End of training
Browse files
README.md
CHANGED
@@ -44,36 +44,36 @@ More information needed
|
|
44 |
| step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl |
|
45 |
| :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
|
46 |
| **teacher eval** | | 43.25 | 61.25 | | | | | 11.6875 | 19.125 |
|
47 |
-
| 0 | 0 |
|
48 |
-
| 2500 | 0.0404 |
|
49 |
-
| 5000 | 0.0808 |
|
50 |
-
| 7500 | 0.1212 |
|
51 |
-
| 10000 | 0.1616 |
|
52 |
-
| 12500 | 0.2020 |
|
53 |
-
| 15000 | 0.2424 |
|
54 |
-
| 17500 | 0.2828 |
|
55 |
-
| 20000 | 0.3232 |
|
56 |
-
| 22500 | 0.3636 |
|
57 |
-
| 25000 | 0.4040 |
|
58 |
-
| 27500 | 0.4444 |
|
59 |
-
| 30000 | 0.4848 |
|
60 |
-
| 32500 | 0.5253 |
|
61 |
-
| 35000 | 0.5657 |
|
62 |
-
| 37500 | 0.6061 |
|
63 |
-
| 40000 | 0.6465 |
|
64 |
-
| 42500 | 0.6869 |
|
65 |
-
| 45000 | 0.7273 |
|
66 |
-
| 47500 | 0.7677 | 50.
|
67 |
-
| 50000 | 0.8081 |
|
68 |
-
| 52500 | 0.8485 |
|
69 |
-
| 55000 | 0.8889 |
|
70 |
-
| 57500 | 0.9293 |
|
71 |
-
| 60000 | 0.9697 |
|
72 |
-
| 61875 | 1.0 |
|
73 |
|
74 |
# Resource Usage Comparison
|
75 |
|
76 |
-
- VRAM Use: 7.
|
77 |
|
78 |
# Distillation (Teacher -> Student) Architecture Difference:
|
79 |
|
@@ -93,7 +93,7 @@ More information needed
|
|
93 |
<br/>
|
94 |
|
95 |
# Train Dataset
|
96 |
-
Trained on 145,
|
97 |
|
98 |
- Num Samples: `247,500`
|
99 |
- Subset: `20231101.en`
|
@@ -103,7 +103,7 @@ Trained on 145,731,638 tokens from the [wikimedia/wikipedia](https://huggingface
|
|
103 |
# Training Objective
|
104 |
|
105 |
```
|
106 |
-
DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=10.0, loss_fn=
|
107 |
```
|
108 |
|
109 |
# Hyperparameters
|
@@ -120,9 +120,9 @@ The following hyperparameters were used during training:
|
|
120 |
- lr_scheduler_type: `linear`
|
121 |
- lr_scheduler_warmup_ratio: `0.5`
|
122 |
- num_epochs: `1.0`
|
123 |
-
- distillation_objective: `DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=10.0, loss_fn=
|
124 |
- train_embeddings: `True`
|
125 |
-
- lr_scheduler: `<torch.optim.lr_scheduler.LambdaLR object at
|
126 |
- student_model_name_or_path: `None`
|
127 |
- student_config_name_or_path: `None`
|
128 |
- student_model_config: `None`
|
|
|
44 |
| step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl |
|
45 |
| :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
|
46 |
| **teacher eval** | | 43.25 | 61.25 | | | | | 11.6875 | 19.125 |
|
47 |
+
| 0 | 0 | 2473901162496.0 | 170424302305280.0 | 30.7740 | 25.2349 | 99.069 | 12.403 | 4060086272.0 | 71468255805440.0 |
|
48 |
+
| 2500 | 0.0404 | 1184.0 | 11776.0 | 9.8284 | 25.2487 | 99.015 | 12.397 | 784.0 | 12800.0 |
|
49 |
+
| 5000 | 0.0808 | 412.0 | 2272.0 | 8.3993 | 25.2618 | 98.964 | 12.39 | 290.0 | 434.0 |
|
50 |
+
| 7500 | 0.1212 | 245.0 | 916.0 | 7.6586 | 25.2887 | 98.858 | 12.377 | 218.0 | 195.0 |
|
51 |
+
| 10000 | 0.1616 | 182.0 | 676.0 | 7.2415 | 25.2556 | 98.988 | 12.393 | 164.0 | 190.0 |
|
52 |
+
| 12500 | 0.2020 | 131.0 | 504.0 | 6.6883 | 25.2962 | 98.829 | 12.373 | 115.5 | 158.0 |
|
53 |
+
| 15000 | 0.2424 | 112.5 | 432.0 | 6.4127 | 25.2743 | 98.915 | 12.384 | 89.5 | 144.0 |
|
54 |
+
| 17500 | 0.2828 | 93.5 | 344.0 | 6.1979 | 25.214 | 99.151 | 12.414 | 70.5 | 127.0 |
|
55 |
+
| 20000 | 0.3232 | 75.0 | 270.0 | 5.9310 | 25.2265 | 99.102 | 12.408 | 63.75 | 128.0 |
|
56 |
+
| 22500 | 0.3636 | 67.0 | 209.0 | 5.6634 | 25.2495 | 99.012 | 12.396 | 49.75 | 83.5 |
|
57 |
+
| 25000 | 0.4040 | 63.5 | 192.0 | 5.5561 | 25.2476 | 99.019 | 12.397 | 44.25 | 86.0 |
|
58 |
+
| 27500 | 0.4444 | 58.0 | 192.0 | 5.4855 | 25.2834 | 98.879 | 12.38 | 40.25 | 70.5 |
|
59 |
+
| 30000 | 0.4848 | 58.75 | 195.0 | 5.4646 | 25.2547 | 98.992 | 12.394 | 41.75 | 65.0 |
|
60 |
+
| 32500 | 0.5253 | 58.5 | 171.0 | 5.4511 | 25.2 | 99.206 | 12.421 | 40.0 | 60.0 |
|
61 |
+
| 35000 | 0.5657 | 57.0 | 165.0 | 5.3711 | 25.2873 | 98.864 | 12.378 | 36.75 | 49.25 |
|
62 |
+
| 37500 | 0.6061 | 57.75 | 155.0 | 5.3390 | 25.2952 | 98.833 | 12.374 | 37.75 | 54.0 |
|
63 |
+
| 40000 | 0.6465 | 55.75 | 154.0 | 5.3225 | 25.2919 | 98.846 | 12.376 | 34.5 | 57.0 |
|
64 |
+
| 42500 | 0.6869 | 54.75 | 146.0 | 5.2939 | 25.2713 | 98.926 | 12.386 | 35.5 | 49.0 |
|
65 |
+
| 45000 | 0.7273 | 50.75 | 133.0 | 5.1563 | 25.2812 | 98.888 | 12.381 | 30.0 | 48.0 |
|
66 |
+
| 47500 | 0.7677 | 50.75 | 124.5 | 5.1271 | 25.3128 | 98.764 | 12.365 | 29.375 | 35.25 |
|
67 |
+
| 50000 | 0.8081 | 49.75 | 123.0 | 5.1093 | 25.2369 | 99.061 | 12.402 | 28.75 | 37.5 |
|
68 |
+
| 52500 | 0.8485 | 48.5 | 119.5 | 5.0960 | 25.2934 | 98.84 | 12.375 | 28.875 | 34.5 |
|
69 |
+
| 55000 | 0.8889 | 48.5 | 118.0 | 5.0747 | 25.23 | 99.088 | 12.406 | 28.0 | 33.25 |
|
70 |
+
| 57500 | 0.9293 | 48.0 | 117.0 | 5.0698 | 25.2235 | 99.114 | 12.409 | 27.75 | 32.0 |
|
71 |
+
| 60000 | 0.9697 | 48.0 | 117.0 | 5.0651 | 25.2107 | 99.164 | 12.415 | 27.75 | 31.875 |
|
72 |
+
| 61875 | 1.0 | 48.0 | 117.5 | 5.0643 | 25.1856 | 99.263 | 12.428 | 27.625 | 32.0 |
|
73 |
|
74 |
# Resource Usage Comparison
|
75 |
|
76 |
+
- VRAM Use: 7.7843 GB
|
77 |
|
78 |
# Distillation (Teacher -> Student) Architecture Difference:
|
79 |
|
|
|
93 |
<br/>
|
94 |
|
95 |
# Train Dataset
|
96 |
+
Trained on 145,744,973 tokens from the [wikimedia/wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia) dataset.
|
97 |
|
98 |
- Num Samples: `247,500`
|
99 |
- Subset: `20231101.en`
|
|
|
103 |
# Training Objective
|
104 |
|
105 |
```
|
106 |
+
DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=10.0, loss_fn=cos, layer_mapper=layer-2))
|
107 |
```
|
108 |
|
109 |
# Hyperparameters
|
|
|
120 |
- lr_scheduler_type: `linear`
|
121 |
- lr_scheduler_warmup_ratio: `0.5`
|
122 |
- num_epochs: `1.0`
|
123 |
+
- distillation_objective: `DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=10.0, loss_fn=cos, layer_mapper=layer-2))`
|
124 |
- train_embeddings: `True`
|
125 |
+
- lr_scheduler: `<torch.optim.lr_scheduler.LambdaLR object at 0x7f6f2143c700>`
|
126 |
- student_model_name_or_path: `None`
|
127 |
- student_config_name_or_path: `None`
|
128 |
- student_model_config: `None`
|
logs/attn_loss_fn=cos, attn_weight=10.0, projector=identity/events.out.tfevents.1724327951.e3f806ea38c9
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9309b9cf44c77a5b34c6655145c99d44043ad6f6cb7f10c8c29258616fd20be2
|
3 |
+
size 588
|