Danbraunai-apollo's workspace
Runs
299
Name
1 visualized
State
Notes
User
Tags
Created
Runtime
Sweep
ID
act_frequency_n_tokens
batch_size
collect_act_frequency_every_n_samples
collect_output_metrics_every_n_samples
cooldown_samples
effective_batch_size
eval_data.dataset_name
eval_data.is_tokenized
eval_data.n_ctx
eval_data.split
eval_data.streaming
eval_data.tokenizer_name
eval_every_n_samples
eval_n_samples
log_every_n_grad_steps
loss.sparsity.coeff
lr
lr_schedule
max_grad_norm
min_lr_factor
n_samples
saes.sae_positions
save_dir
seed
tlens_model_name
train_data.dataset_name
train_data.is_tokenized
train_data.n_ctx
train_data.split
train_data.streaming
train_data.tokenizer_name
wandb_project
wandb_run_name_prefix
warmup_samples
loss.logits_kl.coeff
loss.out_to_in.coeff
saes.dict_size_to_input_ratio
loss.in_to_orig.hook_positions
loss.in_to_orig.total_coeff
grad_norm
grad_updates
loss
loss/eval/in_to_orig/blocks.0.hook_resid_post
Finished
jordantensor
11h 22m 33s
-
h4nxk49h
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.0001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.90588
25000
1.92778
-
Finished
jordantensor
12h 30m 55s
-
0hir5m2s
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.0001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.5397
25000
0.72537
-
Finished
jordantensor
11h 23m 53s
-
9kmzq3et
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.01
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.13872
25000
2.70401
-
Finished
jordantensor
10h 10m 59s
-
dwp54gfv
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.1
0.01
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.10253
25000
0.073461
0
Finished
jordantensor
12h 25m 32s
-
22p6pwv6
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.1
0.01
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.049602
25000
0.030581
0
Finished
jordantensor
10h 13m 36s
-
gcb2xqiy
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.1
0.005
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.11293
25000
0.0619
0
Finished
jordantensor
12h 28m 45s
-
sswuexuw
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.01
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
0.81796
25000
1.20545
-
Finished
jordantensor
12h 27m 23s
-
31jew03v
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.1
0.005
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.06165
25000
0.024119
0
Finished
jordantensor
10h 14m 19s
-
npm87o4a
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.1
0.001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.13046
25000
0.045094
0
Finished
jordantensor
12h 27m 2s
-
444lj988
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.1
0.001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.044596
25000
0.016912
0
Finished
jordantensor
11h 26m 12s
-
qwwc51we
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.10124
25000
2.35431
-
Finished
jordantensor
12h 30m 45s
-
366n7vs2
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.005
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
0.78385
25000
0.99658
-
Finished
jordantensor
10h 15m 22s
-
czf7lcz9
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.01
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.18566
25000
0.13301
0
Finished
jordantensor
10h 15m 25s
-
vg2rq6sn
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.25945
25000
0.082509
0
Finished
jordantensor
12h 25m 22s
-
hknpq33b
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.005
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.052879
25000
0.047705
0
Finished
jordantensor
10h 15m 29s
-
b8jw9s9s
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.005
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.21238
25000
0.11172
0
Finished
jordantensor
12h 25m 18s
-
8r8qezli
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.01
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.042283
25000
0.059351
0
Finished
jordantensor
12h 25m 22s
-
a7gjz3pl
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.092159
25000
0.036135
0
Finished
jordantensor
11h 22m 48s
-
ivrjmq7q
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.01
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.43878
25000
4.21155
-
Finished
jordantensor
11h 23m 5s
-
ah2xa1tz
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.1647
25000
3.68411
-
Finished
jordantensor
12h 30m 16s
-
309c3bz5
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.005
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
0.87671
25000
1.65894
-
Finished
jordantensor
10h 12m 34s
-
0h8vlzfi
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.3
0.0001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.092906
25000
0.15137
-
Finished
jordantensor
11h 24m 22s
-
bn536vxb
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.7641
25000
5.42775
-
Finished
jordantensor
12h 31m 14s
-
pj32iobw
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.005
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.17706
25000
2.57616
-
Finished
jordantensor
11h 22m 37s
-
9ixnz49n
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.54973
25000
4.7195
-
Finished
jordantensor
12h 28m 1s
-
vmgx6cm5
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.24812
25000
2.05323
-
Finished
jordantensor
11h 22m 53s
-
djxc3qyl
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.21945
25000
3.20496
-
Finished
jordantensor
10h 14m 14s
-
ztjvw7a6
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.3
0.01
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.10745
25000
0.21782
-
Finished
jordantensor
10h 12m 46s
-
qlncs71w
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.01
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.10892
25000
0.27293
-
Finished
jordantensor
12h 30m 2s
-
02hegjg0
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
0.80085
25000
1.3178
-
Finished
jordantensor
11h 22m 47s
-
3tqaciw3
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.0001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
3.9716
25000
4.66987
-
Finished
jordantensor
12h 32m 14s
-
sevge0kz
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.0001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
5.03443
25000
2.09609
-
Finished
jordantensor
11h 23m 59s
-
xf2m0yrx
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.01
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.82696
25000
5.92601
-
Finished
jordantensor
12h 29m 8s
-
0e9ax53g
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.01
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.2195
25000
3.01507
-
Finished
jordantensor
10h 14m 31s
-
qr9k3uqm
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.005
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.083873
25000
0.23756
-
Finished
jordantensor
10h 16m 5s
-
ozovijiw
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.75
0.0001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.27858
25000
0.33642
-
Finished
jordantensor
11h 26m 13s
-
vaapohqe
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.56735
25000
5.38573
-
Finished
jordantensor
12h 31m 29s
-
s48bmzwq
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.005
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.4192
25000
2.59097
-
Finished
jordantensor
12h 29m 4s
-
60spq6t1
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.51758
25000
2.05877
-
Finished
jordantensor
11h 24m 22s
-
5edcpji2
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.71938
25000
4.73934
-
Finished
jordantensor
11h 26m 12s
-
5zsyndj2
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.10124
25000
2.35431
-
Finished
jordantensor
10h 13m 10s
-
42gy528b
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.75
0.01
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.21363
25000
0.43425
-
Finished
jordantensor
11h 24m 37s
-
9hf409l8
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.01
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.43878
25000
4.21155
-
Finished
jordantensor
12h 28m 8s
-
bb2a0xl6
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.01
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.6573
25000
1.9936
-
Finished
jordantensor
10h 15m 34s
-
dkntuyi2
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.3
0.001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.057968
25000
0.15401
-
Finished
jordantensor
10h 12m 41s
-
kqww63dc
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.0001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.12231
25000
0.19606
-
Finished
jordantensor
11h 25m 55s
-
5ho3e3c2
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.1647
25000
3.68411
-
Finished
jordantensor
12h 31m 51s
-
xebvx0i0
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.005
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
0.87671
25000
1.65894
-
Finished
jordantensor
11h 22m 51s
-
wwyctbnd
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.21945
25000
3.20496
-
Finished
jordantensor
10h 15m 59s
-
pza0h35r
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.01
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.10892
25000
0.27293
-
Finished
jordantensor
12h 30m 11s
-
cv86y9mv
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
25
0.001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
0.80085
25000
1.3178
-
Finished
danbraunai-apollo
5h 40m 55s
-
lyh74k4z
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.0005
cosine
10
0.1
200000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
recon_samples-200k_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
2.93513
12500
4.63241
-
Finished
danbraunai-apollo
5h 41m 21s
-
b4ciggvs
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
15
0.0005
cosine
10
0.1
200000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
recon_samples-200k_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.94816
12500
2.3772
-
Finished
jordantensor
10h 13m 5s
-
6wvrefj4
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.005
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.083873
25000
0.23756
-
Finished
jordantensor
10h 14m 12s
-
g17ivqd7
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.064958
25000
0.19657
-
Finished
jordantensor
11h 22m 52s
-
m72pt15d
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.0001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
3.9716
25000
4.66987
-
Finished
jordantensor
12h 32m 3s
-
gsfo26d8
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.0001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
5.03443
25000
2.09609
-
Finished
jordantensor
11h 24m 45s
-
4h324blo
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.01
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.82696
25000
5.92601
-
Finished
jordantensor
12h 30m 52s
-
v3syv6c0
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.01
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.2195
25000
3.01507
-
Finished
jordantensor
10h 10m 51s
-
z2tuguhx
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.75
0.0001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.27723
25000
0.33646
-
Finished
jordantensor
10h 14m 48s
-
fn8rcbgs
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.75
0.01
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.21363
25000
0.43425
-
Finished
jordantensor
10h 11m 11s
-
sp44dxbn
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.75
0.001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.11.hook_resid_pre"]
0.05
0.14718
25000
0.33836
-
Finished
jordantensor
11h 24m 8s
-
hsr7afe4
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.56735
25000
5.38573
-
Finished
jordantensor
12h 29m 25s
-
n6t43y3w
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.005
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.4192
25000
2.59097
-
Finished
jordantensor
12h 30m 7s
-
6l8xijmk
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.51758
25000
2.05877
-
Finished
jordantensor
11h 23m 55s
-
4hozwrvr
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
50
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
lr-comparison_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.71938
25000
4.73934
-
Finished
jordantensor
10h 12m 52s
-
kjtj8jye
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.2
0.01
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.1249
25000
0.093602
0
Finished
jordantensor
12h 24m 46s
-
wukxmuxy
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.2
0.01
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.055744
25000
0.040547
0
Finished
jordantensor
10h 11m 1s
-
zwdq744b
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.2
0.005
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.16003
25000
0.079974
0
Finished
jordantensor
12h 26m
-
rs2sf4si
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.2
0.005
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.042444
25000
0.030919
0
Finished
jordantensor
10h 12m 57s
-
gv3g1b9g
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.2
0.001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.17205
25000
0.059629
0
Finished
jordantensor
12h 24m 9s
-
tiy4gem5
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.2
0.001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.050074
25000
0.02279
0
Finished
jordantensor
10h 11m 13s
-
boekuj5j
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.5
0.01
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.31466
25000
0.20268
0
Finished
jordantensor
10h 13m 21s
-
yp1au0qk
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.5
0.005
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.40657
25000
0.16866
0
Finished
jordantensor
12h 21m 49s
-
66fy31il
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.5
0.005
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.061722
25000
0.085976
0
Finished
jordantensor
12h 22m 31s
-
3zm5evtx
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.5
0.01
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.078541
25000
0.10235
0
Finished
jordantensor
10h 12m 58s
-
oeuiporr
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.5
0.001
cosine
10
0.1
400000
blocks.10.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.5034
25000
0.12394
0
Finished
jordantensor
12h 25m 59s
-
7wff2yrj
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.5
0.001
cosine
10
0.1
400000
blocks.2.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.092813
25000
0.067167
0
Finished
danbraunai-apollo
11h 20m 13s
-
k94kl9jj
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.05
0.01
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.054361
25000
0.043744
0
Finished
danbraunai-apollo
11h 20m 54s
-
jshg5ry8
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.01
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.081328
25000
0.10261
0
Finished
danbraunai-apollo
11h 20m 57s
-
941iskek
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.0001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.33997
25000
0.070378
0
Finished
danbraunai-apollo
11h 18m 41s
-
f0nmv6wj
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
3
0.0001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
1.04607
25000
0.16
0
Finished
danbraunai-apollo
11h 20m 19s
-
rpnfcipf
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.0001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
1.76323
25000
0.26139
0
Finished
danbraunai-apollo
11h 17m 50s
-
ivc6vbek
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.1
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.058435
25000
0.046325
0
Finished
danbraunai-apollo
11h 20m 51s
-
m55q9i2r
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.1
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.068023
25000
0.034806
0
Finished
danbraunai-apollo
11h 17m 43s
-
nxa5a19n
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.082586
25000
0.086985
0
Finished
danbraunai-apollo
11h 16m 44s
-
wxoeoohw
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
0.5
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
lr-comparison_
20000
1
0
60
-
-
0.1163
25000
0.069011
0
Finished
danbraunai-apollo
11h 18m 24s
-
eioczw8q
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.5
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
20000
1
0
60
-
-
0.13382
25000
0.14253
0
Finished
danbraunai-apollo
11h 17m 46s
-
fdhxkazx
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1.5
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
20000
1
0
60
-
-
0.17134
25000
0.11354
0
Finished
danbraunai-apollo
11h 17m 24s
-
3rcesdu0
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
3
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
20000
1
0
60
-
-
0.21851
25000
0.20105
0
Finished
danbraunai-apollo
11h 16m 7s
-
y02b5k9w
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
3
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
20000
1
0
60
-
-
0.29792
25000
0.15903
0
Finished
danbraunai-apollo
11h 14m 48s
-
30wwszc2
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
20000
1
0
60
-
-
0.60507
25000
0.37584
0
Finished
danbraunai-apollo
11h 18m 24s
-
j1r0tfeq
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
20000
1
0
60
-
-
1.05582
25000
0.28975
0
Finished
danbraunai-apollo
11h 34m 36s
-
ecvxsgo0
500000
16
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
65
0.0005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
biggerbatch-recon_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.91146
25000
10.38431
-
Finished
danbraunai-apollo
11h 33m 59s
-
xeeor3t4
500000
4
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
65
0.001
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
recon_
20000
0.5
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
2.5
1.89493
25000
2.67218
-
Finished
danbraunai-apollo
11h 32m 40s
-
8osa6x9y
500000
4
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
10
0.0005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/dan_braun/e2e_sae/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e
recon-all_
20000
1
0
60
["blocks.7.hook_resid_pre","blocks.8.hook_resid_pre","blocks.9.hook_resid_pre","blocks.10.hook_resid_pre","blocks.11.hook_resid_pre"]
0.6
0.38781
25000
0.49525
-
Finished
jordantensor
11h 18m 36s
-
awll0ss8
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
20
0.0005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
20000
1
1
60
-
-
0.27011
25000
1.81691
0
Finished
jordantensor
11h 18m 19s
-
iywthubb
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
8
0.0005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
20000
1
1
60
-
-
0.17012
25000
1.03895
0
Finished
jordantensor
11h 15m 27s
-
ikflfruk
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
1
0.0005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
20000
1
1
60
-
-
0.15594
25000
0.33139
0
Finished
jordantensor
11h 16m 14s
-
ifbv3kky
500000
8
40000
-
0
16
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
40000
500
20
20
0.0005
cosine
10
0.1
400000
blocks.6.hook_resid_pre
/data/jordan_tensor/e2e_sae-main/e2e_sae-main/e2e_sae/scripts/train_tlens_saes/out
0
gpt2-small
apollo-research/Skylion007-openwebtext-tokenizer-gpt2
true
1024
train
true
gpt2
gpt2-e2e_play
20000
1
0.5
60
-
-
0.35697
25000
1.42201
0
1-100
of 299