Hugging Face's logo

lapp0
/
distily_bench_obj_cross_v2.13_gpt2

TensorBoard
Safetensors
Distily
gpt2
Generated from Trainer
8-bit precision
bitsandbytes
Model card Files Metrics Community
lapp0 commited on
Commit
c477d3b
•
1 Parent(s): f460f8e

End of training

Browse files
Files changed (2) hide show
  1. README.md +20 -20
  2. logs/hs_layer_mapper=last, hs_loss_fn=kl, hs_weight=1.0/events.out.tfevents.1724096237.f383272e719b +3 -0
README.md CHANGED
@@ -16,14 +16,14 @@ 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: 2160.0
20
- - eval_frwikippl: 9536.0
21
- - eval_zhwikippl: 98816.0
22
- - eval_tinystoriesppl: 1960.0
23
- - eval_loss: 3.2783
24
- - eval_runtime: 12.9016
25
- - eval_samples_per_second: 46.506
26
- - eval_steps_per_second: 11.626
27
 
28
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
29
  should probably proofread and complete it, then remove this comment.
@@ -46,7 +46,7 @@ More information needed
46
  ### Training hyperparameters
47
 
48
  The following hyperparameters were used during training:
49
- - 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=1.0, loss_fn=kl, layer_mapper=all, projector=None), attn_loss_component=LossComponent(label=attn, weight=0, loss_fn=None, layer_mapper=None, projector=None))
50
  - train_embeddings: True
51
  - learning_rate: 0.0001
52
  - train_batch_size: 8
@@ -64,17 +64,17 @@ Peak GPU Memory: 8.0905 GB
64
  | step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl |
65
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
66
  | **teacher eval** | | 43.75 | 61.75 | | | | | 11.8125 | 19.125 |
67
- | 0 | 0 | 1855425871872.0 | 61297773248512.0 | 26.3692 | 12.8695 | 46.622 | 11.655 | 14495514624.0 | 11338713661440.0 |
68
- | 750 | 0.1010 | 2160.0 | 9536.0 | 3.2783 | 12.9016 | 46.506 | 11.626 | 1960.0 | 98816.0 |
69
- | 1500 | 0.2020 | 740.0 | 4640.0 | 2.2764 | 12.9025 | 46.503 | 11.626 | 612.0 | 9728.0 |
70
- | 2250 | 0.3030 | 450.0 | 2656.0 | 1.9411 | 12.9079 | 46.483 | 11.621 | 344.0 | 564.0 |
71
- | 3000 | 0.4040 | 312.0 | 1544.0 | 1.6865 | 12.9248 | 46.422 | 11.606 | 272.0 | 304.0 |
72
- | 3750 | 0.5051 | 237.0 | 964.0 | 1.4813 | 12.9469 | 46.343 | 11.586 | 199.0 | 252.0 |
73
- | 4500 | 0.6061 | 185.0 | 712.0 | 1.2780 | 13.0095 | 46.12 | 11.53 | 144.0 | 254.0 |
74
- | 5250 | 0.7071 | 143.0 | 520.0 | 1.1058 | 13.0283 | 46.054 | 11.513 | 114.5 | 212.0 |
75
- | 6000 | 0.8081 | 130.0 | 478.0 | 1.0297 | 12.9977 | 46.162 | 11.541 | 103.0 | 180.0 |
76
- | 6750 | 0.9091 | 123.5 | 442.0 | 0.9899 | 12.9475 | 46.341 | 11.585 | 99.0 | 160.0 |
77
- | 7425 | 1.0 | 122.5 | 438.0 | 0.9819 | 12.9461 | 46.346 | 11.587 | 97.5 | 160.0 |
78
 
79
  ### Framework versions
80
  - 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: 2176.0
20
+ - eval_frwikippl: 8832.0
21
+ - eval_zhwikippl: 127488.0
22
+ - eval_tinystoriesppl: 1776.0
23
+ - eval_loss: 3.2370
24
+ - eval_runtime: 12.9467
25
+ - eval_samples_per_second: 46.344
26
+ - eval_steps_per_second: 11.586
27
 
28
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
29
  should probably proofread and complete it, then remove this comment.
 
46
  ### Training hyperparameters
47
 
48
  The following hyperparameters were used during training:
49
+ - 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=1.0, loss_fn=kl, layer_mapper=last, projector=None), attn_loss_component=LossComponent(label=attn, weight=0, loss_fn=None, layer_mapper=None, projector=None))
50
  - train_embeddings: True
51
  - learning_rate: 0.0001
52
  - train_batch_size: 8
 
64
  | step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl |
65
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
66
  | **teacher eval** | | 43.75 | 61.75 | | | | | 11.8125 | 19.125 |
67
+ | 0 | 0 | 1821066133504.0 | 158329674399744.0 | 25.4650 | 12.9198 | 46.44 | 11.61 | 12079595520.0 | 98956046499840.0 |
68
+ | 750 | 0.1010 | 2176.0 | 8832.0 | 3.2370 | 12.9467 | 46.344 | 11.586 | 1776.0 | 127488.0 |
69
+ | 1500 | 0.2020 | 780.0 | 4704.0 | 2.2858 | 12.9531 | 46.321 | 11.58 | 580.0 | 6528.0 |
70
+ | 2250 | 0.3030 | 448.0 | 2720.0 | 1.9337 | 12.9786 | 46.23 | 11.558 | 358.0 | 616.0 |
71
+ | 3000 | 0.4040 | 318.0 | 1424.0 | 1.6665 | 12.9898 | 46.19 | 11.548 | 252.0 | 264.0 |
72
+ | 3750 | 0.5051 | 252.0 | 968.0 | 1.4830 | 12.9776 | 46.233 | 11.558 | 206.0 | 494.0 |
73
+ | 4500 | 0.6061 | 187.0 | 680.0 | 1.2771 | 12.9626 | 46.287 | 11.572 | 146.0 | 404.0 |
74
+ | 5250 | 0.7071 | 146.0 | 556.0 | 1.1009 | 12.9778 | 46.233 | 11.558 | 113.0 | 224.0 |
75
+ | 6000 | 0.8081 | 134.0 | 490.0 | 1.0233 | 12.9863 | 46.202 | 11.551 | 104.0 | 179.0 |
76
+ | 6750 | 0.9091 | 125.0 | 464.0 | 0.9838 | 12.985 | 46.207 | 11.552 | 96.0 | 168.0 |
77
+ | 7425 | 1.0 | 124.0 | 462.0 | 0.9755 | 13.0256 | 46.063 | 11.516 | 95.0 | 162.0 |
78
 
79
  ### Framework versions
80
  - Distily 0.2.0
logs/hs_layer_mapper=last, hs_loss_fn=kl, hs_weight=1.0/events.out.tfevents.1724096237.f383272e719b ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1cd16abfbd27535407ccc47dfec09bbd1f35765c2d926745ea121b65037b8d88
3
+ size 578