-
Notifications
You must be signed in to change notification settings - Fork 6
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
10 changed files
with
4,285 additions
and
0 deletions.
There are no files selected for viewing
Large diffs are not rendered by default.
Oops, something went wrong.
Large diffs are not rendered by default.
Oops, something went wrong.
546 changes: 546 additions & 0 deletions
546
log/yelp/pretrain_t5_small_yelp_content_based_modified.log++++++++
Large diffs are not rendered by default.
Oops, something went wrong.
450 changes: 450 additions & 0 deletions
450
log/yelp/pretrain_t5_small_yelp_content_based_norep.log+
Large diffs are not rendered by default.
Oops, something went wrong.
Large diffs are not rendered by default.
Oops, something went wrong.
Large diffs are not rendered by default.
Oops, something went wrong.
521 changes: 521 additions & 0 deletions
521
log/yelp/pretrain_t5_small_yelp_sequential_long_to_short.log+
Large diffs are not rendered by default.
Oops, something went wrong.
217 changes: 217 additions & 0 deletions
217
log/yelp/pretrain_t5_small_yelp_sequential_original.log+
Large diffs are not rendered by default.
Oops, something went wrong.
193 changes: 193 additions & 0 deletions
193
log/yelp/pretrain_t5_small_yelp_sequential_randomize.log+
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,193 @@ | ||
| 04/09/2023, 16:58:13: Namespace(seed=2022, data_dir='data/', logging_dir='log/yelp/pretrain_t5_small_yelp_sequential_randomize.log', model_dir='model/yelp/pretrain_t5_small_yelp_sequential_randomize.pt', task='yelp', max_history=20, sequential_num=10, negative_sample=2, yes_no_sample=5, direct_item_proportion=2, train_sequential_item_batch=64, train_sequential_yesno_batch=32, train_direct_yesno_batch=48, train_direct_candidate_batch=12, train_direct_straightforward_batch=48, meta_title_batch=128, meta_description_batch=12, meta_category_batch=128, meta_epochs=10, meta_lr=0.0001, review_user_batch=16, review_item_batch=16, review_epochs=2, review_lr=0.0001, model_type='t5-small', epochs=20, lr=0.001, clip=1.0, logging_step=1000, warmup_prop=0.05, gradient_accumulation_steps=1, weight_decay=0.01, adam_eps=1e-06, dropout=0.1, alpha=2, multiGPU=True, distributed=True, gpu='0,5', local_rank=0, evaluation_method='sequential_item', evaluation_template_id=0, number_of_items=20034, item_representation='None', cluster_number=55, cluster_size=100, optimal_width_in_CF=False, category_no_repetition=False, last_token_no_repetition=False, hybrid_order='CF_first', data_order='remapped_sequential', remapped_data_order='randomize', resolution=2, max_random_number=30000, min_random_number=1000, overlap=0, base=10, random_initialization_embedding=True, whole_word_embedding='shijie', no_pretrain=False, use_meta_data=False, use_review_data=False, remove_last_item=False, remove_first_item=False, eval_only=False, check_data=False) | ||
| 04/09/2023, 16:58:32: Process Launching at GPU 0 | ||
| 04/09/2023, 16:58:32: Process Launching at GPU 1 | ||
| 04/09/2023, 16:58:33: Building train loader at GPU 0 | ||
| 04/09/2023, 16:58:33: loading data ... | ||
| 04/09/2023, 16:58:33: Building train loader at GPU 1 | ||
| 04/09/2023, 16:58:34: finished loading data | ||
| 04/09/2023, 16:58:34: length of training data is 17583 | ||
| 04/09/2023, 16:58:34: loading model ... | ||
| 04/09/2023, 16:58:34: using only sequential data, and all possible sequences are here | ||
| 04/09/2023, 16:58:34: use pretrained T5 | ||
| 04/09/2023, 16:58:35: randomly initialize number-related embeddings only | ||
| 04/09/2023, 16:58:46: Batch per epoch: 17583 | ||
| 04/09/2023, 16:58:46: Total steps: 351660 | ||
| 04/09/2023, 16:58:46: Warmup proportion: | ||
| 04/09/2023, 16:58:46: Warm up steps: 17583 | ||
| 04/09/2023, 16:58:46: model dataparallel set | ||
| 04/09/2023, 16:58:46: start training | ||
| 04/09/2023, 16:58:46: ---------- training epoch 0 ---------- | ||
| 04/09/2023, 17:04:24: total loss for 1000 steps : 3835.5682282447815 | ||
| 04/09/2023, 17:09:58: total loss for 2000 steps : 2310.6561818122864 | ||
| 04/09/2023, 17:14:25: total loss for 3000 steps : 2210.1801884174347 | ||
| 04/09/2023, 17:19:34: total loss for 4000 steps : 2136.8736176490784 | ||
| 04/09/2023, 17:25:04: total loss for 5000 steps : 2087.160353422165 | ||
| 04/09/2023, 17:30:37: total loss for 6000 steps : 2053.6513986587524 | ||
| 04/09/2023, 17:35:56: total loss for 7000 steps : 2025.9935901165009 | ||
| 04/09/2023, 17:41:04: total loss for 8000 steps : 2009.100499510765 | ||
| 04/09/2023, 17:46:09: total loss for 9000 steps : 1989.7524532079697 | ||
| 04/09/2023, 17:51:16: total loss for 10000 steps : 1975.3841024637222 | ||
| 04/09/2023, 17:56:32: total loss for 11000 steps : 1959.765477180481 | ||
| 04/09/2023, 18:01:28: total loss for 12000 steps : 1942.1314424276352 | ||
| 04/09/2023, 18:06:43: total loss for 13000 steps : 1925.0619105100632 | ||
| 04/09/2023, 18:11:55: total loss for 14000 steps : 1899.207587838173 | ||
| 04/09/2023, 18:16:22: total loss for 15000 steps : 1881.6612648963928 | ||
| 04/09/2023, 18:21:50: total loss for 16000 steps : 1860.0709298849106 | ||
| 04/09/2023, 18:27:43: total loss for 17000 steps : 1842.795482158661 | ||
| 04/09/2023, 18:31:01: ---------- start evaluation after epoch 0 ---------- | ||
| 04/09/2023, 20:25:38: validation hit @ 1 is 0.005060462670872765 | ||
| 04/09/2023, 20:25:38: validation hit @ 5 is 0.018073080967402733 | ||
| 04/09/2023, 20:25:38: validation hit @ 10 is 0.02786540483701367 | ||
| 04/09/2023, 20:25:38: validation ncdg @ 5 is 0.011948187587223255 | ||
| 04/09/2023, 20:25:38: validation ncdg @ 10 is 0.01510155423179728 | ||
| 04/09/2023, 20:38:13: validation hit @ 1 is 0.0044032597266035756 | ||
| 04/09/2023, 20:38:13: validation hit @ 5 is 0.01708727655099895 | ||
| 04/09/2023, 20:38:13: validation hit @ 10 is 0.02641955835962145 | ||
| 04/09/2023, 20:38:13: validation ncdg @ 5 is 0.011150857873012671 | ||
| 04/09/2023, 20:38:13: validation ncdg @ 10 is 0.014128884380907602 | ||
| 04/09/2023, 20:44:37: test hit @ 1 is 0.0053233438485804415 | ||
| 04/09/2023, 20:44:37: test hit @ 5 is 0.01695583596214511 | ||
| 04/09/2023, 20:44:37: test hit @ 10 is 0.025499474237644586 | ||
| 04/09/2023, 20:44:37: test ncdg @ 5 is 0.011457183298331935 | ||
| 04/09/2023, 20:44:37: test ncdg @ 10 is 0.014168458450373022 | ||
| 04/09/2023, 20:44:37: recall increases from 0 ----> 0.02786540483701367 at epoch 0 | ||
| 04/09/2023, 20:44:37: save current best model to best_model/yelp/pretrain_t5_small_yelp_sequential_randomize.pt | ||
| 04/09/2023, 20:46:47: test hit @ 1 is 0.004797581493165089 | ||
| 04/09/2023, 20:46:47: test hit @ 5 is 0.018467402733964248 | ||
| 04/09/2023, 20:46:47: test hit @ 10 is 0.026025236593059938 | ||
| 04/09/2023, 20:46:47: test ncdg @ 5 is 0.01198088309488972 | ||
| 04/09/2023, 20:46:47: test ncdg @ 10 is 0.014438572064991958 | ||
| 04/09/2023, 20:46:47: recall increases from 0 ----> 0.02641955835962145 at epoch 0 | ||
| 04/09/2023, 20:46:47: ---------- training epoch 1 ---------- | ||
| 04/09/2023, 20:47:50: total loss for 18000 steps : 1824.923746228218 | ||
| 04/09/2023, 20:50:19: total loss for 19000 steps : 1807.7785468101501 | ||
| 04/09/2023, 20:52:47: total loss for 20000 steps : 1795.4788318872452 | ||
| 04/09/2023, 20:55:19: total loss for 21000 steps : 1777.045324087143 | ||
| 04/09/2023, 20:57:52: total loss for 22000 steps : 1763.8267604112625 | ||
| 04/09/2023, 21:00:25: total loss for 23000 steps : 1748.7061307430267 | ||
| 04/09/2023, 21:02:59: total loss for 24000 steps : 1737.0215936899185 | ||
| 04/09/2023, 21:05:33: total loss for 25000 steps : 1724.5253168344498 | ||
| 04/09/2023, 21:08:06: total loss for 26000 steps : 1714.2771028280258 | ||
| 04/09/2023, 21:10:40: total loss for 27000 steps : 1705.441875219345 | ||
| 04/09/2023, 21:14:15: total loss for 28000 steps : 1694.000718832016 | ||
| 04/09/2023, 21:19:36: total loss for 29000 steps : 1687.3513606786728 | ||
| 04/09/2023, 21:25:12: total loss for 30000 steps : 1680.9361662864685 | ||
| 04/09/2023, 21:30:47: total loss for 31000 steps : 1671.6695184707642 | ||
| 04/09/2023, 21:36:19: total loss for 32000 steps : 1664.6396279335022 | ||
| 04/09/2023, 21:41:51: total loss for 33000 steps : 1658.289015173912 | ||
| 04/09/2023, 21:47:19: total loss for 34000 steps : 1651.2780312299728 | ||
| 04/09/2023, 21:52:49: total loss for 35000 steps : 1646.9837387800217 | ||
| 04/09/2023, 21:53:43: ---------- start evaluation after epoch 1 ---------- | ||
| 04/09/2023, 22:28:29: validation hit @ 1 is 0.009398002103049422 | ||
| 04/09/2023, 22:28:29: validation hit @ 5 is 0.028391167192429023 | ||
| 04/09/2023, 22:28:29: validation hit @ 10 is 0.0396950578338591 | ||
| 04/09/2023, 22:28:29: validation ncdg @ 5 is 0.019310449768905356 | ||
| 04/09/2023, 22:28:29: validation ncdg @ 10 is 0.02292887505056248 | ||
| 04/09/2023, 23:23:43: test hit @ 1 is 0.007952155625657204 | ||
| 04/09/2023, 23:23:43: test hit @ 5 is 0.028062565720294426 | ||
| 04/09/2023, 23:23:43: test hit @ 10 is 0.03890641430073607 | ||
| 04/09/2023, 23:23:43: test ncdg @ 5 is 0.018238267292797095 | ||
| 04/09/2023, 23:23:43: test ncdg @ 10 is 0.021742485944181128 | ||
| 04/09/2023, 23:23:43: recall increases from 0.02641955835962145 ----> 0.0396950578338591 at epoch 1 | ||
| 04/10/2023, 00:23:44: validation hit @ 1 is 0.009332281808622502 | ||
| 04/10/2023, 00:23:44: validation hit @ 5 is 0.027471083070452155 | ||
| 04/10/2023, 00:23:44: validation hit @ 10 is 0.0410751840168244 | ||
| 04/10/2023, 00:23:44: validation ncdg @ 5 is 0.018777297198331374 | ||
| 04/10/2023, 00:23:44: validation ncdg @ 10 is 0.023160631431622232 | ||
| 04/10/2023, 01:45:13: test hit @ 1 is 0.008937960042060988 | ||
| 04/10/2023, 01:45:13: test hit @ 5 is 0.02983701366982124 | ||
| 04/10/2023, 01:45:13: test hit @ 10 is 0.04173238696109358 | ||
| 04/10/2023, 01:45:13: test ncdg @ 5 is 0.019815193059408945 | ||
| 04/10/2023, 01:45:13: test ncdg @ 10 is 0.023638932672916807 | ||
| 04/10/2023, 01:45:13: recall increases from 0.02786540483701367 ----> 0.0410751840168244 at epoch 1 | ||
| 04/10/2023, 01:45:13: save current best model to best_model/yelp/pretrain_t5_small_yelp_sequential_randomize.pt | ||
| 04/10/2023, 01:45:14: ---------- training epoch 2 ---------- | ||
| 04/10/2023, 01:49:19: total loss for 36000 steps : 1636.5350219011307 | ||
| 04/10/2023, 01:54:53: total loss for 37000 steps : 1628.9545361995697 | ||
| 04/10/2023, 01:59:57: total loss for 38000 steps : 1626.2269533872604 | ||
| 04/10/2023, 02:05:41: total loss for 39000 steps : 1619.9235348701477 | ||
| 04/10/2023, 02:11:16: total loss for 40000 steps : 1613.188186764717 | ||
| 04/10/2023, 02:16:58: total loss for 41000 steps : 1609.055528640747 | ||
| 04/10/2023, 02:22:41: total loss for 42000 steps : 1602.080465555191 | ||
| 04/10/2023, 02:28:30: total loss for 43000 steps : 1601.6773828268051 | ||
| 04/10/2023, 02:34:19: total loss for 44000 steps : 1593.6665391921997 | ||
| 04/10/2023, 02:40:00: total loss for 45000 steps : 1589.210548400879 | ||
| 04/10/2023, 02:45:15: total loss for 46000 steps : 1585.3905110359192 | ||
| 04/10/2023, 02:49:25: total loss for 47000 steps : 1581.6734840869904 | ||
| 04/10/2023, 02:54:19: total loss for 48000 steps : 1574.8261481523514 | ||
| 04/10/2023, 03:00:19: total loss for 49000 steps : 1570.9070399999619 | ||
| 04/10/2023, 03:04:16: total loss for 50000 steps : 1568.7881400585175 | ||
| 04/10/2023, 03:08:10: total loss for 51000 steps : 1561.8046466112137 | ||
| 04/10/2023, 03:12:01: total loss for 52000 steps : 1557.2916151285172 | ||
| 04/10/2023, 03:14:44: ---------- start evaluation after epoch 2 ---------- | ||
| 04/10/2023, 03:43:35: validation hit @ 1 is 0.011303890641430074 | ||
| 04/10/2023, 03:43:35: validation hit @ 5 is 0.03397739221871714 | ||
| 04/10/2023, 03:43:35: validation hit @ 10 is 0.04975026288117771 | ||
| 04/10/2023, 03:43:35: validation ncdg @ 5 is 0.022934379687229965 | ||
| 04/10/2023, 03:43:35: validation ncdg @ 10 is 0.02796724663734019 | ||
| 04/10/2023, 03:52:37: test hit @ 1 is 0.013012618296529969 | ||
| 04/10/2023, 03:52:37: test hit @ 5 is 0.03535751840168244 | ||
| 04/10/2023, 03:52:37: test hit @ 10 is 0.048698738170347006 | ||
| 04/10/2023, 03:52:37: test ncdg @ 5 is 0.02464628593865172 | ||
| 04/10/2023, 03:52:37: test ncdg @ 10 is 0.028950958072227716 | ||
| 04/10/2023, 03:52:37: recall increases from 0.0410751840168244 ----> 0.04975026288117771 at epoch 2 | ||
| 04/10/2023, 03:52:37: save current best model to best_model/yelp/pretrain_t5_small_yelp_sequential_randomize.pt | ||
| 04/10/2023, 05:43:34: validation hit @ 1 is 0.012092534174553101 | ||
| 04/10/2023, 05:43:34: validation hit @ 5 is 0.03430599369085174 | ||
| 04/10/2023, 05:43:34: validation hit @ 10 is 0.04699001051524711 | ||
| 04/10/2023, 05:43:34: validation ncdg @ 5 is 0.023558701955898907 | ||
| 04/10/2023, 05:43:34: validation ncdg @ 10 is 0.02765781762761671 | ||
| 04/10/2023, 06:07:30: test hit @ 1 is 0.009989484752891693 | ||
| 04/10/2023, 06:07:30: test hit @ 5 is 0.03253154574132492 | ||
| 04/10/2023, 06:07:30: test hit @ 10 is 0.04771293375394322 | ||
| 04/10/2023, 06:07:30: test ncdg @ 5 is 0.021738399561017384 | ||
| 04/10/2023, 06:07:30: test ncdg @ 10 is 0.026640575167890163 | ||
| 04/10/2023, 06:07:30: recall increases from 0.0396950578338591 ----> 0.04699001051524711 at epoch 2 | ||
| 04/10/2023, 06:07:30: ---------- training epoch 3 ---------- | ||
| 04/10/2023, 06:08:45: total loss for 53000 steps : 1553.5903067588806 | ||
| 04/10/2023, 06:12:47: total loss for 54000 steps : 1546.18465590477 | ||
| 04/10/2023, 06:17:53: total loss for 55000 steps : 1544.1679309606552 | ||
| 04/10/2023, 06:21:47: total loss for 56000 steps : 1536.844430923462 | ||
| 04/10/2023, 06:24:46: total loss for 57000 steps : 1536.252340555191 | ||
| 04/10/2023, 06:27:50: total loss for 58000 steps : 1529.33762717247 | ||
| 04/10/2023, 06:30:50: total loss for 59000 steps : 1523.3515907526016 | ||
| 04/10/2023, 06:33:53: total loss for 60000 steps : 1521.2324402332306 | ||
| 04/10/2023, 06:37:04: total loss for 61000 steps : 1517.8005579710007 | ||
| 04/10/2023, 06:40:15: total loss for 62000 steps : 1514.4119107723236 | ||
| 04/10/2023, 06:43:27: total loss for 63000 steps : 1510.3928945064545 | ||
| 04/10/2023, 06:46:24: total loss for 64000 steps : 1502.4364550113678 | ||
| 04/10/2023, 06:49:21: total loss for 65000 steps : 1499.4950705766678 | ||
| 04/10/2023, 06:52:21: total loss for 66000 steps : 1493.2867450714111 | ||
| 04/10/2023, 06:55:08: total loss for 67000 steps : 1490.2017785310745 | ||
| 04/10/2023, 06:57:43: total loss for 68000 steps : 1483.8910439014435 | ||
| 04/10/2023, 07:00:28: total loss for 69000 steps : 1480.841351032257 | ||
| 04/10/2023, 07:03:28: total loss for 70000 steps : 1476.737527012825 | ||
| 04/10/2023, 07:26:17: ---------- start evaluation after epoch 3 ---------- | ||
| 04/10/2023, 08:19:19: validation hit @ 1 is 0.010515247108307046 | ||
| 04/10/2023, 08:19:19: validation hit @ 5 is 0.03253154574132492 | ||
| 04/10/2023, 08:19:19: validation hit @ 10 is 0.04777865404837014 | ||
| 04/10/2023, 08:19:19: validation ncdg @ 5 is 0.021682021666896197 | ||
| 04/10/2023, 08:19:19: validation ncdg @ 10 is 0.02658932681583145 | ||
| 04/10/2023, 09:05:43: test hit @ 1 is 0.012026813880126183 | ||
| 04/10/2023, 09:05:43: test hit @ 5 is 0.03496319663512092 | ||
| 04/10/2023, 09:05:43: test hit @ 10 is 0.04679284963196635 | ||
| 04/10/2023, 09:05:43: test ncdg @ 5 is 0.023617123649753657 | ||
| 04/10/2023, 09:05:43: test ncdg @ 10 is 0.02743124568097417 | ||
| 04/10/2023, 09:33:22: validation hit @ 1 is 0.01084384858044164 | ||
| 04/10/2023, 09:33:22: validation hit @ 5 is 0.03194006309148265 | ||
| 04/10/2023, 09:33:22: validation hit @ 10 is 0.04646424815983176 | ||
| 04/10/2023, 09:33:22: validation ncdg @ 5 is 0.02183737869541216 | ||
| 04/10/2023, 09:33:22: validation ncdg @ 10 is 0.026492579755465446 | ||
| 04/10/2023, 10:34:06: test hit @ 1 is 0.010186645636172451 | ||
| 04/10/2023, 10:34:06: test hit @ 5 is 0.02990273396424816 | ||
| 04/10/2023, 10:34:06: test hit @ 10 is 0.045347003154574135 | ||
| 04/10/2023, 10:34:06: test ncdg @ 5 is 0.020233764991760834 | ||
| 04/10/2023, 10:34:06: test ncdg @ 10 is 0.025240475154091436 | ||
| 04/10/2023, 10:34:06: ---------- training epoch 4 ---------- | ||
| 04/10/2023, 10:36:42: total loss for 71000 steps : 1470.8206560611725 | ||
| 04/10/2023, 10:40:03: total loss for 72000 steps : 1466.6910957098007 | ||
| 04/10/2023, 10:43:40: total loss for 73000 steps : 1464.7213594913483 | ||
| 04/10/2023, 10:47:00: total loss for 74000 steps : 1457.4161454439163 | ||
| 04/10/2023, 10:50:04: total loss for 75000 steps : 1453.0070527791977 | ||
| 04/10/2023, 10:52:49: total loss for 76000 steps : 1450.0469020605087 | ||
| 04/10/2023, 10:55:43: total loss for 77000 steps : 1442.1416878700256 |
Oops, something went wrong.