BRlkl's picture
Upload folder using huggingface_hub
2ef210e verified
Raw
History Blame Contribute Delete
4.42 kB
{
"phase_results": [
{
"phase": "simple_multiturn_transcript",
"best_selection_metric_value": 5.384137900958771,
"best_metrics": {
"validation/loss_total": 1.408942354669146,
"validation/loss_response": 2.149438410067044,
"validation/loss_current_user_reconstruction": 1.2758943989573013,
"validation/loss_probe": 1.2539331475071445,
"validation/loss_past_response_contrastive": 0.05162567709699089,
"validation/response_first_token_exact_match": 0.21810699588477367,
"validation/current-user_reconstruction_first_token_exact_match": 0.6152263374485597,
"validation/probe_first_token_exact_match": 1.0,
"validation/supervised_turn_count": 486.0,
"validation/response_exact_match": 0.0,
"validation/response_similarity": 0.10392940422656348,
"validation/response_token_f1": 0.20785880845312696,
"validation/response_line_recall": 0.0,
"validation/response_reconstruction_similarity": 0.4982011305164052,
"validation/response_reconstruction_exact_match": 0.025806451612903226,
"validation/response_reconstruction_token_f1": 0.4982011305164052,
"validation/probe_exact_match": 0.0,
"validation/probe_transcript_similarity": 0.2633016615969853,
"validation/probe_token_f1": 0.5266033231939706,
"validation/probe_line_recall": 0.0,
"validation/response_similarity_by_turn/turn_2": 0.11559769549843006,
"validation/response_reconstruction_similarity_by_turn/turn_2": 0.6686679695190779,
"validation/probe_transcript_similarity_by_turn/turn_2": 0.29072483292965323,
"validation/response_similarity_by_turn/turn_3": 0.10607868284934467,
"validation/response_reconstruction_similarity_by_turn/turn_3": 0.5215039721757743,
"validation/probe_transcript_similarity_by_turn/turn_3": 0.26205390129767436,
"validation/response_similarity_by_turn/turn_4": 0.09581244011492175,
"validation/response_reconstruction_similarity_by_turn/turn_4": 0.3903211526872822,
"validation/probe_transcript_similarity_by_turn/turn_4": 0.24065981727738503,
"validation/response_similarity_by_turn/turn_5": 0.08826680527651742,
"validation/response_reconstruction_similarity_by_turn/turn_5": 0.2735804094022488,
"validation/probe_transcript_similarity_by_turn/turn_5": 0.24492614677816982,
"validation/response_similarity_by_turn/turn_6": 0.09125961673419455,
"validation/response_reconstruction_similarity_by_turn/turn_6": 0.3105487240823803,
"validation/probe_transcript_similarity_by_turn/turn_6": 0.2404807978040937,
"validation/goal_loss": 5.384137900958771
},
"global_step": 805,
"train_dataset": {
"example_count": 7696,
"pair_count_mean": 4.2548076923076925,
"pair_count_max": 6,
"response_target_tokens_mean_turn2_plus": 126.23921517671518,
"response_target_tokens_max_turn2_plus": 1643,
"probe_target_tokens_mean_turn2_plus": 471.31613825363826,
"probe_target_tokens_max_turn2_plus": 3285,
"dataset_counts": {
"chatalpaca_multiturn_enriched": 7696
},
"probe_question_text": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assistant 1]: Y [User 2]: A etc",
"current_user_reconstruction_question_text": "What did me (the user) just ask you? give verbatim message I just previously sent"
},
"validation_dataset": {
"example_count": 151,
"pair_count_mean": 4.139072847682119,
"pair_count_max": 6,
"response_target_tokens_mean_turn2_plus": 123.21854304635761,
"response_target_tokens_max_turn2_plus": 889,
"probe_target_tokens_mean_turn2_plus": 448.5298013245033,
"probe_target_tokens_max_turn2_plus": 1545,
"dataset_counts": {
"chatalpaca_multiturn_enriched": 151
},
"probe_question_text": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assistant 1]: Y [User 2]: A etc",
"current_user_reconstruction_question_text": "What did me (the user) just ask you? give verbatim message I just previously sent"
},
"wandb_enabled": true
}
]
}