Skip to content
Sparse Autoencoder
train
Initializing search
ai-safety-foundation/sparse_autoencoder
Sparse Autoencoder
ai-safety-foundation/sparse_autoencoder
Home
Demo
Source dataset pre-processing
Reference
Reference
activation_resampler
activation_resampler
activation_resampler
utils
utils
component_slice_tensor
activation_store
activation_store
base_store
tensor_store
autoencoder
autoencoder
components
components
linear_encoder
tied_bias
unit_norm_decoder
lightning
model
types
metrics
metrics
loss
loss
l1_absolute_loss
l2_reconstruction_loss
sae_loss
train
train
capacity
feature_density
l0_norm
neuron_activity
neuron_fired_count
validate
validate
reconstruction_score
wrappers
wrappers
classwise
optimizer
optimizer
adam_with_reset
source_data
source_data
abstract_dataset
mock_dataset
pretokenized_dataset
text_dataset
source_model
source_model
replace_activations_hook
reshape_activations
store_activations_hook
zero_ablate_hook
tensor_types
train
train
join_sweep
pipeline
sweep
sweep_config
utils
utils
get_model_device
round_down
wandb_sweep_types
training_runs
training_runs
gpt2
utils
utils
data_parallel
tensor_shape
Contributing
Citation
Table of contents
train
Train step metrics
ยค
Train step metrics.