End of training

final_model/training_config.json

@@ -1,5 +1,5 @@
 {
-  "vocab_size": 4796,
+  "vocab_size": null,
   "max_position_embeddings": 8192,
   "d_model": 512,
   "d_kv": 64,
@@ -9,22 +9,22 @@
   "max_source_length": 2048,
   "max_target_length": 2048,
   "learning_rate": 0.0001,
-  "warmup_steps": 1000,
-  "epochs": 3,
+  "warmup_steps": 100,
+  "epochs": 4,
   "per_device_train_batch_size": 16,
   "gradient_accumulation_steps": 4,
   "fp16": true,
   "gradient_checkpointing": true,
-  "fill_token": "[MASK]",
+  "fill_token": null,
   "content_probability": 0.3,
   "dataset_name": "sandernotenbaert/okai-musiclang-hierarchical",
-  "dataset_config": "hierarchical_all_files_v003_full_content",
+  "dataset_config": "hierarchical_all_files_F-K-content-v001",
   "tokenizer_name": "sandernotenbaert/okai-musiclang-tokenizer-wordlevel",
   "output_dir": "./results/hierarchical_music_t5",
   "logging_steps": 50,
   "save_steps": 500,
   "eval_steps": 500,
-  "resume_from_hub": false,
+  "resume_from_hub": true,
   "push_to_hub": true,
   "hub_model_id": "sandernotenbaert/hierarchical-music-t5",
   "hub_token": null,