Commit Graph

  • a031972ba6 f dev2 Zchen 2025-10-21 01:07:57 +08:00
  • ab12d0b7ee f Zchen 2025-10-21 00:31:59 +08:00
  • e7c9b95b00 f Zchen 2025-10-21 00:19:05 +08:00
  • 5a0079641a f Zchen 2025-10-20 23:34:44 +08:00
  • e399cf262a ff Zchen 2025-10-20 13:37:11 +08:00
  • 7358ff3d79 Enable soft device placement for CTC operations and update related comments Zchen 2025-10-20 11:22:13 +08:00
  • f8fb4d7133 Remove setup script, TPU memory monitor, and training model script Zchen 2025-10-20 11:05:03 +08:00
  • 7c272b7c5b Remove test scripts for data loading and TensorFlow implementation Zchen 2025-10-20 01:37:22 +08:00
  • 0a0e07a193 Remove custom CTC loss implementation for TPU from the TripleGRUDecoder class Zchen 2025-10-20 01:16:50 +08:00
  • 06ddbc6ac2 Refactor input function to implement batch-first approach with dynamic padding and apply data augmentation post-batching for TPU compatibility Zchen 2025-10-20 00:58:29 +08:00
  • fabf70cfa9 Enhance dataset shape analysis by implementing parallel processing and improving sampling logic Zchen 2025-10-20 00:35:17 +08:00
  • e1669b5a4c Increase batch size from 256 to 512 for training in rnn_args.yaml Zchen 2025-10-20 00:21:33 +08:00
  • 6e02894a8a f Zchen 2025-10-20 00:13:39 +08:00
  • 4db3625dc5 f Zchen 2025-10-19 23:55:56 +08:00
  • fed5fd8251 f Zchen 2025-10-19 22:25:21 +08:00
  • 4b373ab317 ff Zchen 2025-10-19 20:16:23 +08:00
  • 40d0fc50de f Zchen 2025-10-19 13:18:20 +08:00
  • 4328114ed6 Add dataset shape analysis function and integrate into input function for TPU optimization Zchen 2025-10-19 11:04:36 +08:00
  • cfd9653da9 Enhance dataset caching logic for training and validation sets with improved messaging Zchen 2025-10-19 10:31:31 +08:00
  • 558be0ad98 Refactor individual dataset creation for improved I/O efficiency and add logging for error handling Zchen 2025-10-19 10:31:18 +08:00
  • d83f990beb f Zchen 2025-10-17 12:20:17 +08:00
  • eb058fe9d3 ff Zchen 2025-10-17 11:57:10 +08:00
  • 57360bec8a Remove CPU optimization call and add logging for TPU strategy and data pipeline performance Zchen 2025-10-17 11:45:20 +08:00
  • eb4e3fc69f fff Zchen 2025-10-17 11:38:57 +08:00
  • 6c7abfcca8 f Zchen 2025-10-17 10:53:58 +08:00
  • 7ede7b5f12 f Zchen 2025-10-17 02:09:14 +08:00
  • ca8c615505 f Zchen 2025-10-17 02:01:48 +08:00
  • 49700456b8 f Zchen 2025-10-17 01:58:28 +08:00
  • 8ee09b6b5e f Zchen 2025-10-17 01:54:32 +08:00
  • a5a3179ca6 f Zchen 2025-10-17 01:49:03 +08:00
  • 59fb73ee9f f Zchen 2025-10-17 01:36:08 +08:00
  • 0a72143513 legacy adam Zchen 2025-10-17 01:26:02 +08:00
  • 7df78244e6 adamw to adam Zchen 2025-10-17 01:07:01 +08:00
  • a96e272f7b fix twice gradient cut Zchen 2025-10-17 00:51:53 +08:00
  • 7a43ebfb71 refactor: streamline model building and ensure dtype consistency in L2 loss calculation Zchen 2025-10-16 23:06:09 +08:00
  • 9453b70fad remove quick test script for TensorFlow implementation fixes Zchen 2025-10-16 23:05:53 +08:00
  • 7efa33d730 f Zchen 2025-10-16 22:42:33 +08:00
  • 982d2dc256 f Zchen 2025-10-16 22:20:08 +08:00
  • bd61136f93 f Zchen 2025-10-16 22:02:11 +08:00
  • 6f94ad5fae f Zchen 2025-10-16 21:51:43 +08:00
  • eefff1ce5e fix Zchen 2025-10-16 21:40:43 +08:00
  • 426b72ef25 fix Zchen 2025-10-16 21:26:00 +08:00
  • dde6378481 fixed Zchen 2025-10-16 21:13:42 +08:00
  • a0b59c6987 fix Zchen 2025-10-16 21:06:01 +08:00
  • ed6e21bfe4 fix 'NoneType' object has no attribute 'extended' Zchen 2025-10-16 20:57:40 +08:00
  • 1e7077bba7 adamw修复 Zchen 2025-10-16 20:44:55 +08:00
  • c2661550ef 内存泄漏修复 Zchen 2025-10-16 20:26:32 +08:00
  • 1b9e0d9bdf 调整batch_size Zchen 2025-10-16 17:37:59 +08:00
  • be578f2e1d 修复数据加载器低效问题 Zchen 2025-10-16 17:14:06 +08:00
  • a545cc5648 tpu维护 Zchen 2025-10-16 13:39:05 +08:00
  • 5a1e446219 HBM Zchen 2025-10-16 11:42:56 +08:00
  • 0ff6634192 简单修复 Zchen 2025-10-16 10:53:42 +08:00
  • df4a914bbd 小参数 Zchen 2025-10-16 09:22:25 +08:00
  • 25561a7615 超大batch_size Zchen 2025-10-16 01:57:19 +08:00
  • 69a7285886 数据加载器多线程加速 Zchen 2025-10-16 01:17:36 +08:00
  • f84d6254e3 tf 环境问题 Zchen 2025-10-16 00:53:42 +08:00
  • f9d3f47d20 fixed : tf call cuda Zchen 2025-10-15 23:37:24 +08:00
  • 01024678c1 tpu not find Zchen 2025-10-15 23:29:32 +08:00
  • ec8509ad31 fix Zchen 2025-10-15 23:21:06 +08:00
  • 6c400a066c fixed:'str' object has no attribute 'base_dtype' Zchen 2025-10-15 23:13:34 +08:00
  • 83621f91f0 fixed:'str' object has no attribute 'base_dtype' Zchen 2025-10-15 23:11:02 +08:00
  • e8f0308fef tpu Zchen 2025-10-15 20:45:25 +08:00
  • 3b242b908d trainer Zchen 2025-10-15 19:04:42 +08:00
  • 7965f7dbfe TPU Zchen 2025-10-15 16:55:52 +08:00
  • b466e97463 tpu test Zchen 2025-10-15 15:22:13 +08:00
  • 082018cd46 tpu-test Zchen 2025-10-15 15:14:01 +08:00
  • 7bdfc0d257 tpu Zchen 2025-10-15 14:38:56 +08:00
  • e7947f310c tpu Zchen 2025-10-15 14:33:49 +08:00
  • 56fa336af0 tpu Zchen 2025-10-15 14:26:11 +08:00
  • 11ee6ebc51 tpu Zchen 2025-10-15 00:44:08 +08:00
  • 5dcbf28c96 tpu Zchen 2025-10-15 00:30:56 +08:00
  • 9025267400 tpu without bf16 Zchen 2025-10-15 00:25:39 +08:00
  • 603bb12220 tpu Zchen 2025-10-15 00:18:05 +08:00
  • 4a3d3f35ec Merge branch 'dev2' of http://ecs.zchens.cn:3000/zchen/b2txt25 into dev2 Zchen 2025-10-15 00:08:56 +08:00
  • aef96f5646 tpu Zchen 2025-10-14 23:54:53 +08:00
  • ec4f6a25ef tpu Zchen 2025-10-14 23:54:53 +08:00
  • 4b6d680283 tpu Zchen 2025-10-14 23:35:42 +08:00
  • cd52ba51ba tpu Zchen 2025-10-14 23:22:59 +08:00
  • 989ba67618 tpu Zchen 2025-10-14 23:11:54 +08:00
  • f67ed2b820 修复B模型未启用的错误 Zchen 2025-10-14 22:48:28 +08:00
  • 9288bde126 Merge branch 'dev2' of http://ecs.zchens.cn:3000/zchen/b2txt25 into dev2 Zchen 2025-10-14 13:31:28 +08:00
  • 06c4c6c267 tpu Zchen 2025-10-12 23:36:58 +08:00
  • eaa327267f 补充信息 Zchen 2025-10-12 23:36:58 +08:00
  • 0d2a0aa8fa final version? maybe Zchen 2025-10-12 23:36:16 +08:00
  • 6cfc568f9a tpu Zchen 2025-10-12 22:59:45 +08:00
  • 5c941d9efa tpu Zchen 2025-10-12 22:52:38 +08:00
  • 69e3892c27 tpu 多线程编译 Zchen 2025-10-12 22:32:12 +08:00
  • cf1d2b0801 tpu Zchen 2025-10-12 22:14:17 +08:00
  • 0cbb83e052 tpu Zchen 2025-10-12 21:56:34 +08:00
  • 4dad570eea tpu Zchen 2025-10-12 21:47:30 +08:00
  • dfb3f7312c tpu Zchen 2025-10-12 21:43:12 +08:00
  • 6e1d8e18f7 tpu Zchen 2025-10-12 21:36:33 +08:00
  • 580648c058 tpu Zchen 2025-10-12 21:31:07 +08:00
  • 00c94fd48b tpu Zchen 2025-10-12 21:20:08 +08:00
  • c6fc211b00 tpu Zchen 2025-10-12 21:08:15 +08:00
  • a5ff1b4c8e tpu dataloader Zchen 2025-10-12 21:03:54 +08:00
  • 12d571c70b tpu Zchen 2025-10-12 20:59:55 +08:00
  • db6108f250 tpu Zchen 2025-10-12 20:56:08 +08:00
  • 7cc9c41b7f tpu maual dataloader Zchen 2025-10-12 20:43:43 +08:00
  • bc015f5efb tpu Zchen 2025-10-12 20:34:07 +08:00