| # Lightning-based configuration for LSTM model | |
| # Extends the base LSTM configuration with Lightning-specific settings | |
| MODEL: | |
| TYPE: 'Transformer' | |
| MAXLEN: 272 | |
| VOCAB_SIZE: 77 | |
| ASR_VOCAB_SIZE: 91 | |
| D_MODEL: 256 | |
| NUM_HEADS: 4 | |
| DFF: 128 | |
| NUM_BLOCKS: 2 | |
| DROPOUT_RATE: 0.2 | |
| OUTPUT_SIZE: 19 | |
| USE_ASR: true | |
| PRETRAINED_PATH: /home/rufael/Projects/forced_alignment/Diac/outputs/old/results/transformer-text+asr/tashkeela+clartts/tensorboard/version_0/checkpoints/best_model.ckpt | |
| LOAD_TEXT_BRANCH_ONLY: false | |
| WITH_CONN: False | |
| TRAIN: | |
| DEVICE: 'cuda' | |
| BATCH_SIZE: 128 | |
| NUM_EPOCHS: 1000 | |
| LEARNING_RATE: 0.001 | |
| SAVE_FREQ: 30 | |
| EVAL_FREQ: 1 | |
| SAVE_DIR: 'outputs/kssa/transformer-text+asr/tashkeela+clartts+kssa' | |
| EARLY_STOPPING_PATIENCE: 50 | |
| ACCUMULATE_GRAD_BATCHES: 1 | |
| VAL_CHECK_INTERVAL: 1.0 | |
| GRAD_CLIP_NORM: 0.5 | |
| INFERENCE: | |
| MAX_LENGTH: 270 | |
| WINDOW_SIZE: 50 | |
| BUFFER_SIZE: 25 | |
| DEVICE: 'cpu' | |
| BATCH_SIZE: 16 | |
| MODEL_PATH: /home/rufael/Projects/forced_alignment/Diac/outputs/kssa/transformer-text+asr/tashkeela+clartts+kssa/tensorboard/version_18/checkpoints/best_model.ckpt | |
| ASR_MODEL_NAME: 'sashat/whisper-medium-ClassicalAr' | |
| USE_ASR: True | |
| FORCED_IDS: null | |
| OUTPUT_PATH: 'outputs/kssa/transformer-text+asr/tashkeela+clartts+kssa/outs/dev_nobom.txt' | |
| DATA: | |
| TRAIN_PATH: 'data/kssa/train+asr.tsv' | |
| VAL_PATH: 'data/kssa/dev+asr.tsv' | |
| TEST_PATH: 'data/kssa/dev+asr.tsv' | |
| MAX_LENGTH: 270 | |
| CONSTANTS_PATH: 'constants/' |