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: 75.
|
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=10.0, loss_fn=raw_mse, layer_mapper=None, projector=None), attn_loss_component=LossComponent(label=attn, weight=10.0, loss_fn=raw_mse, layer_mapper=None, projector=None))
|
49 |
- train_embeddings: True
|
50 |
-
- learning_rate:
|
51 |
- train_batch_size: 16
|
52 |
- eval_batch_size: 8
|
53 |
- seed: 42
|
@@ -62,18 +62,18 @@ Peak GPU Memory: 16.2498 GB
|
|
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 |
-
| 2000 | 0.1293 |
|
67 |
-
| 4000 | 0.2586 |
|
68 |
-
| 6000 | 0.3879 |
|
69 |
-
| 8000 | 0.5172 |
|
70 |
-
| 10000 | 0.6465 |
|
71 |
-
| 12000 | 0.7757 |
|
72 |
-
| 14000 | 0.9050 |
|
73 |
-
| 15469 | 1.0 |
|
74 |
|
75 |
### Framework versions
|
76 |
- Distily 0.2.0
|
77 |
- Transformers 4.44.0
|
78 |
- Pytorch 2.3.0
|
79 |
-
- Datasets 2.
|
|
|
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: 12623.7412
|
19 |
+
- eval_frwikippl: 52327.9961
|
20 |
+
- eval_zhwikippl: 96451.4609
|
21 |
+
- eval_tinystoriesppl: 6281.7222
|
22 |
+
- eval_loss: 18.2212
|
23 |
+
- eval_runtime: 32.9567
|
24 |
+
- eval_samples_per_second: 75.857
|
25 |
+
- eval_steps_per_second: 9.497
|
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=10.0, loss_fn=raw_mse, layer_mapper=None, projector=None), attn_loss_component=LossComponent(label=attn, weight=10.0, loss_fn=raw_mse, layer_mapper=None, projector=None))
|
49 |
- train_embeddings: True
|
50 |
+
- learning_rate: 4e-06
|
51 |
- train_batch_size: 16
|
52 |
- eval_batch_size: 8
|
53 |
- seed: 42
|
|
|
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 | 12930.5508 | 52883.7109 | 18.2508 | 32.8136 | 76.188 | 9.539 | 6471.4722 | 96760.75 |
|
66 |
+
| 2000 | 0.1293 | 12623.7412 | 52327.9961 | 18.2212 | 32.8936 | 76.003 | 9.516 | 6281.7222 | 96451.4609 |
|
67 |
+
| 4000 | 0.2586 | 12623.7412 | 52327.9961 | 18.2212 | 32.8034 | 76.212 | 9.542 | 6281.7222 | 96451.4609 |
|
68 |
+
| 6000 | 0.3879 | 12623.7412 | 52327.9961 | 18.2212 | 32.8868 | 76.018 | 9.517 | 6281.7222 | 96451.4609 |
|
69 |
+
| 8000 | 0.5172 | 12623.7412 | 52327.9961 | 18.2212 | 32.9567 | 75.857 | 9.497 | 6281.7222 | 96451.4609 |
|
70 |
+
| 10000 | 0.6465 | 12623.7412 | 52327.9961 | 18.2212 | 32.9256 | 75.929 | 9.506 | 6281.7222 | 96451.4609 |
|
71 |
+
| 12000 | 0.7757 | 12623.7412 | 52327.9961 | 18.2212 | 32.9173 | 75.948 | 9.509 | 6281.7222 | 96451.4609 |
|
72 |
+
| 14000 | 0.9050 | 12623.7412 | 52327.9961 | 18.2212 | 33.1691 | 75.371 | 9.437 | 6281.7222 | 96451.4609 |
|
73 |
+
| 15469 | 1.0 | 12623.7412 | 52327.9961 | 18.2212 | 33.0829 | 75.568 | 9.461 | 6281.7222 | 96451.4609 |
|
74 |
|
75 |
### Framework versions
|
76 |
- Distily 0.2.0
|
77 |
- Transformers 4.44.0
|
78 |
- Pytorch 2.3.0
|
79 |
+
- Datasets 2.20.0
|
logs/attn_loss_fn=raw_mse, attn_weight=10.0, hs_loss_fn=raw_mse, hs_weight=10.0, learning_rate=4e-06/events.out.tfevents.1723761898.93d6cbb3ad53
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2e29c5c31674ab017aaa74d55b1626f675e08ae2fb3ba79dedcefb5c425a6e77
|
3 |
+
size 307
|
model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 137033984
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6a64218e3d2c30d196c07cc994eaf051554a03e0f07ebd8cd8baad5a362b72d2
|
3 |
size 137033984
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1017948104
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8d665b5b0ccf6b5a8d166fee5cb77c0bd9dc46300ff235f1d51f2fbe7828aef5
|
3 |
size 1017948104
|