andrew-healey commited on
Commit
c246368
·
verified ·
1 Parent(s): 4b767a9

Upload folder using huggingface_hub

Browse files
attention_kindselective_n_heads2_seed1339/args.json CHANGED
@@ -1 +1 @@
1
- {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_2/attention_kindselective_n_heads2_seed1339", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 2, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 1000, "warmup_steps": 200, "group": "wider_is_better_2", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1339, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "none", "selection_head_linear_combo_scale": 1.0, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 80, "total_batch_size": 20480, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 0.00015, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "1.5e-4_20480_2_1339", "n_embd": 128}
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_2/attention_kindselective_n_heads2_seed1339", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 2, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 1000, "warmup_steps": 200, "group": "wider_is_better_2", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1339, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "none", "selection_head_linear_combo_scale": 1.0, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 120, "total_batch_size": 30720, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 0.00015, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "1.5e-4_30720_2_1339", "n_embd": 128}
attention_kindselective_n_heads2_seed1339/dataloader_00999.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a8513131d2bf732f8f6d97ff82d32c505b2a2f3f9f2247bac315bb09ce0e1f97
3
  size 964
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e95b86432b77c3cf61067c8af4e85b6c7def8ef2dcf5f545b891193eafe80ca3
3
  size 964
attention_kindselective_n_heads2_seed1339/log2.txt CHANGED
@@ -1,53 +1,33 @@
1
  max_steps: 1000
2
- 0 val loss 11.7471
3
- 0 val perplexity 126386.8516
4
- 0 train 11.738260 (lr=7.5000e-07) (hash(x)=107123132)
5
- 50 val loss 10.5008
6
- 50 val perplexity 36345.9922
7
- 100 val loss 9.5989
8
- 100 val perplexity 14748.7441
9
- 100 train 9.610984 (lr=7.5750e-05) (hash(x)=101193890)
10
- 150 val loss 8.2620
11
- 150 val perplexity 3873.8713
12
- 200 val loss 7.7626
13
- 200 val perplexity 2350.9575
14
- 200 train 7.601669 (lr=1.5000e-04) (hash(x)=95079403)
15
- 250 val loss 7.6888
16
- 250 val perplexity 2183.7668
17
- 300 val loss 7.6583
18
- 300 val perplexity 2118.1362
19
- 300 train 7.509525 (lr=1.4486e-04) (hash(x)=96917853)
20
- 350 val loss 7.6596
21
- 350 val perplexity 2120.8447
22
- 400 val loss 7.6557
23
- 400 val perplexity 2112.6025
24
- 400 train 7.508607 (lr=1.3023e-04) (hash(x)=93672417)
25
- 450 val loss 7.6534
26
- 450 val perplexity 2107.7827
27
- 500 val loss 7.6489
28
- 500 val perplexity 2098.4023
29
- 500 train 7.561789 (lr=1.0833e-04) (hash(x)=98036365)
30
- 550 val loss 7.6362
31
- 550 val perplexity 2071.9031
32
- 600 val loss 7.6217
33
- 600 val perplexity 2042.1052
34
- 600 train 7.674348 (lr=8.2500e-05) (hash(x)=100769188)
35
- 650 val loss 7.6124
36
- 650 val perplexity 2023.1761
37
- 700 val loss 7.6120
38
- 700 val perplexity 2022.3361
39
- 700 train 7.566431 (lr=5.6669e-05) (hash(x)=102854282)
40
- 750 val loss 7.6004
41
- 750 val perplexity 1999.0525
42
- 800 val loss 7.5911
43
- 800 val perplexity 1980.5055
44
- 800 train 7.497270 (lr=3.4770e-05) (hash(x)=99815715)
45
- 850 val loss 7.5890
46
- 850 val perplexity 1976.2594
47
- 900 val loss 7.5804
48
- 900 val perplexity 1959.4106
49
- 900 train 7.345047 (lr=2.0138e-05) (hash(x)=96846928)
50
- 950 val loss 7.5746
51
- 950 val perplexity 1948.0265
52
- 999 val loss 7.5653
53
- 999 val perplexity 1929.9879
 
1
  max_steps: 1000
2
+ 0 val loss 11.7467
3
+ 0 val perplexity 126331.5391
4
+ 0 train 11.740535 (lr=7.5000e-07) (hash(x)=150724848)
5
+ 100 val loss 9.5441
6
+ 100 val perplexity 13962.1650
7
+ 100 train 9.588784 (lr=7.5750e-05) (hash(x)=149217061)
8
+ 200 val loss 7.7467
9
+ 200 val perplexity 2313.9080
10
+ 200 train 7.728883 (lr=1.5000e-04) (hash(x)=149910534)
11
+ 300 val loss 7.6659
12
+ 300 val perplexity 2134.2258
13
+ 300 train 7.542049 (lr=1.4486e-04) (hash(x)=142185643)
14
+ 400 val loss 7.6514
15
+ 400 val perplexity 2103.6157
16
+ 400 train 7.689255 (lr=1.3023e-04) (hash(x)=148123706)
17
+ 500 val loss 7.6465
18
+ 500 val perplexity 2093.3313
19
+ 500 train 7.502706 (lr=1.0833e-04) (hash(x)=149952383)
20
+ 600 val loss 7.6388
21
+ 600 val perplexity 2077.1519
22
+ 600 train 7.447033 (lr=8.2500e-05) (hash(x)=146678221)
23
+ 700 val loss 7.5877
24
+ 700 val perplexity 1973.6790
25
+ 700 train 7.608317 (lr=5.6669e-05) (hash(x)=156180736)
26
+ 800 val loss 7.5683
27
+ 800 val perplexity 1935.7953
28
+ 800 train 7.559825 (lr=3.4770e-05) (hash(x)=151700982)
29
+ 900 val loss 7.5518
30
+ 900 val perplexity 1904.2003
31
+ 900 train 7.377645 (lr=2.0138e-05) (hash(x)=147288467)
32
+ 999 val loss 7.5496
33
+ 999 val perplexity 1899.9149
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
attention_kindselective_n_heads2_seed1339/model_00999.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5b9427e25cd8c794fcd6493afec02b0ed8412464d8210605219577d79a8b8c17
3
  size 38587970
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de7e7554d231eb11dab5f290e3b552edec3e9175f4e991f20163430cb3f63a00
3
  size 38587970
attention_kindselective_n_heads2_seed1339/optimizer_00999.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9f956661af1400c1252eb28b994c5ebea1c46edd35433ef521a03303b0bb600d
3
  size 70895430
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b8cb2649579aecd769fcef623ad66440afeb86aa5511e02ec52d02ff68c8f34
3
  size 70895430