interp_score.py File Reference

Go to the source code of this file.

Namespaces
namespace	aquin
namespace	aquin.compute
namespace	aquin.compute.interp_score

Functions
dict	aquin.compute.interp_score._generate_sentences (str label, client, int n=N_SAMPLES)
float	aquin.compute.interp_score._get_feature_activation (str sentence, int feature_idx, HookedTransformer model, SparseAutoencoder sae, str model_id="llama-3.2-1b", int|None layer=None)
float	aquin.compute.interp_score._cohen_d_score (list[float] pos_acts, list[float] neg_acts)
float|None	aquin.compute.interp_score._feature_purity_score (list[str] sentences, client)
float|None	aquin.compute.interp_score._finite (float|None x, float|None default=None)
float	aquin.compute.interp_score._kl_div (torch.Tensor p, torch.Tensor q)
float	aquin.compute.interp_score._entropy (torch.Tensor p)
dict	aquin.compute.interp_score.run_mui_score (int feature_idx, str prompt, HookedTransformer model, SparseAutoencoder sae, int n_positions=8, str model_id="llama-3.2-1b", int|None layer=None)
dict	aquin.compute.interp_score.run_interp_score (int feature_idx, str prompt, HookedTransformer model, SparseAutoencoder sae, client, int n_samples=N_SAMPLES, str model_id="llama-3.2-1b", int|None layer=None)

Variables
	aquin.compute.interp_score.DEVICE = resolve_compute_device()
int	aquin.compute.interp_score.N_SAMPLES = 10