Commit Graph

363 Commits (8425924da2d7c9db78bd5ec51807d4e3582fbbdc)

Author SHA1 Message Date
readlnh 9fe4343b7b [fix] gradient accumulation step alignment 2026-03-24 01:45:04 +01:00
jingyaogong f798445826 [update] empty_think_ratio 2026-02-06 19:15:21 +08:00
jingyaogong 533814ea2d [update] empty_think_ratio 2026-02-06 01:36:02 +08:00
jingyaogong ec6a14c65d [feat] data process 2026-02-06 01:17:57 +08:00
jingyaogong 69180447af [update] save interval 2026-01-30 20:30:50 +08:00
jingyaogong 9c1f4325f8 [update] safe half 2026-01-30 20:29:31 +08:00
jingyaogong 3d2f03cfff [fix] data skip 2026-01-18 16:56:29 +08:00
jingyaogong f0aa3ba52e [update] shuffle data 2026-01-18 16:39:34 +08:00
jingyaogong 17be26f5b9 [fix] max length 2026-01-17 13:26:14 +08:00
jingyaogong 33b0dfbec5 [update] pretrain load 2026-01-17 12:00:17 +08:00
jingyaogong b832bd5253 [update] align mask 2026-01-15 11:20:41 +08:00
jingyaogong b189f0046a [update] align loss 2026-01-15 00:56:32 +08:00
jingyaogong caf76fb884 [fix] compile unpack 2026-01-14 20:13:32 +08:00
jingyaogong 6ef2105e34 [feat] add compile 2026-01-14 14:42:30 +08:00
jingyaogong f0f73a208e [update] prompt prefill 2026-01-13 17:46:54 +08:00
jingyaogong 2dafa1f3bc [update] show speed 2026-01-07 23:33:47 +08:00
jingyaogong 379983849e [update] params log 2026-01-07 23:08:45 +08:00
jingyaogong 649328b932 [update] mask log 2026-01-07 22:12:26 +08:00
jingyaogong b3486a72d7 [update] readme 2026-01-07 00:58:38 +08:00
jingyaogong b64fddfdff [update] simplify loader 2026-01-06 01:20:52 +08:00
jingyaogong 3b062dad24 [update] readme 2026-01-06 01:18:10 +08:00
jingyaogong 6bea38c741 [update] rename train tokenizer 2026-01-06 01:17:33 +08:00
jingyaogong ba8448bd20 [update] readme 2026-01-05 23:15:25 +08:00
jingyaogong cb6be70365 [update] rename reason 2026-01-05 23:12:29 +08:00
jingyaogong 4138d7584c [fix] messages num 2026-01-04 11:03:16 +08:00
jingyaogong dbfeff23e3 [fix] dist cleanup 2026-01-02 22:25:55 +08:00
jingyaogong 3c56aca1c1 [update] aux loss 2026-01-01 22:41:46 +08:00
jingyaogong 5b87b9ad8a [fix] experts unused 2025-12-31 21:47:04 +08:00
jingyaogong e82af4bf8f [fix] layers set 8 2025-12-31 21:06:37 +08:00
jingyaogong 1c333636c5 [fix] moe unused 2025-12-31 21:00:06 +08:00
jingyaogong 73a2bcc412 [feat] get params 2025-12-31 20:46:59 +08:00
jingyaogong 759036b822 [feat] get params 2025-12-31 20:44:34 +08:00
jingyaogong 8ace03f068 [feat] update config 2025-12-31 10:29:13 +08:00
jingyaogong 6ee867fffc [feat] update lr 2025-12-31 10:27:09 +08:00
jingyaogong cc40ff6f45 [feat] compatible tokenizer 2025-12-31 10:26:46 +08:00
jingyaogong bc6f6dea99 [feat] stream load data 2025-12-28 16:58:52 +08:00
jingyaogong fc2eeff509 [feat] remove empty_cache 2025-12-27 07:14:36 +08:00
jingyaogong 658ddcae3a [feat] explicit left padding 2025-12-23 18:59:48 +08:00
jingyaogong 4cbca65883 [fix] lora weight 2025-12-22 21:27:29 +08:00
jingyaogong 0664b383e4 Merge pull request #594 from whiteswordLI/fix/lora-load-ddp-weights
Fix: support loading DDP-saved LoRA weights for inference
2025-12-22 21:19:16 +08:00
whitesword 65b648be32 Fix: support loading DDP-saved LoRA weights for inference 2025-12-22 20:50:25 +08:00
jingyaogong dacd4e5121 [feat] adjust seq length 2025-12-14 20:41:58 +08:00
jingyaogong b813b562e2 [feat] update readme 2025-12-11 15:45:50 +08:00
jingyaogong 37300329c5 [fix] dtype & lr 2025-12-09 13:01:38 +08:00
jingyaogong b6b0c8f675 Merge pull request #571 from dyhuachi/dyhuachi-patch-1
[fix] Refactor get_lr function to include min_lr calculation
2025-12-09 12:59:11 +08:00
dyhuachi 8fcc81696d [fix] Refactor get_lr function to include min_lr calculation
这里的退火算法会让参数里的lr的起始值变成原来lr的1.1倍,作出如下修改
2025-12-06 17:09:51 +08:00
jingyaogong 7614a3bf90 [fix] reduce aux_loss_alpha 2025-12-05 23:08:29 +08:00
jingyaogong 95bcb7ba19 [fix] cuda memory #559 2025-12-01 16:17:43 +08:00
jingyaogong 295b9ad443 [feat] update yarn 2025-12-01 16:15:05 +08:00
jingyaogong cafc3b22df [feat] release memory 2025-11-27 19:39:49 +08:00