Slayer1B 1.04 β€” training checkpoints

Full trainer checkpoints of continued pre-training of Slayer1B 1.03 from its final checkpoint (ckpt_7150.pt, sha256 d7cad967…). Each file holds the model weights and the optimizer state.

Status (28 October 2026): complete. Release Slayer1B 1.04 is the state at step 10010 (ckpt_10010.pt). This repository holds every checkpoint of the 1.04 run (7200–10010, 58 files) and, because the run continued without a break, also the checkpoints of the next two releases (10050–13848, 78 files):

release steps files per-release repository
1.04 7151–10010 58 this repository
1.05 10011–12870 58 Slayer1B-1.05-checkpoints (same files)
1.06 12871–13848 20 Slayer1B-1.06-checkpoints (same files)
Start 1.03 ckpt_7150.pt (step 7150, constant lr 6e-4)
Steps 7151 β†’ 10010, 1 048 576 tokens per step (256 sequences Γ— 4096 tokens)
Schedule constant lr 6e-4, no decay
Data SlayerLab/slayer-pl-8x3b revision d50df04a, continuing the single pass over the corpus: tail of pack-03 (steps 7151–7192), pack-04 from step 7193, pack-05 from step 8991
Leak-scan exclusion from pack-04 on (step 7193), the 5133 documents with n-gram overlap with BoolQ / MMLU / ARC test items are excluded (list sha256 84caed78…); steps up to 7192 used the unfiltered data
Hardware 4 Γ— NVIDIA L40S (48 GB), DDP. Training paused (budget) at step 8300 and resumed from ckpt_8300.pt
Checkpoints every 50 steps from 7200, and 10010; all are constant-lr states
Trainer train_gpt_ref r24.1 (sha256 881cb055…); model definition train_gpt_ref_r23.py (unchanged). Steps 10011–13848: train_gpt_ref r24.2 (sha256 46f83b51…) on 8 Γ— L40S
Training curve and evaluations https://track.fabryka.ai/run/f056bb5c-f1da-56d5-9776-cc33063f10a5

SHA-256 (steps 7200–10010)

Each value equals the LFS object id of the file in this repository (checked anonymously after upload).

step sha256
7200 dcff686a2b719ff16d5b8bc625b5f77a2d45d207b7d5fbea35cdf411bd32f5fd
7250 dcecd2d043eae4d6474ec040a61c293846446b9a041970e6ced0f3c5a33f723f
7300 b59643ebb8e2c68784de6140aa34f1c46d6d05d3e381ec3bc46c60d4af40d249
7350 cd0a91338aa512ca399a5df05c0e0e5946cd9f993cdb75ae93745f3c2f3e9bba
7400 7d97c13b3c72e1b023626d5253ff030603764dd47c6af259c9c11dd0b6a92b6b
7450 cbc2b17fa6a716bc970c6e6303396f40f1ab7561595e12ace976e5833edbfe0a
7500 20bb66250b63a43ac54adbab63493ffac56b41e919463e13e274908f32781867
7550 0b19d6e08d8b9b7d07bc2a0f0ee86ea0057916b8aaf84d39aec47e0515568e6e
7600 37b547147ea6ec0856cb23bc78bec3f78319ac9c45ec63480f833c5335601d59
7650 0cf517a74b8d15ae226dee131c2185053cb0dbde8e5ba8d84a9558b45b0f5ab0
7700 6de388d452035069608033eacd262ce44edc6f4b8e819512320118e3b68989f0
7750 8d8ade244783dded8335a9f5fbe000b215e5af6e8f39d6493016f90ec5946ab0
7800 473893a4c4d1386be504501ac41692359aaf66817837e08bb775e0973b9b9f2c
7850 c0ec371d48741888918040b88906d3c8077e6213d276fb076047bbe0eb33902b
7900 7a5beaa46a025c1bd9b42b43faaae0a6362692a1a66913cc989b73b590fc4cac
7950 8f8fb65da465e410495382a7b9f401349cd3d1af9c864bc61b02058288ffb7a3
8000 b4b4aa12a8bd7f9b4949226b49e953ed88dded2fddd984a080f1d02c9c1712df
8050 2bdc077d7f71ff91e74fbc0deb120c24d41cd63cddcb079100f2fe3ad7b4ecb4
8100 5e2e4dd12bfc5a8c5fa690b442b0062a2b4b559064a57dcb1830e68357548908
8150 05dcdcf26b4d5cbf86b7fd56aaedc3bc7dafe03143a2eb98ee0c3dc6df1e5a19
8200 39d7f609e88cb7f9363fa2920c29d0c13daa948c9911c1f7c5bcd55d619d72dc
8250 b2c9fa7c9f83b5e4209b1d4c01267fcdf33459fd85255dd21c986c033ff21eef
8300 1487a2673b64e26c241ff46a7e98222f9cc6a059a7c509ebd9cf3581bc3c68d6
8350 66f4c2e7946379a9e372be1b2865c788b0008a97afcb73c0b5c9d4e341caa9f3
8400 112c3cf9878f65cdd4904dd7fd4387ff8b188192ba744b0523c74d0aedfb82e8
8450 5ac0a07ec745fc71e6f9fd136fcb4f3824df07bff3a7568bf570bd7ff269665e
8500 3083a4eb82c32eb6c12a0a8b481b1ee3deaf70294d9ec728371a9bfb614d58d5
8550 73b425383f706df43f2f4b482186e1ba733c0e47ce6d832e95c53b6a9142eeaa
8600 daa4308f0502c5c6562a9b96a33ab5c80dd051a8a4d61e541f262655be9c62b7
8650 38147d42c718faeaf34b861827a2502689aac857c2781a398a018af817176b2b
8700 a7849393b913cd9a7ed07602e623110edc3fb12c5f96529f7d639bd26acb96db
8750 90a0f8a261839aa615481204c3d7f9286579f50df1f4f8bbb01925ff25e88b2f
8800 24b648b0ed510e2958121d1c51c07731b605c8c09a64154a11dcb03ecdc7a191
8850 d4a7476c44b7f9ff9dc111acc0dd51c8041c93a31c7555c4a08de46ec1d8bbc5
8900 68d0c316249adf620256ef90e5becbddd06920667a3fa858d0028ec2f5d17e7a
8950 9c47394c72b08c4da14b32504fdf1eb1711c11fa13d2f49a862379954029ed9c
9000 d8ad7edec7e27425f5be4ed0c7d6b1cf32d84ea00e33ebd77fbde00c598dca64
9050 cca1ab10de62de41eef5038e87f3b3ffcd17876c2cea284ef9dfd83e6c048028
9100 1968f7969b207d9740fdbf4a4efdf0dfebedc17c9fc1b8e7d2cdbc20c43048f5
9150 9798249cfdbce9cebb687df9b7cad1c8c87a30ecde243a3ba0853f75007b5bb6
9200 9a3a7dd1aaa9a57df1fd68d6a549f0a92e519394e04b5e4371c68200a3aaabab
9250 209c1afd3360955e51c4b8fde7ea48789ad62ce2f31f0e839e32627ca6f9fd66
9300 4540338c3bad2642b1d8f45ea093fb25ae809f29b520a6787e034dab2995dd45
9350 2bb43e35480e82be7b509fcb8cc6b796897c7e1203f50840665f4d23ab7753ba
9400 2783ac61aa374ce936ff59c5f3e2df821afefecd774979478d88a3b84e7acad7
9450 93601b632ae2e1ae9272ed0a9ae792c0b33b825f867e2dc7b454e8615d6a7cb8
9500 14ee41b7b04165ecbc3dea3e1db88aca070df14a85732e21133c17a976db5312
9550 4af10fe35c5d087b223b3af96852483aabbc876dda3729601fa0fe3a62ad7fca
9600 a5fc866ef4fe196cb959b6a5d9aa34f236bb15fa8294c56da27b3c6b2038e896
9650 8f3f58e37fe7b2c51b10e4e8569041f73457d198a1fdf6c53f2b63c38a1e4a28
9700 75de6f614ea667790fb4209a8b7ee9fe281cea7b03da9eb307263f4c37b291d3
9750 c42785a68715e7dff90e2471aef433a77cf9b01b7c20c323c39a29ef148993c9
9800 c9fc88ae437d6c5c0d4532d1225ff3af1e56b2709280f96f864c1695f0552c3d
9850 6d7bdbcdb035a8445dc079ee011feac4ead3aeb5eaa15b1950db4f7965b316c8
9900 e0bb1cc460798d8eaed3260550a0dd874b53dbbfba6f86c7c66c2ff2c366fa2f
9950 b8249f8f20b3a55c992d0ec3ae2d44ad27a93b769cd25926f432dcd611536192
10000 37fb6d1893978dfce3a2f87a6bc177e069f4af7ec7b7581249d2c908acfa8ba0
10010 04438959d7b317d1a8d972b264e9610a06fd8addde7349c6e9f458ee56dc1e51

Steps 10050–13848: see the per-release repositories above (identical files, SHA-256 listed there).

File format

torch.save dictionary of the trainer: model, opt (Muon + AdamW), step, config. Load with torch.load(path, map_location="cpu", weights_only=True). Model class: train_gpt_ref_r23.py in the model repositories (1.00–1.06). Every file is 8 860 287 151 bytes.

License

Apache-2.0 for the checkpoints. The training data includes CC BY-SA sources (shared/core_sa, shared/legal_sa of slayer-pl-8x3b); see the dataset card for the licenses of the data.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Space using SlayerLab/Slayer1B-1.04-checkpoints 1