Alina Kolesnikova commited on
Commit
a0cf62e
1 Parent(s): fcd48fc

Remove ipynb checkpoints

Browse files

Edit README

Upd. README with performance test

README typo fixed

.ipynb_checkpoints/README-checkpoint.md DELETED
@@ -1,16 +0,0 @@
1
- ---
2
- language:
3
- - ru
4
- ---
5
- # distilrubert-base-cased-conversational
6
- Conversational DistilRuBERT \(Russian, cased, 6‑layer, 768‑hidden, 12‑heads, 135.4M parameters\) was trained on OpenSubtitles\[1\], [Dirty](https://d3.ru/), [Pikabu](https://pikabu.ru/), and a Social Media segment of Taiga corpus\[2\] (as [Conversational RuBERT](https://huggingface.co/DeepPavlov/rubert-base-cased-conversational)).
7
-
8
- Our DistilRuBERT was highly inspired by \[3\], \[4\]. Namely, we used
9
- * KL loss (between teacher and student output logits)
10
- * MLM loss (between tokens labels and student output logits)
11
- * Cosine embedding loss between mean of two consecutive hidden states of the teacher and one hidden state of the student
12
-
13
- \[1\]: P. Lison and J. Tiedemann, 2016, OpenSubtitles2016: Extracting Large Parallel Corpora from Movie and TV Subtitles. In Proceedings of the 10th International Conference on Language Resources and Evaluation \(LREC 2016\)
14
- \[2\]: Shavrina T., Shapovalova O. \(2017\) TO THE METHODOLOGY OF CORPUS CONSTRUCTION FOR MACHINE LEARNING: «TAIGA» SYNTAX TREE CORPUS AND PARSER. in proc. of “CORPORA2017”, international conference , Saint-Petersbourg, 2017.
15
- \[3\]: Sanh, V., Debut, L., Chaumond, J., & Wolf, T. \(2019\). DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. arXiv preprint arXiv:1910.01108.
16
- \[4\]: <https://github.com/huggingface/transformers/tree/master/examples/research_projects/distillation>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
README.md CHANGED
@@ -10,7 +10,20 @@ Our DistilRuBERT was highly inspired by \[3\], \[4\]. Namely, we used
10
  * MLM loss (between tokens labels and student output logits)
11
  * Cosine embedding loss between mean of two consecutive hidden states of the teacher and one hidden state of the student
12
 
 
 
 
 
 
 
 
 
 
 
13
  \[1\]: P. Lison and J. Tiedemann, 2016, OpenSubtitles2016: Extracting Large Parallel Corpora from Movie and TV Subtitles. In Proceedings of the 10th International Conference on Language Resources and Evaluation \(LREC 2016\)
 
14
  \[2\]: Shavrina T., Shapovalova O. \(2017\) TO THE METHODOLOGY OF CORPUS CONSTRUCTION FOR MACHINE LEARNING: «TAIGA» SYNTAX TREE CORPUS AND PARSER. in proc. of “CORPORA2017”, international conference , Saint-Petersbourg, 2017.
 
15
  \[3\]: Sanh, V., Debut, L., Chaumond, J., & Wolf, T. \(2019\). DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. arXiv preprint arXiv:1910.01108.
 
16
  \[4\]: <https://github.com/huggingface/transformers/tree/master/examples/research_projects/distillation>
10
  * MLM loss (between tokens labels and student output logits)
11
  * Cosine embedding loss between mean of two consecutive hidden states of the teacher and one hidden state of the student
12
 
13
+ The model was trained for about 100 hrs. on 8 nVIDIA Tesla P100-SXM2.0 16Gb.
14
+
15
+ To evaluate improvements in the inference speed, we ran teacher and student models on random sequences with seq_len=512, batch_size = 16 (for throughput) and batch_size=1 (for latency).
16
+ All tests were performed on Intel(R) Xeon(R) CPU E5-2698 v4 @ 2.20GHz and nVIDIA Tesla P100-SXM2.0 16Gb.
17
+
18
+ | Model | Size, Mb. | CPU latency, sec.| GPU latency, sec. | CPU throughput, samples/sec. | GPU throughput, samples/sec. |
19
+ |-------------------------------------------------|------------|------------------|-------------------|------------------------------|------------------------------|
20
+ | Teacher (RuBERT-base-cased-conversational) | 679 | 0.655 | 0.031 | 0.3754 | 36.4902 |
21
+ | Student (DistilRuBERT-base-cased-conversational)| 517 | 0.3285 | 0.0212 | 0.5803 | 52.2495 |
22
+
23
  \[1\]: P. Lison and J. Tiedemann, 2016, OpenSubtitles2016: Extracting Large Parallel Corpora from Movie and TV Subtitles. In Proceedings of the 10th International Conference on Language Resources and Evaluation \(LREC 2016\)
24
+
25
  \[2\]: Shavrina T., Shapovalova O. \(2017\) TO THE METHODOLOGY OF CORPUS CONSTRUCTION FOR MACHINE LEARNING: «TAIGA» SYNTAX TREE CORPUS AND PARSER. in proc. of “CORPORA2017”, international conference , Saint-Petersbourg, 2017.
26
+
27
  \[3\]: Sanh, V., Debut, L., Chaumond, J., & Wolf, T. \(2019\). DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. arXiv preprint arXiv:1910.01108.
28
+
29
  \[4\]: <https://github.com/huggingface/transformers/tree/master/examples/research_projects/distillation>