End of training
Browse files
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:
|
| 20 |
- eval_frwikippl: 776.0
|
| 21 |
-
- eval_zhwikippl:
|
| 22 |
-
- eval_tinystoriesppl:
|
| 23 |
-
- eval_loss:
|
| 24 |
-
- eval_runtime: 25.
|
| 25 |
-
- eval_samples_per_second: 98.
|
| 26 |
-
- eval_steps_per_second: 12.
|
| 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.
|
|
@@ -53,7 +53,7 @@ The following hyperparameters were used during training:
|
|
| 53 |
- eval_batch_size: 8
|
| 54 |
- seed: 42
|
| 55 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
-
- lr_scheduler_type:
|
| 57 |
- lr_scheduler_warmup_ratio: 0.5
|
| 58 |
- num_epochs: 1.0
|
| 59 |
|
|
@@ -64,69 +64,69 @@ Peak GPU Memory: 7.2012 GB
|
|
| 64 |
| step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl |
|
| 65 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
|
| 66 |
| **teacher eval** | | 43.25 | 61.25 | | | | | 11.6875 | 19.125 |
|
| 67 |
-
| 0 | 0 |
|
| 68 |
-
| 1000 | 0.0162 |
|
| 69 |
-
| 2000 | 0.0323 |
|
| 70 |
-
| 3000 | 0.0485 |
|
| 71 |
-
| 4000 | 0.0646 |
|
| 72 |
-
| 5000 | 0.0808 |
|
| 73 |
-
| 6000 | 0.0970 |
|
| 74 |
-
| 7000 | 0.1131 |
|
| 75 |
-
| 8000 | 0.1293 |
|
| 76 |
-
| 9000 | 0.1455 |
|
| 77 |
-
| 10000 | 0.1616 |
|
| 78 |
-
| 11000 | 0.1778 |
|
| 79 |
-
| 12000 | 0.1939 |
|
| 80 |
-
| 13000 | 0.2101 |
|
| 81 |
-
| 14000 | 0.2263 |
|
| 82 |
-
| 15000 | 0.2424 |
|
| 83 |
-
| 16000 | 0.2586 |
|
| 84 |
-
| 17000 | 0.2747 |
|
| 85 |
-
| 18000 | 0.2909 |
|
| 86 |
-
| 19000 | 0.3071 |
|
| 87 |
-
| 20000 | 0.3232 |
|
| 88 |
-
| 21000 | 0.3394 |
|
| 89 |
-
| 22000 | 0.3556 |
|
| 90 |
-
| 23000 | 0.3717 |
|
| 91 |
-
| 24000 | 0.3879 |
|
| 92 |
-
| 25000 | 0.4040 |
|
| 93 |
-
| 26000 | 0.4202 |
|
| 94 |
-
| 27000 | 0.4364 |
|
| 95 |
-
| 28000 | 0.4525 |
|
| 96 |
-
| 29000 | 0.4687 |
|
| 97 |
-
| 30000 | 0.4848 |
|
| 98 |
-
| 31000 | 0.5010 |
|
| 99 |
-
| 32000 | 0.5172 |
|
| 100 |
-
| 33000 | 0.5333 |
|
| 101 |
-
| 34000 | 0.5495 |
|
| 102 |
-
| 35000 | 0.5657 |
|
| 103 |
-
| 36000 | 0.5818 |
|
| 104 |
-
| 37000 | 0.5980 |
|
| 105 |
-
| 38000 | 0.6141 |
|
| 106 |
-
| 39000 | 0.6303 |
|
| 107 |
-
| 40000 | 0.6465 |
|
| 108 |
-
| 41000 | 0.6626 |
|
| 109 |
-
| 42000 | 0.6788 |
|
| 110 |
-
| 43000 | 0.6949 |
|
| 111 |
-
| 44000 | 0.7111 |
|
| 112 |
-
| 45000 | 0.7273 |
|
| 113 |
-
| 46000 | 0.7434 |
|
| 114 |
-
| 47000 | 0.7596 |
|
| 115 |
-
| 48000 | 0.7758 |
|
| 116 |
-
| 49000 | 0.7919 |
|
| 117 |
-
| 50000 | 0.8081 |
|
| 118 |
-
| 51000 | 0.8242 |
|
| 119 |
-
| 52000 | 0.8404 |
|
| 120 |
-
| 53000 | 0.8566 |
|
| 121 |
-
| 54000 | 0.8727 |
|
| 122 |
-
| 55000 | 0.8889 |
|
| 123 |
-
| 56000 | 0.9051 |
|
| 124 |
-
| 57000 | 0.9212 |
|
| 125 |
-
| 58000 | 0.9374 |
|
| 126 |
-
| 59000 | 0.9535 |
|
| 127 |
-
| 60000 | 0.9697 |
|
| 128 |
-
| 61000 | 0.9859 |
|
| 129 |
-
| 61875 | 1.0 |
|
| 130 |
|
| 131 |
### Framework versions
|
| 132 |
- 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: 210.0
|
| 20 |
- eval_frwikippl: 776.0
|
| 21 |
+
- eval_zhwikippl: 169.0
|
| 22 |
+
- eval_tinystoriesppl: 166.0
|
| 23 |
+
- eval_loss: 1.2670
|
| 24 |
+
- eval_runtime: 25.479
|
| 25 |
+
- eval_samples_per_second: 98.12
|
| 26 |
+
- eval_steps_per_second: 12.285
|
| 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.
|
|
|
|
| 53 |
- eval_batch_size: 8
|
| 54 |
- seed: 42
|
| 55 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 56 |
+
- lr_scheduler_type: inverse_sqrt
|
| 57 |
- lr_scheduler_warmup_ratio: 0.5
|
| 58 |
- num_epochs: 1.0
|
| 59 |
|
|
|
|
| 64 |
| step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl |
|
| 65 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
|
| 66 |
| **teacher eval** | | 43.25 | 61.25 | | | | | 11.6875 | 19.125 |
|
| 67 |
+
| 0 | 0 | 1477468749824.0 | 128093104635904.0 | 19.0496 | 25.4251 | 98.328 | 12.311 | 7985954816.0 | 67345087201280.0 |
|
| 68 |
+
| 1000 | 0.0162 | 73728.0 | 1736704.0 | 5.1613 | 25.443 | 98.259 | 12.302 | 15872.0 | 3735552.0 |
|
| 69 |
+
| 2000 | 0.0323 | 4000.0 | 42752.0 | 3.2656 | 25.4471 | 98.243 | 12.3 | 2688.0 | 544768.0 |
|
| 70 |
+
| 3000 | 0.0485 | 1120.0 | 8160.0 | 2.3670 | 25.4374 | 98.28 | 12.305 | 728.0 | 16192.0 |
|
| 71 |
+
| 4000 | 0.0646 | 620.0 | 4512.0 | 1.9891 | 25.4378 | 98.279 | 12.305 | 410.0 | 1464.0 |
|
| 72 |
+
| 5000 | 0.0808 | 422.0 | 2080.0 | 1.7081 | 25.4876 | 98.087 | 12.28 | 286.0 | 372.0 |
|
| 73 |
+
| 6000 | 0.0970 | 312.0 | 1296.0 | 1.4979 | 25.4407 | 98.268 | 12.303 | 231.0 | 237.0 |
|
| 74 |
+
| 7000 | 0.1131 | 262.0 | 996.0 | 1.4032 | 25.4618 | 98.186 | 12.293 | 209.0 | 193.0 |
|
| 75 |
+
| 8000 | 0.1293 | 226.0 | 860.0 | 1.3139 | 25.4752 | 98.135 | 12.286 | 179.0 | 221.0 |
|
| 76 |
+
| 9000 | 0.1455 | 210.0 | 776.0 | 1.2670 | 25.479 | 98.12 | 12.285 | 166.0 | 169.0 |
|
| 77 |
+
| 10000 | 0.1616 | 186.0 | 652.0 | 1.1738 | 25.4466 | 98.245 | 12.3 | 153.0 | 147.0 |
|
| 78 |
+
| 11000 | 0.1778 | 167.0 | 568.0 | 1.0682 | 25.5929 | 97.683 | 12.23 | 145.0 | 149.0 |
|
| 79 |
+
| 12000 | 0.1939 | 205.0 | 576.0 | 0.9983 | 25.4381 | 98.278 | 12.304 | 190.0 | 141.0 |
|
| 80 |
+
| 13000 | 0.2101 | 141.0 | 472.0 | 0.9487 | 25.4558 | 98.21 | 12.296 | 118.0 | 132.0 |
|
| 81 |
+
| 14000 | 0.2263 | 139.0 | 478.0 | 0.8964 | 25.4863 | 98.092 | 12.281 | 113.5 | 145.0 |
|
| 82 |
+
| 15000 | 0.2424 | 129.0 | 458.0 | 0.8617 | 25.5441 | 97.87 | 12.253 | 98.5 | 119.0 |
|
| 83 |
+
| 16000 | 0.2586 | 116.0 | 430.0 | 0.8348 | 25.4844 | 98.099 | 12.282 | 94.5 | 109.0 |
|
| 84 |
+
| 17000 | 0.2747 | 120.0 | 452.0 | 0.8054 | 25.4588 | 98.198 | 12.294 | 86.5 | 110.0 |
|
| 85 |
+
| 18000 | 0.2909 | 102.0 | 380.0 | 0.7811 | 25.4713 | 98.15 | 12.288 | 82.5 | 104.0 |
|
| 86 |
+
| 19000 | 0.3071 | 108.5 | 408.0 | 0.8030 | 25.4558 | 98.21 | 12.296 | 86.0 | 111.5 |
|
| 87 |
+
| 20000 | 0.3232 | 94.0 | 366.0 | 0.7253 | 25.5141 | 97.985 | 12.268 | 74.0 | 123.5 |
|
| 88 |
+
| 21000 | 0.3394 | 87.5 | 322.0 | 0.6815 | 25.4799 | 98.117 | 12.284 | 70.0 | 142.0 |
|
| 89 |
+
| 22000 | 0.3556 | 78.5 | 282.0 | 0.6363 | 25.4302 | 98.308 | 12.308 | 64.5 | 128.0 |
|
| 90 |
+
| 23000 | 0.3717 | 77.0 | 270.0 | 0.6005 | 25.4991 | 98.043 | 12.275 | 61.25 | 110.5 |
|
| 91 |
+
| 24000 | 0.3879 | 75.0 | 264.0 | 0.5879 | 25.4595 | 98.195 | 12.294 | 65.5 | 109.5 |
|
| 92 |
+
| 25000 | 0.4040 | 70.5 | 235.0 | 0.5705 | 25.5194 | 97.965 | 12.265 | 58.0 | 140.0 |
|
| 93 |
+
| 26000 | 0.4202 | 72.5 | 229.0 | 0.5606 | 25.5376 | 97.895 | 12.256 | 57.5 | 131.0 |
|
| 94 |
+
| 27000 | 0.4364 | 71.5 | 236.0 | 0.5460 | 25.5324 | 97.915 | 12.259 | 56.25 | 120.0 |
|
| 95 |
+
| 28000 | 0.4525 | 70.0 | 238.0 | 0.5403 | 25.5379 | 97.894 | 12.256 | 56.0 | 111.0 |
|
| 96 |
+
| 29000 | 0.4687 | 76.5 | 249.0 | 0.5596 | 25.5341 | 97.908 | 12.258 | 58.25 | 113.5 |
|
| 97 |
+
| 30000 | 0.4848 | 72.0 | 222.0 | 0.5487 | 25.5124 | 97.992 | 12.269 | 54.0 | 162.0 |
|
| 98 |
+
| 31000 | 0.5010 | 70.5 | 253.0 | 0.5503 | 25.4372 | 98.281 | 12.305 | 56.0 | 136.0 |
|
| 99 |
+
| 32000 | 0.5172 | 72.5 | 247.0 | 0.5404 | 25.5107 | 97.998 | 12.269 | 53.75 | 117.0 |
|
| 100 |
+
| 33000 | 0.5333 | 70.0 | 235.0 | 0.5380 | 25.5177 | 97.971 | 12.266 | 51.25 | 127.0 |
|
| 101 |
+
| 34000 | 0.5495 | 71.5 | 251.0 | 0.5315 | 25.5159 | 97.978 | 12.267 | 54.25 | 141.0 |
|
| 102 |
+
| 35000 | 0.5657 | 73.5 | 247.0 | 0.5319 | 25.4405 | 98.268 | 12.303 | 55.75 | 98.0 |
|
| 103 |
+
| 36000 | 0.5818 | 71.5 | 231.0 | 0.5224 | 25.4323 | 98.3 | 12.307 | 53.0 | 147.0 |
|
| 104 |
+
| 37000 | 0.5980 | 69.0 | 237.0 | 0.5154 | 25.4606 | 98.191 | 12.294 | 53.0 | 115.5 |
|
| 105 |
+
| 38000 | 0.6141 | 69.5 | 224.0 | 0.5031 | 25.5405 | 97.884 | 12.255 | 51.0 | 98.0 |
|
| 106 |
+
| 39000 | 0.6303 | 64.5 | 219.0 | 0.4864 | 25.4803 | 98.115 | 12.284 | 50.0 | 458.0 |
|
| 107 |
+
| 40000 | 0.6465 | 67.0 | 214.0 | 0.4865 | 25.5475 | 97.857 | 12.252 | 53.0 | 110.5 |
|
| 108 |
+
| 41000 | 0.6626 | 67.5 | 218.0 | 0.4886 | 25.4575 | 98.203 | 12.295 | 47.5 | 95.0 |
|
| 109 |
+
| 42000 | 0.6788 | 63.25 | 210.0 | 0.4788 | 25.43 | 98.309 | 12.308 | 51.0 | 92.0 |
|
| 110 |
+
| 43000 | 0.6949 | 63.0 | 203.0 | 0.4746 | 25.4807 | 98.113 | 12.284 | 48.25 | 84.0 |
|
| 111 |
+
| 44000 | 0.7111 | 62.75 | 211.0 | 0.4785 | 25.4599 | 98.194 | 12.294 | 48.5 | 90.0 |
|
| 112 |
+
| 45000 | 0.7273 | 63.5 | 205.0 | 0.4745 | 25.4377 | 98.279 | 12.305 | 47.25 | 92.5 |
|
| 113 |
+
| 46000 | 0.7434 | 63.25 | 204.0 | 0.4697 | 25.5287 | 97.929 | 12.261 | 45.5 | 116.0 |
|
| 114 |
+
| 47000 | 0.7596 | 63.75 | 229.0 | 0.4705 | 25.4923 | 98.069 | 12.278 | 49.0 | 101.5 |
|
| 115 |
+
| 48000 | 0.7758 | 63.75 | 205.0 | 0.4669 | 25.4951 | 98.058 | 12.277 | 46.0 | 75.5 |
|
| 116 |
+
| 49000 | 0.7919 | 62.75 | 200.0 | 0.4651 | 25.5061 | 98.016 | 12.272 | 49.75 | 87.0 |
|
| 117 |
+
| 50000 | 0.8081 | 63.25 | 200.0 | 0.4666 | 25.5442 | 97.869 | 12.253 | 46.5 | 149.0 |
|
| 118 |
+
| 51000 | 0.8242 | 62.75 | 191.0 | 0.4644 | 25.4188 | 98.353 | 12.314 | 46.5 | 85.0 |
|
| 119 |
+
| 52000 | 0.8404 | 61.25 | 187.0 | 0.4531 | 25.5074 | 98.011 | 12.271 | 47.25 | 62.75 |
|
| 120 |
+
| 53000 | 0.8566 | 60.25 | 207.0 | 0.4613 | 25.4158 | 98.364 | 12.315 | 44.5 | 129.0 |
|
| 121 |
+
| 54000 | 0.8727 | 65.0 | 209.0 | 0.4588 | 25.4519 | 98.225 | 12.298 | 45.5 | 95.0 |
|
| 122 |
+
| 55000 | 0.8889 | 60.25 | 196.0 | 0.4571 | 25.4656 | 98.172 | 12.291 | 45.5 | 75.5 |
|
| 123 |
+
| 56000 | 0.9051 | 63.75 | 195.0 | 0.4519 | 25.5712 | 97.766 | 12.24 | 46.0 | 82.0 |
|
| 124 |
+
| 57000 | 0.9212 | 65.5 | 212.0 | 0.4563 | 25.4339 | 98.294 | 12.306 | 46.0 | 100.0 |
|
| 125 |
+
| 58000 | 0.9374 | 62.5 | 204.0 | 0.4474 | 25.4328 | 98.298 | 12.307 | 45.25 | 80.5 |
|
| 126 |
+
| 59000 | 0.9535 | 62.0 | 214.0 | 0.4461 | 25.4201 | 98.347 | 12.313 | 44.0 | 85.0 |
|
| 127 |
+
| 60000 | 0.9697 | 62.0 | 211.0 | 0.4494 | 25.5124 | 97.992 | 12.269 | 44.25 | 85.0 |
|
| 128 |
+
| 61000 | 0.9859 | 62.0 | 219.0 | 0.4483 | 25.4698 | 98.155 | 12.289 | 46.0 | 62.75 |
|
| 129 |
+
| 61875 | 1.0 | 61.25 | 203.0 | 0.4444 | 25.5209 | 97.959 | 12.264 | 45.5 | 100.0 |
|
| 130 |
|
| 131 |
### Framework versions
|
| 132 |
- Distily 0.2.0
|
logs/lr_scheduler_type=inverse_sqrt, warmup_ratio=0.5/events.out.tfevents.1724187476.f383272e719b
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4538f4c6cdaacdc1cd047c19ca528f739cfdc3fa8eacb6891952317ee1b9bd38
|
| 3 |
+
size 312
|