MarkushGrapher-2 / trainer_state.json
TimStrohmeyer's picture
Upload folder using huggingface_hub
95464de verified
Invalid JSON: Unexpected token 'N', ..."du_ar_m": NaN, "... is not valid JSON
{
"best_metric": 0.534,
"best_model_checkpoint": "./models/12_44_AM_November_04_2025/checkpoint-1000",
"epoch": 0.0263866167080057,
"eval_steps": 1000,
"global_step": 1000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 2.5e-06,
"loss": 0.1033,
"step": 50
},
{
"epoch": 0.0,
"learning_rate": 5e-06,
"loss": 0.1094,
"step": 100
},
{
"epoch": 0.0,
"learning_rate": 7.5e-06,
"loss": 0.0969,
"step": 150
},
{
"epoch": 0.01,
"learning_rate": 1e-05,
"loss": 0.1012,
"step": 200
},
{
"epoch": 0.01,
"learning_rate": 1.25e-05,
"loss": 0.0892,
"step": 250
},
{
"epoch": 0.01,
"learning_rate": 1.5e-05,
"loss": 0.0773,
"step": 300
},
{
"epoch": 0.01,
"learning_rate": 1.75e-05,
"loss": 0.0871,
"step": 350
},
{
"epoch": 0.01,
"learning_rate": 2e-05,
"loss": 0.0629,
"step": 400
},
{
"epoch": 0.01,
"learning_rate": 2.25e-05,
"loss": 0.0733,
"step": 450
},
{
"epoch": 0.01,
"learning_rate": 2.5e-05,
"loss": 0.0682,
"step": 500
},
{
"epoch": 0.01,
"learning_rate": 2.7500000000000004e-05,
"loss": 0.0731,
"step": 550
},
{
"epoch": 0.02,
"learning_rate": 3e-05,
"loss": 0.0639,
"step": 600
},
{
"epoch": 0.02,
"learning_rate": 3.2500000000000004e-05,
"loss": 0.0694,
"step": 650
},
{
"epoch": 0.02,
"learning_rate": 3.5e-05,
"loss": 0.0755,
"step": 700
},
{
"epoch": 0.02,
"learning_rate": 3.7500000000000003e-05,
"loss": 0.0724,
"step": 750
},
{
"epoch": 0.02,
"learning_rate": 4e-05,
"loss": 0.0655,
"step": 800
},
{
"epoch": 0.02,
"learning_rate": 4.25e-05,
"loss": 0.0736,
"step": 850
},
{
"epoch": 0.02,
"learning_rate": 4.5e-05,
"loss": 0.075,
"step": 900
},
{
"epoch": 0.03,
"learning_rate": 4.75e-05,
"loss": 0.0653,
"step": 950
},
{
"epoch": 0.03,
"learning_rate": 5e-05,
"loss": 0.058,
"step": 1000
},
{
"epoch": 0.03,
"eval_mdu_accuracy": 0.9150433120501037,
"eval_mdu_ar_cxsmi_equality": 0.712,
"eval_mdu_ar_inchi_equality": 0.8,
"eval_mdu_ar_is_in_training": 0.0,
"eval_mdu_ar_m": NaN,
"eval_mdu_ar_m_size": 0,
"eval_mdu_ar_markush_equality": 0.656,
"eval_mdu_ar_r": 0.9,
"eval_mdu_ar_r_size": 107,
"eval_mdu_ar_sg": 0.574,
"eval_mdu_ar_sg_size": 28,
"eval_mdu_ar_size": 125,
"eval_mdu_ar_stable_equality": 0.912,
"eval_mdu_ar_stable_precision": 0.963,
"eval_mdu_ar_stable_recall": 0.952,
"eval_mdu_ar_string_equality": 0.728,
"eval_mdu_ar_string_equality_opt": 0.648,
"eval_mdu_ar_tanimoto": 0.948,
"eval_mdu_ar_valid": 0.976,
"eval_mdu_invalid_gt": 0.0,
"eval_mdu_iou": 0,
"eval_mdu_loss": 0.022680645808577538,
"eval_mdu_lum_test_ar_cxsmi_equality": 0.534,
"eval_mdu_lum_test_ar_inchi_equality": 0.68,
"eval_mdu_lum_test_ar_is_in_training": 0.0,
"eval_mdu_lum_test_ar_m": NaN,
"eval_mdu_lum_test_ar_m_size": 0,
"eval_mdu_lum_test_ar_markush_equality": 0.126,
"eval_mdu_lum_test_ar_r": 0.798,
"eval_mdu_lum_test_ar_r_size": 100,
"eval_mdu_lum_test_ar_sg": 0.346,
"eval_mdu_lum_test_ar_sg_size": 26,
"eval_mdu_lum_test_ar_size": 103,
"eval_mdu_lum_test_ar_stable_equality": 0.223,
"eval_mdu_lum_test_ar_stable_precision": 0.668,
"eval_mdu_lum_test_ar_stable_recall": 0.624,
"eval_mdu_lum_test_ar_string_equality": 0.136,
"eval_mdu_lum_test_ar_string_equality_opt": 0.126,
"eval_mdu_lum_test_ar_tanimoto": 0.839,
"eval_mdu_lum_test_ar_valid": 0.932,
"eval_mdu_lum_test_invalid_gt": 0.0,
"eval_mdu_mae": 0,
"eval_mdu_runtime": 7938.8722,
"eval_mdu_samples_per_second": 2.967,
"eval_mdu_steps_per_second": 0.594,
"eval_mdu_uspto_clean_ar_cxsmi_equality": 0.917,
"eval_mdu_uspto_clean_ar_inchi_equality": 0.927,
"eval_mdu_uspto_clean_ar_is_in_training": 0.0,
"eval_mdu_uspto_clean_ar_m": NaN,
"eval_mdu_uspto_clean_ar_m_size": 0,
"eval_mdu_uspto_clean_ar_markush_equality": 0.917,
"eval_mdu_uspto_clean_ar_r": 0.778,
"eval_mdu_uspto_clean_ar_r_size": 9,
"eval_mdu_uspto_clean_ar_sg": NaN,
"eval_mdu_uspto_clean_ar_sg_size": 0,
"eval_mdu_uspto_clean_ar_size": 100,
"eval_mdu_uspto_clean_ar_stable_equality": 1.0,
"eval_mdu_uspto_clean_ar_stable_precision": 1.0,
"eval_mdu_uspto_clean_ar_stable_recall": 1.0,
"eval_mdu_uspto_clean_ar_string_equality": 0.896,
"eval_mdu_uspto_clean_ar_string_equality_opt": 0.844,
"eval_mdu_uspto_clean_ar_tanimoto": 0.961,
"eval_mdu_uspto_clean_ar_valid": 0.99,
"eval_mdu_uspto_clean_invalid_gt": 0.04,
"eval_mdu_uspto_markush_ar_cxsmi_equality": 0.5,
"eval_mdu_uspto_markush_ar_inchi_equality": 0.743,
"eval_mdu_uspto_markush_ar_is_in_training": 0.0,
"eval_mdu_uspto_markush_ar_m": NaN,
"eval_mdu_uspto_markush_ar_m_size": 0,
"eval_mdu_uspto_markush_ar_markush_equality": 0.5,
"eval_mdu_uspto_markush_ar_r": 0.827,
"eval_mdu_uspto_markush_ar_r_size": 67,
"eval_mdu_uspto_markush_ar_sg": 0.441,
"eval_mdu_uspto_markush_ar_sg_size": 31,
"eval_mdu_uspto_markush_ar_size": 74,
"eval_mdu_uspto_markush_ar_stable_equality": 1.0,
"eval_mdu_uspto_markush_ar_stable_precision": 1.0,
"eval_mdu_uspto_markush_ar_stable_recall": 1.0,
"eval_mdu_uspto_markush_ar_string_equality": 0.149,
"eval_mdu_uspto_markush_ar_string_equality_opt": 0.108,
"eval_mdu_uspto_markush_ar_tanimoto": 0.886,
"eval_mdu_uspto_markush_ar_valid": 0.959,
"eval_mdu_uspto_markush_invalid_gt": 0.0,
"eval_mdu_uspto_molscribe_ar_cxsmi_equality": 0.917,
"eval_mdu_uspto_molscribe_ar_inchi_equality": 0.927,
"eval_mdu_uspto_molscribe_ar_is_in_training": 0.0,
"eval_mdu_uspto_molscribe_ar_m": NaN,
"eval_mdu_uspto_molscribe_ar_m_size": 0,
"eval_mdu_uspto_molscribe_ar_markush_equality": 0.917,
"eval_mdu_uspto_molscribe_ar_r": 0.778,
"eval_mdu_uspto_molscribe_ar_r_size": 9,
"eval_mdu_uspto_molscribe_ar_sg": NaN,
"eval_mdu_uspto_molscribe_ar_sg_size": 0,
"eval_mdu_uspto_molscribe_ar_size": 100,
"eval_mdu_uspto_molscribe_ar_stable_equality": 1.0,
"eval_mdu_uspto_molscribe_ar_stable_precision": 1.0,
"eval_mdu_uspto_molscribe_ar_stable_recall": 1.0,
"eval_mdu_uspto_molscribe_ar_string_equality": 0.896,
"eval_mdu_uspto_molscribe_ar_string_equality_opt": 0.844,
"eval_mdu_uspto_molscribe_ar_tanimoto": 0.961,
"eval_mdu_uspto_molscribe_ar_valid": 0.99,
"eval_mdu_uspto_molscribe_invalid_gt": 0.04,
"eval_mdu_wildmol_m_ar_cxsmi_equality": 0.417,
"eval_mdu_wildmol_m_ar_inchi_equality": 0.594,
"eval_mdu_wildmol_m_ar_is_in_training": 0.0,
"eval_mdu_wildmol_m_ar_m": NaN,
"eval_mdu_wildmol_m_ar_m_size": 0,
"eval_mdu_wildmol_m_ar_markush_equality": 0.417,
"eval_mdu_wildmol_m_ar_r": 0.702,
"eval_mdu_wildmol_m_ar_r_size": 94,
"eval_mdu_wildmol_m_ar_sg": 0.429,
"eval_mdu_wildmol_m_ar_sg_size": 14,
"eval_mdu_wildmol_m_ar_size": 100,
"eval_mdu_wildmol_m_ar_stable_equality": 0.99,
"eval_mdu_wildmol_m_ar_stable_precision": 0.99,
"eval_mdu_wildmol_m_ar_stable_recall": 0.99,
"eval_mdu_wildmol_m_ar_string_equality": 0.24,
"eval_mdu_wildmol_m_ar_string_equality_opt": 0.094,
"eval_mdu_wildmol_m_ar_tanimoto": 0.82,
"eval_mdu_wildmol_m_ar_valid": 0.958,
"eval_mdu_wildmol_m_invalid_gt": 0.04,
"step": 1000
}
],
"logging_steps": 50,
"max_steps": 50000,
"num_train_epochs": 2,
"save_steps": 1000,
"total_flos": 1.952644890624e+16,
"trial_name": null,
"trial_params": null
}