del raysesult

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_6ef1de40_37_batch_size=2048,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=8,drop__2024-06-27_20-46-41/params.json

-{
-  "batch_size": 2048,
-  "decoder_int_ff": 512,
-  "decoder_int_num_layer": 1,
-  "decoder_rt_ff": 1024,
-  "decoder_rt_num_layer": 8,
-  "drop_rate": 0.49537163003492246,
-  "embedding_dim": 256,
-  "encoder_ff": 512,
-  "encoder_num_layer": 8,
-  "lr": 0.001560166878876501,
-  "n_head": 16
-}
\ No newline at end of file

ray_results/test_experiment/train_model_6ef1de40_37_batch_size=2048,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=8,drop__2024-06-27_20-46-41/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
-1972.8930990864912,1719514947,checkpoint_000000,True,False,1,6ef1de40,2024-06-27_21-02-27,281.80140829086304,281.80140829086304,242786,r8i6n8,10.159.28.66,281.80140829086304,1
-1970.5526190329724,1719515215,checkpoint_000001,True,False,2,6ef1de40,2024-06-27_21-06-55,268.26743173599243,550.0688400268555,242786,r8i6n8,10.159.28.66,550.0688400268555,2
-1968.2562842181349,1719515484,checkpoint_000002,True,False,3,6ef1de40,2024-06-27_21-11-24,268.5265476703644,818.5953876972198,242786,r8i6n8,10.159.28.66,818.5953876972198,3
-1975.871436892532,1719515753,checkpoint_000003,True,False,4,6ef1de40,2024-06-27_21-15-53,268.67456102371216,1087.269948720932,242786,r8i6n8,10.159.28.66,1087.269948720932,4
-1968.5231750968874,1719516021,checkpoint_000004,True,False,5,6ef1de40,2024-06-27_21-20-21,268.2965958118439,1355.5665445327759,242786,r8i6n8,10.159.28.66,1355.5665445327759,5
-1970.8189610759105,1719516289,checkpoint_000005,True,False,6,6ef1de40,2024-06-27_21-24-49,267.979775428772,1623.5463199615479,242786,r8i6n8,10.159.28.66,1623.5463199615479,6
-1969.1982075848916,1719516557,checkpoint_000006,True,False,7,6ef1de40,2024-06-27_21-29-17,268.1068859100342,1891.653205871582,242786,r8i6n8,10.159.28.66,1891.653205871582,7
-1983.0505034679502,1719516825,checkpoint_000007,True,False,8,6ef1de40,2024-06-27_21-33-45,267.9620032310486,2159.6152091026306,242786,r8i6n8,10.159.28.66,2159.6152091026306,8
-1978.1448330616388,1719517093,checkpoint_000008,True,False,9,6ef1de40,2024-06-27_21-38-13,267.67944145202637,2427.294650554657,242786,r8i6n8,10.159.28.66,2427.294650554657,9
-1973.7387137826033,1719517360,checkpoint_000009,True,False,10,6ef1de40,2024-06-27_21-42-40,267.4155812263489,2694.710231781006,242786,r8i6n8,10.159.28.66,2694.710231781006,10

ray_results/test_experiment/train_model_6ef1de40_37_batch_size=2048,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=8,drop__2024-06-27_20-46-41/result.json

-{"loss": 1972.8930990864912, "timestamp": 1719514947, "checkpoint_dir_name": "checkpoint_000000", "should_checkpoint": true, "done": false, "training_iteration": 1, "trial_id": "6ef1de40", "date": "2024-06-27_21-02-27", "time_this_iter_s": 281.80140829086304, "time_total_s": 281.80140829086304, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.49537163003492246, "lr": 0.001560166878876501, "batch_size": 2048}, "time_since_restore": 281.80140829086304, "iterations_since_restore": 1}
-{"loss": 1970.5526190329724, "timestamp": 1719515215, "checkpoint_dir_name": "checkpoint_000001", "should_checkpoint": true, "done": false, "training_iteration": 2, "trial_id": "6ef1de40", "date": "2024-06-27_21-06-55", "time_this_iter_s": 268.26743173599243, "time_total_s": 550.0688400268555, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.49537163003492246, "lr": 0.001560166878876501, "batch_size": 2048}, "time_since_restore": 550.0688400268555, "iterations_since_restore": 2}
-{"loss": 1968.2562842181349, "timestamp": 1719515484, "checkpoint_dir_name": "checkpoint_000002", "should_checkpoint": true, "done": false, "training_iteration": 3, "trial_id": "6ef1de40", "date": "2024-06-27_21-11-24", "time_this_iter_s": 268.5265476703644, "time_total_s": 818.5953876972198, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.49537163003492246, "lr": 0.001560166878876501, "batch_size": 2048}, "time_since_restore": 818.5953876972198, "iterations_since_restore": 3}
-{"loss": 1975.871436892532, "timestamp": 1719515753, "checkpoint_dir_name": "checkpoint_000003", "should_checkpoint": true, "done": false, "training_iteration": 4, "trial_id": "6ef1de40", "date": "2024-06-27_21-15-53", "time_this_iter_s": 268.67456102371216, "time_total_s": 1087.269948720932, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.49537163003492246, "lr": 0.001560166878876501, "batch_size": 2048}, "time_since_restore": 1087.269948720932, "iterations_since_restore": 4}
-{"loss": 1968.5231750968874, "timestamp": 1719516021, "checkpoint_dir_name": "checkpoint_000004", "should_checkpoint": true, "done": false, "training_iteration": 5, "trial_id": "6ef1de40", "date": "2024-06-27_21-20-21", "time_this_iter_s": 268.2965958118439, "time_total_s": 1355.5665445327759, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.49537163003492246, "lr": 0.001560166878876501, "batch_size": 2048}, "time_since_restore": 1355.5665445327759, "iterations_since_restore": 5}
-{"loss": 1970.8189610759105, "timestamp": 1719516289, "checkpoint_dir_name": "checkpoint_000005", "should_checkpoint": true, "done": false, "training_iteration": 6, "trial_id": "6ef1de40", "date": "2024-06-27_21-24-49", "time_this_iter_s": 267.979775428772, "time_total_s": 1623.5463199615479, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.49537163003492246, "lr": 0.001560166878876501, "batch_size": 2048}, "time_since_restore": 1623.5463199615479, "iterations_since_restore": 6}
-{"loss": 1969.1982075848916, "timestamp": 1719516557, "checkpoint_dir_name": "checkpoint_000006", "should_checkpoint": true, "done": false, "training_iteration": 7, "trial_id": "6ef1de40", "date": "2024-06-27_21-29-17", "time_this_iter_s": 268.1068859100342, "time_total_s": 1891.653205871582, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.49537163003492246, "lr": 0.001560166878876501, "batch_size": 2048}, "time_since_restore": 1891.653205871582, "iterations_since_restore": 7}
-{"loss": 1983.0505034679502, "timestamp": 1719516825, "checkpoint_dir_name": "checkpoint_000007", "should_checkpoint": true, "done": false, "training_iteration": 8, "trial_id": "6ef1de40", "date": "2024-06-27_21-33-45", "time_this_iter_s": 267.9620032310486, "time_total_s": 2159.6152091026306, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.49537163003492246, "lr": 0.001560166878876501, "batch_size": 2048}, "time_since_restore": 2159.6152091026306, "iterations_since_restore": 8}
-{"loss": 1978.1448330616388, "timestamp": 1719517093, "checkpoint_dir_name": "checkpoint_000008", "should_checkpoint": true, "done": false, "training_iteration": 9, "trial_id": "6ef1de40", "date": "2024-06-27_21-38-13", "time_this_iter_s": 267.67944145202637, "time_total_s": 2427.294650554657, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.49537163003492246, "lr": 0.001560166878876501, "batch_size": 2048}, "time_since_restore": 2427.294650554657, "iterations_since_restore": 9}
-{"loss": 1973.7387137826033, "timestamp": 1719517360, "checkpoint_dir_name": "checkpoint_000009", "should_checkpoint": true, "done": false, "training_iteration": 10, "trial_id": "6ef1de40", "date": "2024-06-27_21-42-40", "time_this_iter_s": 267.4155812263489, "time_total_s": 2694.710231781006, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 8, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.49537163003492246, "lr": 0.001560166878876501, "batch_size": 2048}, "time_since_restore": 2694.710231781006, "iterations_since_restore": 10}

ray_results/test_experiment/train_model_7509f383_13_batch_size=2048,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=2048,decoder_rt_num_layer=1,drop__2024-06-27_14-47-40/params.json

-{
-  "batch_size": 2048,
-  "decoder_int_ff": 512,
-  "decoder_int_num_layer": 1,
-  "decoder_rt_ff": 2048,
-  "decoder_rt_num_layer": 1,
-  "drop_rate": 0.8871532583599777,
-  "embedding_dim": 16,
-  "encoder_ff": 1024,
-  "encoder_num_layer": 8,
-  "lr": 0.038469373500283253,
-  "n_head": 4
-}
\ No newline at end of file

ray_results/test_experiment/train_model_7509f383_13_batch_size=2048,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=2048,decoder_rt_num_layer=1,drop__2024-06-27_14-47-40/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
-2070.736326202633,1719493262,checkpoint_000000,True,False,1,7509f383,2024-06-27_15-01-02,80.20380187034607,80.20380187034607,69992,r8i6n8,10.159.28.66,80.20380187034607,1

ray_results/test_experiment/train_model_7509f383_13_batch_size=2048,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=2048,decoder_rt_num_layer=1,drop__2024-06-27_14-47-40/result.json

-{"loss": 2070.736326202633, "timestamp": 1719493262, "checkpoint_dir_name": "checkpoint_000000", "should_checkpoint": true, "done": false, "training_iteration": 1, "trial_id": "7509f383", "date": "2024-06-27_15-01-02", "time_this_iter_s": 80.20380187034607, "time_total_s": 80.20380187034607, "pid": 69992, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 8, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 16, "encoder_ff": 1024, "decoder_rt_ff": 2048, "decoder_int_ff": 512, "n_head": 4, "drop_rate": 0.8871532583599777, "lr": 0.038469373500283253, "batch_size": 2048}, "time_since_restore": 80.20380187034607, "iterations_since_restore": 1}

ray_results/test_experiment/train_model_769b08ee_61_batch_size=2048,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=512,decoder_rt_num_layer=1,drop_r_2024-06-28_09-00-38/params.json

-{
-  "batch_size": 2048,
-  "decoder_int_ff": 512,
-  "decoder_int_num_layer": 1,
-  "decoder_rt_ff": 512,
-  "decoder_rt_num_layer": 1,
-  "drop_rate": 0.9311098017111933,
-  "embedding_dim": 256,
-  "encoder_ff": 1024,
-  "encoder_num_layer": 4,
-  "lr": 0.0033598362705093534,
-  "n_head": 16
-}
\ No newline at end of file

ray_results/test_experiment/train_model_769b08ee_61_batch_size=2048,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=512,decoder_rt_num_layer=1,drop_r_2024-06-28_09-00-38/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
-2152.0349524790845,1719561170,checkpoint_000000,True,False,1,769b08ee,2024-06-28_09-52-50,112.00384831428528,112.00384831428528,242786,r8i6n8,10.159.28.66,112.00384831428528,1
-2143.4170431302286,1719561268,checkpoint_000001,True,False,2,769b08ee,2024-06-28_09-54-28,97.78639459609985,209.79024291038513,242786,r8i6n8,10.159.28.66,209.79024291038513,2
-2122.2596243310163,1719561366,checkpoint_000002,True,False,3,769b08ee,2024-06-28_09-56-07,98.308185338974,308.09842824935913,242786,r8i6n8,10.159.28.66,308.09842824935913,3
-2108.745537224717,1719561465,checkpoint_000003,True,False,4,769b08ee,2024-06-28_09-57-45,98.25336289405823,406.35179114341736,242786,r8i6n8,10.159.28.66,406.35179114341736,4
-2093.87230772484,1719561563,checkpoint_000004,True,False,5,769b08ee,2024-06-28_09-59-23,98.16895484924316,504.5207459926605,242786,r8i6n8,10.159.28.66,504.5207459926605,5
-2087.8821137195496,1719561661,checkpoint_000005,True,False,6,769b08ee,2024-06-28_10-01-01,98.08770871162415,602.6084547042847,242786,r8i6n8,10.159.28.66,602.6084547042847,6
-2082.8323003813975,1719561759,checkpoint_000006,True,False,7,769b08ee,2024-06-28_10-02-39,97.76127862930298,700.3697333335876,242786,r8i6n8,10.159.28.66,700.3697333335876,7
-2078.3941189022516,1719561857,checkpoint_000007,True,False,8,769b08ee,2024-06-28_10-04-17,98.10805487632751,798.4777882099152,242786,r8i6n8,10.159.28.66,798.4777882099152,8
-2082.373000738189,1719561955,checkpoint_000008,True,False,9,769b08ee,2024-06-28_10-05-56,98.56372618675232,897.0415143966675,242786,r8i6n8,10.159.28.66,897.0415143966675,9
-2067.5648145300197,1719562054,checkpoint_000009,True,False,10,769b08ee,2024-06-28_10-07-34,98.67155456542969,995.7130689620972,242786,r8i6n8,10.159.28.66,995.7130689620972,10

ray_results/test_experiment/train_model_769b08ee_61_batch_size=2048,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=512,decoder_rt_num_layer=1,drop_r_2024-06-28_09-00-38/result.json

-{"loss": 2152.0349524790845, "timestamp": 1719561170, "checkpoint_dir_name": "checkpoint_000000", "should_checkpoint": true, "done": false, "training_iteration": 1, "trial_id": "769b08ee", "date": "2024-06-28_09-52-50", "time_this_iter_s": 112.00384831428528, "time_total_s": 112.00384831428528, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.9311098017111933, "lr": 0.0033598362705093534, "batch_size": 2048}, "time_since_restore": 112.00384831428528, "iterations_since_restore": 1}
-{"loss": 2143.4170431302286, "timestamp": 1719561268, "checkpoint_dir_name": "checkpoint_000001", "should_checkpoint": true, "done": false, "training_iteration": 2, "trial_id": "769b08ee", "date": "2024-06-28_09-54-28", "time_this_iter_s": 97.78639459609985, "time_total_s": 209.79024291038513, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.9311098017111933, "lr": 0.0033598362705093534, "batch_size": 2048}, "time_since_restore": 209.79024291038513, "iterations_since_restore": 2}
-{"loss": 2122.2596243310163, "timestamp": 1719561366, "checkpoint_dir_name": "checkpoint_000002", "should_checkpoint": true, "done": false, "training_iteration": 3, "trial_id": "769b08ee", "date": "2024-06-28_09-56-07", "time_this_iter_s": 98.308185338974, "time_total_s": 308.09842824935913, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.9311098017111933, "lr": 0.0033598362705093534, "batch_size": 2048}, "time_since_restore": 308.09842824935913, "iterations_since_restore": 3}
-{"loss": 2108.745537224717, "timestamp": 1719561465, "checkpoint_dir_name": "checkpoint_000003", "should_checkpoint": true, "done": false, "training_iteration": 4, "trial_id": "769b08ee", "date": "2024-06-28_09-57-45", "time_this_iter_s": 98.25336289405823, "time_total_s": 406.35179114341736, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.9311098017111933, "lr": 0.0033598362705093534, "batch_size": 2048}, "time_since_restore": 406.35179114341736, "iterations_since_restore": 4}
-{"loss": 2093.87230772484, "timestamp": 1719561563, "checkpoint_dir_name": "checkpoint_000004", "should_checkpoint": true, "done": false, "training_iteration": 5, "trial_id": "769b08ee", "date": "2024-06-28_09-59-23", "time_this_iter_s": 98.16895484924316, "time_total_s": 504.5207459926605, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.9311098017111933, "lr": 0.0033598362705093534, "batch_size": 2048}, "time_since_restore": 504.5207459926605, "iterations_since_restore": 5}
-{"loss": 2087.8821137195496, "timestamp": 1719561661, "checkpoint_dir_name": "checkpoint_000005", "should_checkpoint": true, "done": false, "training_iteration": 6, "trial_id": "769b08ee", "date": "2024-06-28_10-01-01", "time_this_iter_s": 98.08770871162415, "time_total_s": 602.6084547042847, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.9311098017111933, "lr": 0.0033598362705093534, "batch_size": 2048}, "time_since_restore": 602.6084547042847, "iterations_since_restore": 6}
-{"loss": 2082.8323003813975, "timestamp": 1719561759, "checkpoint_dir_name": "checkpoint_000006", "should_checkpoint": true, "done": false, "training_iteration": 7, "trial_id": "769b08ee", "date": "2024-06-28_10-02-39", "time_this_iter_s": 97.76127862930298, "time_total_s": 700.3697333335876, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.9311098017111933, "lr": 0.0033598362705093534, "batch_size": 2048}, "time_since_restore": 700.3697333335876, "iterations_since_restore": 7}
-{"loss": 2078.3941189022516, "timestamp": 1719561857, "checkpoint_dir_name": "checkpoint_000007", "should_checkpoint": true, "done": false, "training_iteration": 8, "trial_id": "769b08ee", "date": "2024-06-28_10-04-17", "time_this_iter_s": 98.10805487632751, "time_total_s": 798.4777882099152, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.9311098017111933, "lr": 0.0033598362705093534, "batch_size": 2048}, "time_since_restore": 798.4777882099152, "iterations_since_restore": 8}
-{"loss": 2082.373000738189, "timestamp": 1719561955, "checkpoint_dir_name": "checkpoint_000008", "should_checkpoint": true, "done": false, "training_iteration": 9, "trial_id": "769b08ee", "date": "2024-06-28_10-05-56", "time_this_iter_s": 98.56372618675232, "time_total_s": 897.0415143966675, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.9311098017111933, "lr": 0.0033598362705093534, "batch_size": 2048}, "time_since_restore": 897.0415143966675, "iterations_since_restore": 9}
-{"loss": 2067.5648145300197, "timestamp": 1719562054, "checkpoint_dir_name": "checkpoint_000009", "should_checkpoint": true, "done": false, "training_iteration": 10, "trial_id": "769b08ee", "date": "2024-06-28_10-07-34", "time_this_iter_s": 98.67155456542969, "time_total_s": 995.7130689620972, "pid": 242786, "hostname": "r8i6n8", "node_ip": "10.159.28.66", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 256, "encoder_ff": 1024, "decoder_rt_ff": 512, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.9311098017111933, "lr": 0.0033598362705093534, "batch_size": 2048}, "time_since_restore": 995.7130689620972, "iterations_since_restore": 10}