res

ray_results/test_experiment/train_model_6a62b85a_3_batch_size=2048,decoder_int_ff=1024,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=2,drop__2024-06-25_16-09-34/progress.csv

+loss,timestamp,checkpoint_dir_name,should_checkpoint,done,training_iteration,trial_id,date,time_this_iter_s,time_total_s,pid,hostname,node_ip,time_since_restore,iterations_since_restore
+510.42659056656004,1719324728,checkpoint_000000,True,False,1,6a62b85a,2024-06-25_16-12-08,64.32230615615845,64.32230615615845,69318,r3i5n6,10.159.8.159,64.32230615615845,1
+310.75900604968933,1719327245,checkpoint_000001,True,False,2,6a62b85a,2024-06-25_16-54-05,63.18879699707031,127.51110315322876,130020,r3i5n6,10.159.8.159,63.18879699707031,1

ray_results/test_experiment/train_model_6a62b85a_3_batch_size=2048,decoder_int_ff=1024,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=2,drop__2024-06-25_16-09-34/result.json

+{"loss": 510.42659056656004, "timestamp": 1719324728, "checkpoint_dir_name": "checkpoint_000000", "should_checkpoint": true, "done": false, "training_iteration": 1, "trial_id": "6a62b85a", "date": "2024-06-25_16-12-08", "time_this_iter_s": 64.32230615615845, "time_total_s": 64.32230615615845, "pid": 69318, "hostname": "r3i5n6", "node_ip": "10.159.8.159", "config": {"encoder_num_layer": 1, "decoder_rt_num_layer": 2, "decoder_int_num_layer": 1, "embedding_dim": 16, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 1024, "n_head": 4, "drop_rate": 0.06860896590384413, "lr": 0.00016561766448063182, "batch_size": 2048}, "time_since_restore": 64.32230615615845, "iterations_since_restore": 1}
+{"loss": 310.75900604968933, "timestamp": 1719327245, "checkpoint_dir_name": "checkpoint_000001", "should_checkpoint": true, "done": false, "training_iteration": 2, "trial_id": "6a62b85a", "date": "2024-06-25_16-54-05", "time_this_iter_s": 63.18879699707031, "time_total_s": 127.51110315322876, "pid": 130020, "hostname": "r3i5n6", "node_ip": "10.159.8.159", "config": {"encoder_num_layer": 1, "decoder_rt_num_layer": 2, "decoder_int_num_layer": 1, "embedding_dim": 16, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 1024, "n_head": 4, "drop_rate": 0.06860896590384413, "lr": 0.00016561766448063182, "batch_size": 2048}, "time_since_restore": 63.18879699707031, "iterations_since_restore": 1}

ray_results/test_experiment/train_model_6be79717_24_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=2,drop__2024-06-25_19-14-34/params.json

+{
+  "batch_size": 1024,
+  "decoder_int_ff": 512,
+  "decoder_int_num_layer": 1,
+  "decoder_rt_ff": 1024,
+  "decoder_rt_num_layer": 2,
+  "drop_rate": 0.7446531774853204,
+  "embedding_dim": 64,
+  "encoder_ff": 2048,
+  "encoder_num_layer": 1,
+  "lr": 0.006000257122673396,
+  "n_head": 8
+}
\ No newline at end of file

ray_results/test_experiment/train_model_6be79717_24_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=2,drop__2024-06-25_19-14-34/progress.csv

+loss,timestamp,checkpoint_dir_name,should_checkpoint,done,training_iteration,trial_id,date,time_this_iter_s,time_total_s,pid,hostname,node_ip,time_since_restore,iterations_since_restore
+2020.6805025836613,1719335855,checkpoint_000000,True,False,1,6be79717,2024-06-25_19-17-35,71.47419285774231,71.47419285774231,2195253,r8i6n2,10.159.28.60,71.47419285774231,1
+2003.3902054433747,1719335912,checkpoint_000001,True,False,2,6be79717,2024-06-25_19-18-32,57.25079274177551,128.72498559951782,2195253,r8i6n2,10.159.28.60,128.72498559951782,2

ray_results/test_experiment/train_model_6be79717_24_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=2,drop__2024-06-25_19-14-34/result.json

+{"loss": 2020.6805025836613, "timestamp": 1719335855, "checkpoint_dir_name": "checkpoint_000000", "should_checkpoint": true, "done": false, "training_iteration": 1, "trial_id": "6be79717", "date": "2024-06-25_19-17-35", "time_this_iter_s": 71.47419285774231, "time_total_s": 71.47419285774231, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 1, "decoder_rt_num_layer": 2, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 2048, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 8, "drop_rate": 0.7446531774853204, "lr": 0.006000257122673396, "batch_size": 1024}, "time_since_restore": 71.47419285774231, "iterations_since_restore": 1}
+{"loss": 2003.3902054433747, "timestamp": 1719335912, "checkpoint_dir_name": "checkpoint_000001", "should_checkpoint": true, "done": false, "training_iteration": 2, "trial_id": "6be79717", "date": "2024-06-25_19-18-32", "time_this_iter_s": 57.25079274177551, "time_total_s": 128.72498559951782, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 1, "decoder_rt_num_layer": 2, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 2048, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 8, "drop_rate": 0.7446531774853204, "lr": 0.006000257122673396, "batch_size": 1024}, "time_since_restore": 128.72498559951782, "iterations_since_restore": 2}

ray_results/test_experiment/train_model_6c8b476a_20_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=512,decoder_rt_num_layer=4,drop_r_2024-06-25_18-43-49/params.json

+{
+  "batch_size": 1024,
+  "decoder_int_ff": 512,
+  "decoder_int_num_layer": 1,
+  "decoder_rt_ff": 512,
+  "decoder_rt_num_layer": 4,
+  "drop_rate": 0.5088357248512263,
+  "embedding_dim": 64,
+  "encoder_ff": 1024,
+  "encoder_num_layer": 8,
+  "lr": 0.010524289583233754,
+  "n_head": 16
+}
\ No newline at end of file

ray_results/test_experiment/train_model_6c8b476a_20_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=512,decoder_rt_num_layer=4,drop_r_2024-06-25_18-43-49/progress.csv

+loss,timestamp,checkpoint_dir_name,should_checkpoint,done,training_iteration,trial_id,date,time_this_iter_s,time_total_s,pid,hostname,node_ip,time_since_restore,iterations_since_restore
+2006.8072533795214,1719334267,checkpoint_000000,True,False,1,6c8b476a,2024-06-25_18-51-07,127.51717782020569,127.51717782020569,2195253,r8i6n2,10.159.28.60,127.51717782020569,1
+1995.246540219765,1719334380,checkpoint_000001,True,False,2,6c8b476a,2024-06-25_18-53-00,113.46883034706116,240.98600816726685,2195253,r8i6n2,10.159.28.60,240.98600816726685,2
+1996.2762984628753,1719338625,checkpoint_000002,True,False,3,6c8b476a,2024-06-25_20-03-45,126.75374698638916,367.739755153656,2195253,r8i6n2,10.159.28.60,126.75374698638916,1
+1991.5076505405696,1719338738,checkpoint_000003,True,False,4,6c8b476a,2024-06-25_20-05-38,113.23560690879822,480.9753620624542,2195253,r8i6n2,10.159.28.60,239.98935389518738,2

ray_results/test_experiment/train_model_6c8b476a_20_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=512,decoder_rt_num_layer=4,drop_r_2024-06-25_18-43-49/result.json

+{"loss": 2006.8072533795214, "timestamp": 1719334267, "checkpoint_dir_name": "checkpoint_000000", "should_checkpoint": true, "done": false, "training_iteration": 1, "trial_id": "6c8b476a", "date": "2024-06-25_18-51-07", "time_this_iter_s": 127.51717782020569, "time_total_s": 127.51717782020569, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 4, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.5088357248512263, "lr": 0.010524289583233754, "batch_size": 1024}, "time_since_restore": 127.51717782020569, "iterations_since_restore": 1}
+{"loss": 1995.246540219765, "timestamp": 1719334380, "checkpoint_dir_name": "checkpoint_000001", "should_checkpoint": true, "done": false, "training_iteration": 2, "trial_id": "6c8b476a", "date": "2024-06-25_18-53-00", "time_this_iter_s": 113.46883034706116, "time_total_s": 240.98600816726685, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 4, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.5088357248512263, "lr": 0.010524289583233754, "batch_size": 1024}, "time_since_restore": 240.98600816726685, "iterations_since_restore": 2}
+{"loss": 1996.2762984628753, "timestamp": 1719338625, "checkpoint_dir_name": "checkpoint_000002", "should_checkpoint": true, "done": false, "training_iteration": 3, "trial_id": "6c8b476a", "date": "2024-06-25_20-03-45", "time_this_iter_s": 126.75374698638916, "time_total_s": 367.739755153656, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 4, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.5088357248512263, "lr": 0.010524289583233754, "batch_size": 1024}, "time_since_restore": 126.75374698638916, "iterations_since_restore": 1}
+{"loss": 1991.5076505405696, "timestamp": 1719338738, "checkpoint_dir_name": "checkpoint_000003", "should_checkpoint": true, "done": false, "training_iteration": 4, "trial_id": "6c8b476a", "date": "2024-06-25_20-05-38", "time_this_iter_s": 113.23560690879822, "time_total_s": 480.9753620624542, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 4, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.5088357248512263, "lr": 0.010524289583233754, "batch_size": 1024}, "time_since_restore": 239.98935389518738, "iterations_since_restore": 2}

ray_results/test_experiment/train_model_7ddcd025_3_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=512,decoder_rt_num_layer=8,drop_ra_2024-06-25_17-14-16/params.json

+{
+  "batch_size": 1024,
+  "decoder_int_ff": 512,
+  "decoder_int_num_layer": 1,
+  "decoder_rt_ff": 512,
+  "decoder_rt_num_layer": 8,
+  "drop_rate": 0.474834603771515,
+  "embedding_dim": 64,
+  "encoder_ff": 2048,
+  "encoder_num_layer": 2,
+  "lr": 0.015369204694575145,
+  "n_head": 1
+}
\ No newline at end of file

ray_results/test_experiment/train_model_7ddcd025_3_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=512,decoder_rt_num_layer=8,drop_ra_2024-06-25_17-14-16/progress.csv

+loss,timestamp,checkpoint_dir_name,should_checkpoint,done,training_iteration,trial_id,date,time_this_iter_s,time_total_s,pid,hostname,node_ip,time_since_restore,iterations_since_restore
+2079.0407700425994,1719328617,checkpoint_000000,True,False,1,7ddcd025,2024-06-25_17-16-57,90.59219408035278,90.59219408035278,2195253,r8i6n2,10.159.28.60,90.59219408035278,1

ray_results/test_experiment/train_model_7ddcd025_3_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=512,decoder_rt_num_layer=8,drop_ra_2024-06-25_17-14-16/result.json

+{"loss": 2079.0407700425994, "timestamp": 1719328617, "checkpoint_dir_name": "checkpoint_000000", "should_checkpoint": true, "done": false, "training_iteration": 1, "trial_id": "7ddcd025", "date": "2024-06-25_17-16-57", "time_this_iter_s": 90.59219408035278, "time_total_s": 90.59219408035278, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 2, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 2048, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 1, "drop_rate": 0.474834603771515, "lr": 0.015369204694575145, "batch_size": 1024}, "time_since_restore": 90.59219408035278, "iterations_since_restore": 1}

ray_results/test_experiment/train_model_7dfe67c7_4_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=2048,decoder_rt_num_layer=2,drop_r_2024-06-25_17-15-26/params.json

+{
+  "batch_size": 1024,
+  "decoder_int_ff": 512,
+  "decoder_int_num_layer": 1,
+  "decoder_rt_ff": 2048,
+  "decoder_rt_num_layer": 2,
+  "drop_rate": 0.6066107852445892,
+  "embedding_dim": 256,
+  "encoder_ff": 2048,
+  "encoder_num_layer": 2,
+  "lr": 0.0011313709304158431,
+  "n_head": 2
+}
\ No newline at end of file