End of training
Browse files
README.md
CHANGED
@@ -15,14 +15,14 @@ This student model is distilled from the teacher model [roneneldan/TinyStories-3
|
|
15 |
The [Distily](https://github.com/lapp0/distily) library was used for this distillation.
|
16 |
|
17 |
It achieves the following results on the evaluation set:
|
18 |
-
- eval_enwikippl:
|
19 |
-
- eval_frwikippl:
|
20 |
-
- eval_zhwikippl:
|
21 |
-
- eval_tinystoriesppl:
|
22 |
-
- eval_loss:
|
23 |
-
- eval_runtime:
|
24 |
-
- eval_samples_per_second: 76.
|
25 |
-
- eval_steps_per_second: 9.
|
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.
|
@@ -47,7 +47,7 @@ More information needed
|
|
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=0, loss_fn=None, layer_mapper=None, projector=None))
|
49 |
- train_embeddings: True
|
50 |
-
- learning_rate:
|
51 |
- train_batch_size: 1
|
52 |
- eval_batch_size: 8
|
53 |
- seed: 42
|
@@ -56,33 +56,33 @@ The following hyperparameters were used during training:
|
|
56 |
- num_epochs: 1.0
|
57 |
|
58 |
### Resource Usage
|
59 |
-
Peak GPU Memory: 6.
|
60 |
|
61 |
### Eval-Phase Metrics
|
62 |
| step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl |
|
63 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
|
64 |
| **teacher eval** | | 169.9865 | 47377.9414 | | | | | 3.9789 | 4998.1294 |
|
65 |
-
| 0 | 0 |
|
66 |
-
| 5000 | 0.0505 |
|
67 |
-
| 10000 | 0.1010 |
|
68 |
-
| 15000 | 0.1515 |
|
69 |
-
| 20000 | 0.2020 |
|
70 |
-
| 25000 | 0.2525 |
|
71 |
-
| 30000 | 0.3030 |
|
72 |
-
| 35000 | 0.3535 |
|
73 |
-
| 40000 | 0.4040 |
|
74 |
-
| 45000 | 0.4545 |
|
75 |
-
| 50000 | 0.5051 |
|
76 |
-
| 55000 | 0.5556 |
|
77 |
-
| 60000 | 0.6061 |
|
78 |
-
| 65000 | 0.6566 |
|
79 |
-
| 70000 | 0.7071 |
|
80 |
-
| 75000 | 0.7576 |
|
81 |
-
| 80000 | 0.8081 |
|
82 |
-
| 85000 | 0.8586 |
|
83 |
-
| 90000 | 0.9091 |
|
84 |
-
| 95000 | 0.9596 |
|
85 |
-
| 99000 | 1.0 |
|
86 |
|
87 |
### Framework versions
|
88 |
- Distily 0.2.0
|
|
|
15 |
The [Distily](https://github.com/lapp0/distily) library was used for this distillation.
|
16 |
|
17 |
It achieves the following results on the evaluation set:
|
18 |
+
- eval_enwikippl: 133.3661
|
19 |
+
- eval_frwikippl: 19650.0977
|
20 |
+
- eval_zhwikippl: 54146.3867
|
21 |
+
- eval_tinystoriesppl: 9.1470
|
22 |
+
- eval_loss: 1.2078
|
23 |
+
- eval_runtime: 12.9975
|
24 |
+
- eval_samples_per_second: 76.938
|
25 |
+
- eval_steps_per_second: 9.617
|
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.
|
|
|
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=0, loss_fn=None, layer_mapper=None, projector=None))
|
49 |
- train_embeddings: True
|
50 |
+
- learning_rate: 4e-06
|
51 |
- train_batch_size: 1
|
52 |
- eval_batch_size: 8
|
53 |
- seed: 42
|
|
|
56 |
- num_epochs: 1.0
|
57 |
|
58 |
### Resource Usage
|
59 |
+
Peak GPU Memory: 6.6064 GB
|
60 |
|
61 |
### Eval-Phase Metrics
|
62 |
| step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl |
|
63 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
|
64 |
| **teacher eval** | | 169.9865 | 47377.9414 | | | | | 3.9789 | 4998.1294 |
|
65 |
+
| 0 | 0 | 50480.5703 | 85684.4844 | 6.8305 | 13.0278 | 76.759 | 9.595 | 33932.0586 | 94692.1562 |
|
66 |
+
| 5000 | 0.0505 | 133.2834 | 19661.1758 | 1.2085 | 13.0193 | 76.809 | 9.601 | 9.1500 | 54349.0273 |
|
67 |
+
| 10000 | 0.1010 | 133.3712 | 19627.9590 | 1.2079 | 13.0358 | 76.712 | 9.589 | 9.1530 | 54117.5273 |
|
68 |
+
| 15000 | 0.1515 | 133.2834 | 19650.0977 | 1.2077 | 13.0514 | 76.62 | 9.578 | 9.1402 | 54088.6328 |
|
69 |
+
| 20000 | 0.2020 | 133.3299 | 19639.0254 | 1.2077 | 13.0057 | 76.889 | 9.611 | 9.1545 | 54204.1992 |
|
70 |
+
| 25000 | 0.2525 | 133.4539 | 19650.0977 | 1.2079 | 13.0301 | 76.745 | 9.593 | 9.1538 | 54349.0273 |
|
71 |
+
| 30000 | 0.3030 | 133.5160 | 19650.0977 | 1.2079 | 13.03 | 76.746 | 9.593 | 9.1542 | 54088.6328 |
|
72 |
+
| 35000 | 0.3535 | 133.2834 | 19627.9590 | 1.2078 | 13.0569 | 76.588 | 9.573 | 9.1451 | 54117.5273 |
|
73 |
+
| 40000 | 0.4040 | 133.3712 | 19627.9590 | 1.2078 | 12.9991 | 76.928 | 9.616 | 9.1523 | 54146.3867 |
|
74 |
+
| 45000 | 0.4545 | 133.3041 | 19650.0977 | 1.2077 | 12.9923 | 76.969 | 9.621 | 9.1477 | 54088.6328 |
|
75 |
+
| 50000 | 0.5051 | 133.2834 | 19650.0977 | 1.2078 | 13.1989 | 75.764 | 9.47 | 9.1470 | 54204.1992 |
|
76 |
+
| 55000 | 0.5556 | 133.4953 | 19650.0977 | 1.2078 | 13.1556 | 76.013 | 9.502 | 9.1485 | 54117.5273 |
|
77 |
+
| 60000 | 0.6061 | 133.4901 | 19661.1758 | 1.2077 | 13.206 | 75.723 | 9.465 | 9.1477 | 54117.5273 |
|
78 |
+
| 65000 | 0.6566 | 133.4488 | 19650.0977 | 1.2077 | 13.0052 | 76.892 | 9.612 | 9.1470 | 54117.5273 |
|
79 |
+
| 70000 | 0.7071 | 133.3661 | 19650.0977 | 1.2078 | 12.9996 | 76.925 | 9.616 | 9.1470 | 54117.5273 |
|
80 |
+
| 75000 | 0.7576 | 133.4074 | 19650.0977 | 1.2079 | 13.0082 | 76.874 | 9.609 | 9.1470 | 54117.5273 |
|
81 |
+
| 80000 | 0.8081 | 133.4488 | 19650.0977 | 1.2078 | 12.9816 | 77.032 | 9.629 | 9.1485 | 54117.5273 |
|
82 |
+
| 85000 | 0.8586 | 133.3661 | 19650.0977 | 1.2077 | 12.9875 | 76.997 | 9.625 | 9.1470 | 54117.5273 |
|
83 |
+
| 90000 | 0.9091 | 133.3661 | 19650.0977 | 1.2077 | 12.985 | 77.012 | 9.626 | 9.1462 | 54117.5273 |
|
84 |
+
| 95000 | 0.9596 | 133.4074 | 19650.0977 | 1.2078 | 13.0478 | 76.641 | 9.58 | 9.1470 | 54146.3867 |
|
85 |
+
| 99000 | 1.0 | 133.3661 | 19650.0977 | 1.2078 | 12.9975 | 76.938 | 9.617 | 9.1470 | 54146.3867 |
|
86 |
|
87 |
### Framework versions
|
88 |
- Distily 0.2.0
|
logs/copy_teacher_modules=_(_lm_head___False)_, learning_rate=4e-06/events.out.tfevents.1724012794.5f530b1cf724
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a83ba5dba130affcfe3553bfe73b6e9eeadd76f9c351871e8c5c0a98ee6166bb
|
3 |
+
size 312
|