sled_mt5-base-7 / trainer_state.json
rusheeliyer's picture
Upload 14 files
9a57d9d
raw
history blame
7.84 kB
{
"best_metric": 28.7934,
"best_model_checkpoint": "models/sled/output_sled_mt5_base2/checkpoint-11985",
"epoch": 7.0,
"global_step": 16779,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.21,
"learning_rate": 1.3904338153503894e-05,
"loss": 18.6448,
"step": 500
},
{
"epoch": 0.42,
"learning_rate": 2.7808676307007788e-05,
"loss": 3.8147,
"step": 1000
},
{
"epoch": 0.63,
"learning_rate": 4.171301446051168e-05,
"loss": 2.6096,
"step": 1500
},
{
"epoch": 0.83,
"learning_rate": 5.5617352614015575e-05,
"loss": 2.3758,
"step": 2000
},
{
"epoch": 1.0,
"eval_mean_prediction_length_characters": 7800.0,
"eval_mean_prediction_length_tokens": 2047.0,
"eval_num_predicted": 1,
"eval_rouge/geometric_mean": 14.0723,
"eval_rouge/rouge1": 24.3546,
"eval_rouge/rouge2": 7.1185,
"eval_rouge/rougeL": 16.074,
"eval_rouge/rougeLsum": 21.4321,
"eval_runtime": 358.1416,
"eval_samples_per_second": 0.003,
"eval_steps_per_second": 0.003,
"step": 2397
},
{
"epoch": 1.04,
"learning_rate": 6.952169076751947e-05,
"loss": 2.1056,
"step": 2500
},
{
"epoch": 1.25,
"learning_rate": 8.342602892102336e-05,
"loss": 2.0678,
"step": 3000
},
{
"epoch": 1.46,
"learning_rate": 9.733036707452726e-05,
"loss": 1.9603,
"step": 3500
},
{
"epoch": 1.67,
"learning_rate": 9.87515065360487e-05,
"loss": 1.8989,
"step": 4000
},
{
"epoch": 1.88,
"learning_rate": 9.720634135789116e-05,
"loss": 1.8726,
"step": 4500
},
{
"epoch": 2.0,
"eval_mean_prediction_length_characters": 7920.0,
"eval_mean_prediction_length_tokens": 2047.0,
"eval_num_predicted": 1,
"eval_rouge/geometric_mean": 9.5424,
"eval_rouge/rouge1": 14.5332,
"eval_rouge/rouge2": 5.973,
"eval_rouge/rougeL": 10.0096,
"eval_rouge/rougeLsum": 13.9557,
"eval_runtime": 361.8939,
"eval_samples_per_second": 0.003,
"eval_steps_per_second": 0.003,
"step": 4794
},
{
"epoch": 2.09,
"learning_rate": 9.566117617973361e-05,
"loss": 1.7568,
"step": 5000
},
{
"epoch": 2.29,
"learning_rate": 9.411601100157608e-05,
"loss": 1.7487,
"step": 5500
},
{
"epoch": 2.5,
"learning_rate": 9.257084582341852e-05,
"loss": 1.7171,
"step": 6000
},
{
"epoch": 2.71,
"learning_rate": 9.102568064526098e-05,
"loss": 1.7211,
"step": 6500
},
{
"epoch": 2.92,
"learning_rate": 8.948051546710344e-05,
"loss": 1.6807,
"step": 7000
},
{
"epoch": 3.0,
"eval_mean_prediction_length_characters": 7675.0,
"eval_mean_prediction_length_tokens": 2047.0,
"eval_num_predicted": 1,
"eval_rouge/geometric_mean": 13.6689,
"eval_rouge/rouge1": 24.0,
"eval_rouge/rouge2": 7.1289,
"eval_rouge/rougeL": 14.9268,
"eval_rouge/rougeLsum": 21.2683,
"eval_runtime": 361.2079,
"eval_samples_per_second": 0.003,
"eval_steps_per_second": 0.003,
"step": 7191
},
{
"epoch": 3.13,
"learning_rate": 8.79353502889459e-05,
"loss": 1.6141,
"step": 7500
},
{
"epoch": 3.34,
"learning_rate": 8.639018511078834e-05,
"loss": 1.6026,
"step": 8000
},
{
"epoch": 3.55,
"learning_rate": 8.48450199326308e-05,
"loss": 1.616,
"step": 8500
},
{
"epoch": 3.75,
"learning_rate": 8.329985475447326e-05,
"loss": 1.6012,
"step": 9000
},
{
"epoch": 3.96,
"learning_rate": 8.175468957631572e-05,
"loss": 1.6069,
"step": 9500
},
{
"epoch": 4.0,
"eval_mean_prediction_length_characters": 9281.0,
"eval_mean_prediction_length_tokens": 2047.0,
"eval_num_predicted": 1,
"eval_rouge/geometric_mean": 16.6758,
"eval_rouge/rouge1": 26.9307,
"eval_rouge/rouge2": 10.6046,
"eval_rouge/rougeL": 16.2376,
"eval_rouge/rougeLsum": 24.5545,
"eval_runtime": 363.6434,
"eval_samples_per_second": 0.003,
"eval_steps_per_second": 0.003,
"step": 9588
},
{
"epoch": 4.17,
"learning_rate": 8.020952439815816e-05,
"loss": 1.5703,
"step": 10000
},
{
"epoch": 4.38,
"learning_rate": 7.866435922000062e-05,
"loss": 1.5467,
"step": 10500
},
{
"epoch": 4.59,
"learning_rate": 7.711919404184308e-05,
"loss": 1.5567,
"step": 11000
},
{
"epoch": 4.8,
"learning_rate": 7.557402886368554e-05,
"loss": 1.5021,
"step": 11500
},
{
"epoch": 5.0,
"eval_mean_prediction_length_characters": 7806.0,
"eval_mean_prediction_length_tokens": 2047.0,
"eval_num_predicted": 1,
"eval_rouge/geometric_mean": 28.7934,
"eval_rouge/rouge1": 48.5567,
"eval_rouge/rouge2": 21.775,
"eval_rouge/rougeL": 22.5773,
"eval_rouge/rougeLsum": 46.1856,
"eval_runtime": 395.7669,
"eval_samples_per_second": 0.003,
"eval_steps_per_second": 0.003,
"step": 11985
},
{
"epoch": 5.01,
"learning_rate": 7.402886368552798e-05,
"loss": 1.5094,
"step": 12000
},
{
"epoch": 5.21,
"learning_rate": 7.248369850737045e-05,
"loss": 1.4837,
"step": 12500
},
{
"epoch": 5.42,
"learning_rate": 7.09385333292129e-05,
"loss": 1.4792,
"step": 13000
},
{
"epoch": 5.63,
"learning_rate": 6.939336815105535e-05,
"loss": 1.4671,
"step": 13500
},
{
"epoch": 5.84,
"learning_rate": 6.78482029728978e-05,
"loss": 1.4644,
"step": 14000
},
{
"epoch": 6.0,
"eval_mean_prediction_length_characters": 7814.0,
"eval_mean_prediction_length_tokens": 2047.0,
"eval_num_predicted": 1,
"eval_rouge/geometric_mean": 17.4993,
"eval_rouge/rouge1": 30.9627,
"eval_rouge/rouge2": 10.46,
"eval_rouge/rougeL": 16.5457,
"eval_rouge/rougeLsum": 28.7373,
"eval_runtime": 421.5615,
"eval_samples_per_second": 0.002,
"eval_steps_per_second": 0.002,
"step": 14382
},
{
"epoch": 6.05,
"learning_rate": 6.630303779474027e-05,
"loss": 1.4353,
"step": 14500
},
{
"epoch": 6.26,
"learning_rate": 6.475787261658271e-05,
"loss": 1.4384,
"step": 15000
},
{
"epoch": 6.47,
"learning_rate": 6.321270743842517e-05,
"loss": 1.439,
"step": 15500
},
{
"epoch": 6.68,
"learning_rate": 6.166754226026763e-05,
"loss": 1.4215,
"step": 16000
},
{
"epoch": 6.88,
"learning_rate": 6.012237708211008e-05,
"loss": 1.4284,
"step": 16500
},
{
"epoch": 7.0,
"eval_mean_prediction_length_characters": 7701.0,
"eval_mean_prediction_length_tokens": 2047.0,
"eval_num_predicted": 1,
"eval_rouge/geometric_mean": 13.7624,
"eval_rouge/rouge1": 25.5464,
"eval_rouge/rouge2": 7.1949,
"eval_rouge/rougeL": 14.1816,
"eval_rouge/rougeLsum": 23.2152,
"eval_runtime": 798.1174,
"eval_samples_per_second": 0.001,
"eval_steps_per_second": 0.001,
"step": 16779
}
],
"max_steps": 35955,
"num_train_epochs": 15,
"total_flos": 5.646514808583475e+17,
"trial_name": null,
"trial_params": null
}