INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

hips

-0.77

waters

-0.75

 lows

-0.72

 relativity

-0.65

IAS

-0.65

uation

-0.63

iating

-0.62

armor

-0.62

 recovery

-0.61

urgy

-0.61

POSITIVE LOGITS

ĸļ

0.82

 theoretically

0.73

herical

0.73

>]

0.69

ÃĥÃĤ

0.69

ython

0.68

 hypot

0.67

roman

0.66

Pict

0.63

}:

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.