lapp0 commited on
Commit
f735c9d
1 Parent(s): e267b10

End of training

Browse files
README.md CHANGED
@@ -16,13 +16,13 @@ This student model is distilled from the teacher model [gpt2](https://huggingfac
16
  The [Distily](https://github.com/lapp0/distily) library was used for this distillation.
17
 
18
  It achieves the following results on the evaluation set:
19
- - eval_enwikippl: 207.9922
20
- - eval_frwikippl: 1314.4666
21
- - eval_zhwikippl: 759.8159
22
- - eval_loss: 1.3326
23
- - eval_runtime: 17.3702
24
- - eval_samples_per_second: 57.57
25
- - eval_steps_per_second: 7.196
26
 
27
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
28
  should probably proofread and complete it, then remove this comment.
@@ -45,7 +45,7 @@ More information needed
45
  ### Training hyperparameters
46
 
47
  The following hyperparameters were used during training:
48
- - distillation_objective: DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl, layer_mapper=None, projector=None), hs_loss_component=LossComponent(label=hs, weight=0, loss_fn=None, layer_mapper=None, projector=None), attn_loss_component=LossComponent(label=attn, weight=2.0, loss_fn=reverse_kl, layer_mapper=None, projector=None))
49
  - train_embeddings: True
50
  - learning_rate: 4e-05
51
  - train_batch_size: 8
@@ -62,20 +62,20 @@ Peak GPU Memory: 8.2195 GB
62
  | step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | zhwikippl |
63
  | --- | --- | --- | --- | --- | --- | --- | --- | --- |
64
  | **teacher eval** | | 30.2086 | 57.2728 | | | | | 18.1784 |
65
- | 0 | 0 | 55429.6875 | 57698.8047 | 6.1518 | 17.3225 | 57.728 | 7.216 | 56988.9141 |
66
- | 1000 | 0.0808 | 693.0135 | 4581.8110 | 2.0460 | 17.3292 | 57.706 | 7.213 | 22366.3984 |
67
- | 2000 | 0.1616 | 504.2434 | 3241.0867 | 1.8627 | 17.42 | 57.405 | 7.176 | 1925.1605 |
68
- | 3000 | 0.2424 | 416.2050 | 2635.7954 | 1.7568 | 17.2717 | 57.898 | 7.237 | 924.6143 |
69
- | 4000 | 0.3232 | 367.7481 | 2426.7476 | 1.6637 | 17.2866 | 57.848 | 7.231 | 843.0013 |
70
- | 5000 | 0.4040 | 314.2136 | 2124.5867 | 1.5737 | 17.3864 | 57.516 | 7.19 | 970.9272 |
71
- | 6000 | 0.4848 | 274.5013 | 1727.5643 | 1.5012 | 17.3269 | 57.714 | 7.214 | 815.5406 |
72
- | 7000 | 0.5657 | 250.4276 | 1508.2014 | 1.4380 | 17.3171 | 57.747 | 7.218 | 763.2737 |
73
- | 8000 | 0.6465 | 227.7920 | 1387.4103 | 1.3836 | 17.3674 | 57.579 | 7.197 | 706.1053 |
74
- | 9000 | 0.7273 | 207.9922 | 1314.4666 | 1.3326 | 17.3702 | 57.57 | 7.196 | 759.8159 |
75
- | 10000 | 0.8081 | 190.8745 | 1171.5941 | 1.2857 | 17.3634 | 57.592 | 7.199 | 598.8307 |
76
- | 11000 | 0.8889 | 175.8197 | 1119.1125 | 1.2359 | 17.3533 | 57.626 | 7.203 | 493.6122 |
77
- | 12000 | 0.9697 | 159.0854 | 1000.5724 | 1.1915 | 17.3916 | 57.499 | 7.187 | 562.3265 |
78
- | 12375 | 1.0 | 157.0114 | 957.9113 | 1.1794 | 17.3573 | 57.613 | 7.202 | 671.0787 |
79
 
80
  ### Framework versions
81
  - Distily 0.2.0
 
16
  The [Distily](https://github.com/lapp0/distily) library was used for this distillation.
17
 
18
  It achieves the following results on the evaluation set:
19
+ - eval_enwikippl: 208.9635
20
+ - eval_frwikippl: 1351.4938
21
+ - eval_zhwikippl: 781.2166
22
+ - eval_loss: 19.7940
23
+ - eval_runtime: 17.3332
24
+ - eval_samples_per_second: 57.693
25
+ - eval_steps_per_second: 7.212
26
 
27
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
28
  should probably proofread and complete it, then remove this comment.
 
45
  ### Training hyperparameters
46
 
47
  The following hyperparameters were used during training:
48
+ - distillation_objective: DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl, layer_mapper=None, projector=None), hs_loss_component=LossComponent(label=hs, weight=0, loss_fn=None, layer_mapper=None, projector=None), attn_loss_component=LossComponent(label=attn, weight=2.0, loss_fn=ce, layer_mapper=None, projector=None))
49
  - train_embeddings: True
50
  - learning_rate: 4e-05
51
  - train_batch_size: 8
 
62
  | step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | zhwikippl |
63
  | --- | --- | --- | --- | --- | --- | --- | --- | --- |
64
  | **teacher eval** | | 30.2086 | 57.2728 | | | | | 18.1784 |
65
+ | 0 | 0 | 55429.6875 | 57698.8047 | 24.5150 | 17.4179 | 57.412 | 7.177 | 56988.9141 |
66
+ | 1000 | 0.0808 | 702.9320 | 4403.8062 | 20.5050 | 17.3512 | 57.633 | 7.204 | 19095.4688 |
67
+ | 2000 | 0.1616 | 507.8192 | 3252.5339 | 20.3170 | 17.3451 | 57.653 | 7.207 | 2454.8054 |
68
+ | 3000 | 0.2424 | 418.1162 | 2743.4949 | 20.2070 | 17.3188 | 57.741 | 7.218 | 1193.5658 |
69
+ | 4000 | 0.3232 | 372.6640 | 2567.2002 | 20.1200 | 17.2361 | 58.018 | 7.252 | 1026.6641 |
70
+ | 5000 | 0.4040 | 320.0249 | 2154.7588 | 20.0340 | 17.3151 | 57.753 | 7.219 | 1183.4081 |
71
+ | 6000 | 0.4848 | 278.3867 | 1778.2332 | 19.9610 | 17.3435 | 57.658 | 7.207 | 869.0625 |
72
+ | 7000 | 0.5657 | 251.7534 | 1568.9419 | 19.9040 | 17.4023 | 57.464 | 7.183 | 807.5215 |
73
+ | 8000 | 0.6465 | 230.5502 | 1399.7903 | 19.8380 | 17.3855 | 57.519 | 7.19 | 816.4125 |
74
+ | 9000 | 0.7273 | 208.9635 | 1351.4938 | 19.7940 | 17.3332 | 57.693 | 7.212 | 781.2166 |
75
+ | 10000 | 0.8081 | 192.8560 | 1211.9225 | 19.7530 | 17.3032 | 57.793 | 7.224 | 608.5041 |
76
+ | 11000 | 0.8889 | 179.3916 | 1140.7820 | 19.6930 | 17.2721 | 57.897 | 7.237 | 624.0573 |
77
+ | 12000 | 0.9697 | 161.3999 | 997.4732 | 19.6480 | 17.21 | 58.106 | 7.263 | 560.2280 |
78
+ | 12375 | 1.0 | 158.3214 | 948.9705 | 19.6380 | 17.3071 | 57.78 | 7.222 | 575.3149 |
79
 
80
  ### Framework versions
81
  - Distily 0.2.0
logs/attn_loss_fn=ce, attn_weight=2.0/events.out.tfevents.1723680689.93d6cbb3ad53 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e1a844ba82d26f8d754402c5f272177311e462e03ec1cfb8aa86fe69a0533e74
3
+ size 249