AKSARA-CLM-v1 / model /checkpoint.json
emylton's picture
Upload model/checkpoint.json with huggingface_hub
4b9cc3f verified
Raw
History Blame Contribute Delete
2.31 kB
{
"state": {
"phase_index": 4,
"epoch_index": 5,
"global_step": 8440,
"best_metric": 1.857707867548929,
"metrics": {
"loss": 1.857707867548929,
"running_terms": {
"state_consistency": 699.7146280673333,
"constraint_satisfaction": 101.4670789560328,
"semantic_alignment": 129.8230349663645,
"gos_coherence": 1246.1481617949903,
"multi_state_margin": 1679.8049648925662
},
"last_batch": {
"state_consistency": 0.4334685243666172,
"constraint_satisfaction": 0.09194006177131087,
"semantic_alignment": 0.0810413584113121,
"gos_coherence": 0.7358924970030785,
"multi_state_margin": 1.115578532218933
}
},
"config_snapshot": {
"seed": 42,
"device": "cuda",
"curriculum": {
"phase_index": 4,
"objectives": {
"state_consistency": 1.0,
"constraint_satisfaction": 1.0,
"semantic_alignment": 1.0,
"gos_coherence": 0.1,
"multi_state_margin": 1.0
},
"phase_advanced": true
},
"model": {
"vocab_size": 5000,
"affix_vocab_size": 40,
"kbbi_input_dim": 16,
"bsu_config": {
"d_morpheme": 64,
"d_semantic": 64,
"d_role": 32,
"d_context": 64,
"dropout": 0.1
},
"meb_config": {
"n_layers": 4,
"n_dep_heads": 4,
"dropout": 0.1
}
}
}
},
"metadata": {
"loss": 1.857707867548929,
"diagnostics": {
"running_terms": {
"state_consistency": 699.7146280673333,
"constraint_satisfaction": 101.4670789560328,
"semantic_alignment": 129.8230349663645,
"gos_coherence": 1246.1481617949903,
"multi_state_margin": 1679.8049648925662
},
"last_batch": {
"state_consistency": 0.4334685243666172,
"constraint_satisfaction": 0.09194006177131087,
"semantic_alignment": 0.0810413584113121,
"gos_coherence": 0.7358924970030785,
"multi_state_margin": 1.115578532218933
},
"validation": {},
"test": {},
"avg_loss": 1.857707867548929
}
}
}