Tim_hua's workspace
Runs
22
Name
15 visualized
State
Notes
User
Tags
Created
Runtime
Sweep
base_save_dir
cache_dir
crosscoder.dec_init_norm
crosscoder.k
crosscoder.n_latents
crosscoder.use_decoder_bias
crosscoder.use_encoder_bias
cuda_device
data.activations_harvester.cache_mode
data.activations_harvester.harvesting_batch_size
data.activations_harvester.inference_dtype
data.activations_harvester.llms
data.n_tokens_for_norm_estimate
data.token_sequence_loader.hf_dataset_name
data.token_sequence_loader.sequence_length
experiment_name
hookpoints
seed
train.batch_size
train.dead_latents_threshold_n_examples
train.gradient_accumulation_steps_per_batch
train.k_aux
train.lambda_aux
train.log_every_n_steps
train.num_steps
train.optimizer.betas
train.optimizer.learning_rate
train.optimizer.type
train.optimizer.warmdown_pct
train.optimizer.warmup_pct
train.save_every_n_steps
train.topk_style
train.upload_saves_to_wandb
wandb.entity
wandb.mode
wandb.project
data.activations_harvester.max_seq_len_with_padding
data.activations_harvester.remove_bos
data.filter_first_n_tokens
data.token_sequence_loader.use_padding
train.buffer_refill_ratio
train.buffer_size
train.clip_grad_norm
train.final_lambda_s
Failed
tkutsyk
8h 30m 31s
-
.checkpoints
.cache
0.01
-
16384
true
true
2
cache
16
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
2048
l1_lambda_2_padding_buffer_1k_refil_0.25_2025-05-05_11-39-39
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
-
-
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
-
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
512
true
0
true
0.25
1000
-
2
Failed
tkutsyk
8h 33m 25s
-
.checkpoints
.cache
0.01
120
16384
true
true
3
cache
32
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
2048
k120_continuous_buffer_new_padding_buffer_1k_refil_0.25_2025-05-05_11-37-06
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
512
true
0
true
0.25
1000
-
-
Killed
tkutsyk
9h 9m 1s
-
.checkpoints
.cache
0.01
120
16384
true
true
1
cache
32
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
2048
k120_continuous_buffer_new_padding_logic_2025-05-01_21-00-14
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
512
true
0
true
0.25
400
-
-
Killed
tkutsyk
6h 44m 13s
-
.checkpoints
.cache
0.01
120
16384
true
true
1
no_cache
8
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
256
k120_continuous_buffer_seed_1052025_2025-05-01_13-02-56
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
1052025
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
0
-
0.25
400
-
-
Killed
tkutsyk
8h 24m 10s
-
.checkpoints
.cache
0.05
120
16384
true
true
0
no_cache
8
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
256
k120_grad_clip_10_dec_init_5e2_lr_1e5_warmup_5e2_2025-05-01_12-58-24
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.00001
adam
0.05
0.05
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
1
-
0.25
400
10
-
Failed
tkutsyk
19h 17m 43s
-
.checkpoints
.cache
0.01
120
16384
true
true
0
no_cache
8
float32
[{"hf_model_name":null,"base_archicteture_name":null,"name":"google/gemma-2-2b","revision":null}]
100000
HuggingFaceFW/fineweb
256
k120_continuous_buffer_filter_1st_token_2025-04-27_18-43-37
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
1
-
0.25
400
-
-
Killed
tkutsyk
11h 18m 24s
-
.checkpoints
.cache
0.01
120
16384
true
true
2
no_cache
8
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
256
k120_grad_clip_warmup_1e1_2025-04-22_19-50-34
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.05
0.1
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
0.25
400
1
-
Finished
tkutsyk
2d 6h 26m 41s
-
.checkpoints
.cache
0.01
120
16384
true
true
1
no_cache
8
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
256
k120_grad_clip_lr_5e5_warmup_5e2_2025-04-22_19-34-41
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.00005
adam
0.05
0.05
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
0.25
400
1
-
Finished
tkutsyk
2d 6h 15s
-
.checkpoints
.cache
0.03
120
16384
true
true
0
no_cache
8
float32
[{"revision":null,"hf_model_name":null,"base_archicteture_name":null,"name":"google/gemma-2-2b"}]
100000
HuggingFaceFW/fineweb
256
k120_grad_clip_dec_init_3e2_lr_5e5_warmup_5e2_2025-04-22_19-33-57
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.00005
adam
0.05
0.05
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
0.25
400
1
-
Killed
tkutsyk
4h 16m 2s
-
.checkpoints
.cache
0.01
120
16384
true
true
1
no_cache
8
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
256
k120_grad_clip_1_2025-04-22_15-14-22
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
0.25
400
1
-
Crashed
tkutsyk
4h 15m 30s
-
.checkpoints
.cache
1
120
16384
true
true
0
no_cache
8
float32
[{"base_archicteture_name":null,"name":"google/gemma-2-2b","revision":null,"hf_model_name":null}]
100000
HuggingFaceFW/fineweb
256
k120_dec_init_1_2025-04-22_15-12-28
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
0.25
400
-
-
Killed
tkutsyk
1d 5h 23m 6s
-
.checkpoints
.cache
0.01
120
16384
true
true
0
no_cache
8
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
256
k120_continuous_optimized_buffer_2025-04-21_09-36-58
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
0.25
400
-
-
Killed
tkutsyk
9h 3m 49s
-
.checkpoints
.cache
0.01
-
16384
true
true
3
no_cache
8
float32
[{"base_archicteture_name":null,"name":"google/gemma-2-2b","revision":null,"hf_model_name":null}]
100000
HuggingFaceFW/fineweb
256
l1_lambda_2_2025-04-20_22-24-33
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
-
-
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
-
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
0.25
400
-
2
Killed
tkutsyk
12h 19m 24s
-
.checkpoints
.cache
0.01
-
16384
true
true
1
no_cache
8
float32
[{"revision":null,"hf_model_name":null,"base_archicteture_name":null,"name":"google/gemma-2-2b"}]
100000
HuggingFaceFW/fineweb
256
l1_lambda_5_2025-04-20_19-08-35
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
-
-
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
-
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
0.25
400
-
5
Crashed
tim_hua
1d 52m 16s
-
.checkpoints
.cache
0.01
120
16384
true
true
0
no_cache
8
float32
[{"base_archicteture_name":null,"name":"google/gemma-2-2b","revision":null,"hf_model_name":null}]
100000
HuggingFaceFW/fineweb
256
k120_continuous_buffer_2025-04-13_22-37-59
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
-
-
-
-
Crashed
tim_hua
1d 5h 31m 37s
-
.checkpoints
.cache
0.01
80
16384
true
true
0
no_cache
8
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
256
k80_2025-04-13_17-58-47
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
-
-
-
-
Crashed
tim_hua
1d 5h 32m 7s
-
.checkpoints
.cache
0.01
120
16384
true
true
0
no_cache
8
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
256
k120_2025-04-13_17-58-20
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.005
0.005
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
-
-
-
-
Crashed
tim_hua
5m 30s
-
.checkpoints
.cache
0.01
120
16384
true
true
0
no_cache
8
float32
[{"base_archicteture_name":null,"name":"google/gemma-2-2b","revision":null,"hf_model_name":null}]
100000
HuggingFaceFW/fineweb
256
k120_2025-04-13_17-51-08
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.05
0.05
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
-
-
-
-
Crashed
tim_hua
7m 45s
-
.checkpoints
.cache
0.01
80
16384
true
true
0
no_cache
8
float32
[{"hf_model_name":null,"base_archicteture_name":null,"name":"google/gemma-2-2b","revision":null}]
100000
HuggingFaceFW/fineweb
256
k80_2025-04-13_17-48-54
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
300000
[0.9,0.999]
0.0002
adam
0.05
0.05
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
-
-
-
-
Killed
tim_hua
26m 11s
-
.checkpoints
.cache
0.01
80
16384
true
true
0
no_cache
8
float32
[{"name":"google/gemma-2-2b","revision":null,"hf_model_name":null,"base_archicteture_name":null}]
100000
HuggingFaceFW/fineweb
256
with_bigger_buffer_2025-04-13_17-03-07
["blocks.12.hook_resid_post","blocks.13.hook_resid_post","blocks.14.hook_resid_post","blocks.15.hook_resid_post","blocks.16.hook_resid_post","blocks.17.hook_resid_post","blocks.18.hook_resid_post","blocks.19.hook_resid_post","blocks.20.hook_resid_post","blocks.21.hook_resid_post"]
10101
1024
500000
4
512
0.0625
10
30000
[0.9,0.999]
0.0002
adam
0.05
0.05
10000
batch_topk
false
jnainani-university-of-massachusetts-amherst
online
acausal_cc_test2
-
-
-
-
-
-
-
-
1-20
of 22