Commit Graph

199 Commits

Author SHA1 Message Date
Colin cda7f04e49 Fix model path. 2025-03-18 15:58:08 +08:00
Colin 7faf629d45 Refine seed config. 2025-03-14 17:38:24 +08:00
Colin e3493163f3 Update configuration to str for tensorboard. 2025-03-13 23:02:11 +08:00
Colin b3817f84fe Refine model of wit. 2025-03-13 16:52:33 +08:00
Colin f411b1cc5e donot use auto optimizer. 2025-03-13 14:28:53 +08:00
Colin 990e27ba15 Fix train define. 2025-03-12 20:02:02 +08:00
Colin 90e94db2c1 Rename QwenModule to lightmodule. 2025-03-10 19:14:47 +08:00
Colin 1efda9fe25 Update rwkv train. 2025-03-10 16:26:53 +08:00
Colin 0600d46f2f Add safe softmax demo code. 2025-03-09 14:39:34 +08:00
Colin c4e9637c10 Add rwkv in wit. 2025-03-07 18:53:46 +08:00
Colin 251ea7f004 Add rwkv flow graph. 2025-03-06 23:22:50 +08:00
Colin 821e7206b8 Refine rwkv. 2025-03-05 19:39:08 +08:00
Colin 240858c030 Update rwkv. 2025-03-03 21:30:58 +08:00
Colin 4f18296e40 Format rwkv/RWKV-v7/rwkv_v7_demo.py 2025-03-03 15:47:21 +08:00
Colin 002f132818 Add hook of attention for query qkv. 2025-03-03 15:38:00 +08:00
Colin 3eea09d78c Add rwkv v7 demo. 2025-03-03 14:53:15 +08:00
Colin e3b63f4635 Refine model define. 2025-02-28 13:38:28 +08:00
Colin bff65b189d Add query file. Refine print tree. 2025-02-26 16:55:20 +08:00
Colin 3c34d12fba Refine dataset and nodetree. 2025-02-24 21:38:31 +08:00
Colin a1d5fce300 Refine node tree define and print. 2025-02-23 16:46:37 +08:00
Colin 3a7ce45654 Refine meaning dataset map. 2025-02-22 16:50:16 +08:00
Colin f0469b351c refine meaning dataset tree print. 2025-02-22 01:53:58 +08:00
Colin 81f9e54ca3 Update inference and val dataset. 2025-02-21 17:28:21 +08:00
Colin 383c40afd7 Update inference for debug. 2025-02-21 15:51:27 +08:00
Colin 7cf31a1f78 Rename mask level and index. 2025-02-21 15:33:37 +08:00
Colin bca06af2dc Refine meaning dataset. 2025-02-20 17:30:46 +08:00
Colin 0b19fd576a Refine train code. 2025-02-19 19:39:59 +08:00
Colin 3feec36059 refine code. 2025-02-19 18:01:41 +08:00
Colin c92db47135 Benchmark single 4090D 15fps
cd /wit
python train.py
2025-02-18 21:40:49 +08:00
Colin f8480678d8 Refine meaning dataset document. 2025-02-18 19:36:16 +08:00
Colin 383125edc9 Refine dataset org. 2025-02-18 14:21:15 +08:00
Colin e635ce0df4 Regine wit config method. 2025-02-17 19:41:40 +08:00
Colin cdee69bf54 Regine minist unsuper. 2025-02-17 14:27:15 +08:00
Colin f74a5d29bd Update unsuper minist. 2024-11-04 14:23:29 +08:00
Colin df05002c90 Update unsuper minist. 2024-11-03 15:15:33 +08:00
Colin 1bb41f0ee7 Update. 2024-10-31 15:15:13 +08:00
Colin 2ad977a072 Update show. 2024-10-31 00:01:57 +08:00
Colin 59449df047 Refine minist unsuper. 2024-10-28 18:52:28 +08:00
Colin 5b2cd4da61 Add model.conv1.weight normal after update grad. 2024-10-28 16:31:42 +08:00
Colin 6a0b47c674 Update show tools. 2024-10-27 19:48:53 +08:00
Colin a0da2565fe Update minist unsuper. 2024-10-22 19:13:19 +08:00
Colin 385c438c1c Update unsuper. 2024-10-22 13:54:26 +08:00
Colin f3690fd47f Update unsuper learning. 2024-10-07 16:39:29 +08:00
Colin 22464e7724 Update unsuper. 2024-10-05 16:17:41 +08:00
Colin 45d5701835 Unsuper train with max confidense of conv output 2024-10-04 01:30:18 +08:00
Colin 81f203ce59 update the max grad's weight. 2024-09-22 15:18:08 +08:00
Colin 03516f6302 Remove unsuper pool layer. 2024-09-21 17:44:57 +08:00
Colin 39cecd1146 mov optimizer to grad sub. 2024-09-17 01:59:00 +08:00
Colin ce64a2f7aa Update show. 2024-09-16 18:46:09 +08:00
Colin 8583bc56d7 Delete png. 2024-09-16 18:29:15 +08:00