/nanochat from mirror 2026-01-14 15:33:49 +00:00

7cacaf0ec0 Merge 04862cbfea into 7312ec9898

7312ec9898 fix buggy midtrain and update all kwargs to be idiomatic. that is, argparse uses dashes variables use underscores. the underscores are just a remnant of the previous Configurator object. This is the right way

3b50b77ed3 fix base_loss to report correct loss by switching the dataloader to the new default

f92efce169 add negative result about not allowing attention across BOS tokens. A lot more code complexity for basically no gain in performance

43c29dd9d5 Big DataLoader refactor: BOS-aligned dataloaders with epoch tracking for pre/mid-training

Compare 8 commits »

tacit synced commits to refs/pull/436/merge at tacit/nanochat from mirror 2026-01-14 15:33:49 +00:00

4bd88b4eaa Merge 68e66be05c into 7312ec9898

68e66be05c fix: wrap FA3 import in try-except block to support both CUDA and MPS

Compare 2 commits »

tacit synced commits to refs/pull/433/merge at tacit/nanochat from mirror 2026-01-14 15:33:48 +00:00

e2e17e1fe1 Merge c0618a6b7e into 7312ec9898

7312ec9898 fix buggy midtrain and update all kwargs to be idiomatic. that is, argparse uses dashes variables use underscores. the underscores are just a remnant of the previous Configurator object. This is the right way

3b50b77ed3 fix base_loss to report correct loss by switching the dataloader to the new default

f92efce169 add negative result about not allowing attention across BOS tokens. A lot more code complexity for basically no gain in performance

43c29dd9d5 Big DataLoader refactor: BOS-aligned dataloaders with epoch tracking for pre/mid-training

Compare 5 commits »