Feature Extraction
Transformers
audio
speech
sparse-autoencoder
sae
interpretability
mechanistic-interpretability
hubert
Instructions to use Egorgij21/Audio-SAE-HuBERT-large with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Egorgij21/Audio-SAE-HuBERT-large with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="Egorgij21/Audio-SAE-HuBERT-large")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Egorgij21/Audio-SAE-HuBERT-large", dtype="auto") - Notebooks
- Google Colab
- Kaggle
File size: 585 Bytes
eb37d01 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 | {
"trainer": {
"trainer_class": "BatchTopKTrainer",
"dict_class": "BatchTopKSAE",
"lr": 0.0002,
"steps": 200001,
"auxk_alpha": 0.0,
"warmup_steps": 10000,
"decay_start": 160000,
"threshold_beta": 0.999,
"threshold_start_step": 1000,
"top_k_aux": 512,
"seed": 21,
"activation_dim": 1024,
"dict_size": 8192,
"k": 50,
"device": "cuda:6",
"layer": 20,
"lm_name": "hubert",
"wandb_name": "BatchTopKSAE",
"submodule_name": null
}
} |