res

ray_results/test_experiment/train_model_2465ee2c_11_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_17-34-53/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": 0.3089996310219745,
+  "embedding_dim": 64,
+  "encoder_ff": 2048,
+  "encoder_num_layer": 4,
+  "lr": 0.062282868368169514,
+  "n_head": 8
+}
\ No newline at end of file

ray_results/test_experiment/train_model_2465ee2c_11_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_17-34-53/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
+2699.211756428396,1719329892,checkpoint_000000,True,False,1,2465ee2c,2024-06-25_17-38-12,85.10484099388123,85.10484099388123,2195253,r8i6n2,10.159.28.60,85.10484099388123,1

ray_results/test_experiment/train_model_2465ee2c_11_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_17-34-53/result.json

+{"loss": 2699.211756428396, "timestamp": 1719329892, "checkpoint_dir_name": "checkpoint_000000", "should_checkpoint": true, "done": false, "training_iteration": 1, "trial_id": "2465ee2c", "date": "2024-06-25_17-38-12", "time_this_iter_s": 85.10484099388123, "time_total_s": 85.10484099388123, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 4, "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": 0.3089996310219745, "lr": 0.062282868368169514, "batch_size": 1024}, "time_since_restore": 85.10484099388123, "iterations_since_restore": 1}

ray_results/test_experiment/train_model_25a9f43a_13_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=1,drop__2024-06-25_17-38-12/params.json

+{
+  "batch_size": 1024,
+  "decoder_int_ff": 512,
+  "decoder_int_num_layer": 1,
+  "decoder_rt_ff": 1024,
+  "decoder_rt_num_layer": 1,
+  "drop_rate": 0.7474706763430264,
+  "embedding_dim": 64,
+  "encoder_ff": 512,
+  "encoder_num_layer": 4,
+  "lr": 0.00045083297498032556,
+  "n_head": 16
+}
\ No newline at end of file

ray_results/test_experiment/train_model_25a9f43a_13_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=1,drop__2024-06-25_17-38-12/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
+1284.8956039308564,1719330178,checkpoint_000000,True,False,1,25a9f43a,2024-06-25_17-42-59,79.39262270927429,79.39262270927429,2195253,r8i6n2,10.159.28.60,79.39262270927429,1
+517.7363889198604,1719330947,checkpoint_000001,True,False,2,25a9f43a,2024-06-25_17-55-47,77.99471831321716,157.38734102249146,2195253,r8i6n2,10.159.28.60,77.99471831321716,1
+424.9116310660295,1719331554,checkpoint_000002,True,False,3,25a9f43a,2024-06-25_18-05-54,94.57213068008423,251.95947170257568,2195253,r8i6n2,10.159.28.60,94.57213068008423,1
+381.2108863169753,1719331634,checkpoint_000003,True,False,4,25a9f43a,2024-06-25_18-07-14,80.07035231590271,332.0298240184784,2195253,r8i6n2,10.159.28.60,174.64248299598694,2

ray_results/test_experiment/train_model_25a9f43a_13_batch_size=1024,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=1024,decoder_rt_num_layer=1,drop__2024-06-25_17-38-12/result.json

+{"loss": 1284.8956039308564, "timestamp": 1719330178, "checkpoint_dir_name": "checkpoint_000000", "should_checkpoint": true, "done": false, "training_iteration": 1, "trial_id": "25a9f43a", "date": "2024-06-25_17-42-59", "time_this_iter_s": 79.39262270927429, "time_total_s": 79.39262270927429, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.7474706763430264, "lr": 0.00045083297498032556, "batch_size": 1024}, "time_since_restore": 79.39262270927429, "iterations_since_restore": 1}
+{"loss": 517.7363889198604, "timestamp": 1719330947, "checkpoint_dir_name": "checkpoint_000001", "should_checkpoint": true, "done": false, "training_iteration": 2, "trial_id": "25a9f43a", "date": "2024-06-25_17-55-47", "time_this_iter_s": 77.99471831321716, "time_total_s": 157.38734102249146, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.7474706763430264, "lr": 0.00045083297498032556, "batch_size": 1024}, "time_since_restore": 77.99471831321716, "iterations_since_restore": 1}
+{"loss": 424.9116310660295, "timestamp": 1719331554, "checkpoint_dir_name": "checkpoint_000002", "should_checkpoint": true, "done": false, "training_iteration": 3, "trial_id": "25a9f43a", "date": "2024-06-25_18-05-54", "time_this_iter_s": 94.57213068008423, "time_total_s": 251.95947170257568, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.7474706763430264, "lr": 0.00045083297498032556, "batch_size": 1024}, "time_since_restore": 94.57213068008423, "iterations_since_restore": 1}
+{"loss": 381.2108863169753, "timestamp": 1719331634, "checkpoint_dir_name": "checkpoint_000003", "should_checkpoint": true, "done": false, "training_iteration": 4, "trial_id": "25a9f43a", "date": "2024-06-25_18-07-14", "time_this_iter_s": 80.07035231590271, "time_total_s": 332.0298240184784, "pid": 2195253, "hostname": "r8i6n2", "node_ip": "10.159.28.60", "config": {"encoder_num_layer": 4, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 1, "embedding_dim": 64, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 512, "n_head": 16, "drop_rate": 0.7474706763430264, "lr": 0.00045083297498032556, "batch_size": 1024}, "time_since_restore": 174.64248299598694, "iterations_since_restore": 2}

ray_results/test_experiment/train_model_25b42194_15_batch_size=2048,decoder_int_ff=2048,decoder_int_num_layer=8,decoder_rt_ff=1024,decoder_rt_num_layer=1,drop_2024-06-25_16-42-52/params.json

+{
+  "batch_size": 2048,
+  "decoder_int_ff": 2048,
+  "decoder_int_num_layer": 8,
+  "decoder_rt_ff": 1024,
+  "decoder_rt_num_layer": 1,
+  "drop_rate": 0.3256301932631026,
+  "embedding_dim": 1024,
+  "encoder_ff": 512,
+  "encoder_num_layer": 1,
+  "lr": 0.0018178629337167497,
+  "n_head": 16
+}
\ No newline at end of file

ray_results/test_experiment/train_model_25b42194_15_batch_size=2048,decoder_int_ff=2048,decoder_int_num_layer=8,decoder_rt_ff=1024,decoder_rt_num_layer=1,drop_2024-06-25_16-42-52/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
+5061.415592704232,1719326792,checkpoint_000000,True,False,1,25b42194,2024-06-25_16-46-32,194.36064219474792,194.36064219474792,122466,r3i5n6,10.159.8.159,194.36064219474792,1

ray_results/test_experiment/train_model_25b42194_15_batch_size=2048,decoder_int_ff=2048,decoder_int_num_layer=8,decoder_rt_ff=1024,decoder_rt_num_layer=1,drop_2024-06-25_16-42-52/result.json

+{"loss": 5061.415592704232, "timestamp": 1719326792, "checkpoint_dir_name": "checkpoint_000000", "should_checkpoint": true, "done": false, "training_iteration": 1, "trial_id": "25b42194", "date": "2024-06-25_16-46-32", "time_this_iter_s": 194.36064219474792, "time_total_s": 194.36064219474792, "pid": 122466, "hostname": "r3i5n6", "node_ip": "10.159.8.159", "config": {"encoder_num_layer": 1, "decoder_rt_num_layer": 1, "decoder_int_num_layer": 8, "embedding_dim": 1024, "encoder_ff": 512, "decoder_rt_ff": 1024, "decoder_int_ff": 2048, "n_head": 16, "drop_rate": 0.3256301932631026, "lr": 0.0018178629337167497, "batch_size": 2048}, "time_since_restore": 194.36064219474792, "iterations_since_restore": 1}

ray_results/test_experiment/train_model_3510e84b_15_batch_size=2048,decoder_int_ff=512,decoder_int_num_layer=1,decoder_rt_ff=2048,decoder_rt_num_layer=1,drop__2024-06-25_17-42-59/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": 0.3408951993486181,
+  "embedding_dim": 16,
+  "encoder_ff": 2048,
+  "encoder_num_layer": 4,
+  "lr": 0.009004430142657121,
+  "n_head": 4
+}
\ No newline at end of file