asrtbsc_phi-freezed-best
ASR transcripts multi-label DAC
Model description
Backbone: Phi 3 mini
Pooling: Weighted mean
Multi-label classification head: 2 dense layers with two dropouts 0.3 and Tanh activation inbetween
Training and evaluation data
Trained on ASR transcripts.
Evaluated on ground truth (GT) and normalized Whisper small transcripts (E2E).
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0002
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 20
Framework versions
- Transformers 4.41.1
- Pytorch 2.3.0+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1
- Downloads last month
- 7
Dataset used to train Masioki/asrtbsc_phi-freezed-best
Evaluation results
- F1 macro E2E on asapp/slue-phase-2self-reportedTBA
- F1 macro GT on asapp/slue-phase-2self-reportedTBA