Jbloom's workspace
Runs
28
State
Notes
User
Tags
Created
Runtime
Sweep
adam_beta1
adam_beta2
b_dec_init_method
cached_activations_path
checkpoint_path
context_size
d_in
d_sae
dataset_path
dead_feature_estimation_method
dead_feature_threshold
dead_feature_window
device
dtype
expansion_factor
feature_reinit_scale
feature_sampling_window
fine_tune_tokens
finetuning_method
hook_point
hook_point_layer
is_dataset_tokenized
l1_coefficient
log_to_wandb
lr
lr_scheduler_name
lr_warm_up_steps
model_name
mse_loss_normalization
n_batches_in_buffer
n_checkpoints
normalize_activations
resample_batches
run_name
seed
store_batch_size
tokens_per_buffer
total_training_tokens
train_batch_size
use_cached_activations
use_ghost_grads
use_pre_encoder_bias
wandb_log_frequency
wandb_project
Finished
jbloom
1h 25m 11s
-
0
0.9999
mean
activations/apollo-research_sae-Skylion007-openwebtext-tokenizer-gpt2/gpt2-small/blocks.3.hook_resid_pre
checkpoints/ysoc8r4p
128
768
12288
apollo-research/sae-Skylion007-openwebtext-tokenizer-gpt2
-
1.0000e-8
5000
cuda
torch.float32
16
-
1000
100000000
decoder
blocks.3.hook_resid_pre
3
true
0.00008
true
0.0004
constantwithwarmup
10000
gpt2-small
variance
128
10
false
-
12288-L1-8e-05-LR-0.0004-Tokens-2.000e+08
42
32
67108864
200000000
4096
false
-
false
100
mats_sae_training_gpt2_ghost_grad_experiment
Finished
jbloom
2h 10m 22s
-
0
0.9999
mean
activations/apollo-research_sae-Skylion007-openwebtext-tokenizer-gpt2/gpt2-small/blocks.3.hook_resid_pre
checkpoints/penlplwz
128
768
12288
apollo-research/sae-Skylion007-openwebtext-tokenizer-gpt2
-
1.0000e-8
5000
cuda
torch.float32
16
-
1000
100000000
decoder
blocks.3.hook_resid_pre
3
true
0.00008
true
0.0004
constantwithwarmup
10000
gpt2-small
variance
128
10
false
-
12288-L1-8e-05-LR-0.0004-Tokens-2.000e+08
42
32
67108864
200000000
4096
false
-
true
100
mats_sae_training_gpt2_ghost_grad_experiment
Finished
jbloom
1h 25m 29s
-
0.9
0.999
mean
activations/apollo-research_sae-Skylion007-openwebtext-tokenizer-gpt2/gpt2-small/blocks.3.hook_resid_pre
checkpoints/0bli492w
128
768
12288
apollo-research/sae-Skylion007-openwebtext-tokenizer-gpt2
-
1.0000e-8
5000
cuda
torch.float32
16
-
1000
0
-
blocks.3.hook_resid_pre
3
true
0.00008
true
0.0004
constantwithwarmup
10000
gpt2-small
variance
128
10
false
-
12288-L1-8e-05-LR-0.0004-Tokens-3.000e+08
42
32
67108864
300000000
4096
false
-
false
100
mats_sae_training_gpt2_ghost_grad_experiment
Finished
jbloom
2h 23m 37s
-
0
0.9999
mean
activations/apollo-research_sae-Skylion007-openwebtext-tokenizer-gpt2/gpt2-small/blocks.3.hook_resid_pre
checkpoints/ylits63n
128
768
12288
apollo-research/sae-Skylion007-openwebtext-tokenizer-gpt2
-
1.0000e-8
5000
cuda
torch.float32
16
-
1000
0
decoder
blocks.3.hook_resid_pre
3
true
0.00008
true
0.0004
constantwithwarmup
10000
gpt2-small
variance
128
10
false
-
12288-L1-8e-05-LR-0.0004-Tokens-3.000e+08
42
32
67108864
300000000
4096
false
-
true
100
mats_sae_training_gpt2_ghost_grad_experiment
1-4
of 4