patrickvonplaten Follow

Patrick von Platen patrickvonplaten

96 followers · 22 following · 56

Highlights

Arctic Code Vault Contributor
Pro

Organizations

Popular repositories

notebooks

Some notebooks for NLP

Jupyter Notebook 15 5
AdvertisementDetection

Python 2
TRexGameRL

JavaScript 2
course_work

TeX 1 1
DM_Stat_avance

TeX
AdvancedAutomaticSpeechRecognition

Python

1,089 contributions in the last year

Contribution activity

July 2020

patrickvonplaten/tokenizers Rust Jul 5

Created a pull request in huggingface/transformers that received 7 comments

Jul 3

[Generation] better error message

If cur_len of input context is as long or longer than max_length a nice error message should be shown.

+8 −0 • 7 comments

Finalize Reformer blog with fixed Latex Jul 10
Reformer - fix more underscore Jul 3
fix reformer underscore Jul 3
fix reformer formulas Jul 3
fix latex in reformer blog Jul 3
Never doing words in latex in markdown again Jul 3
try out new word regex Jul 3
Last time words fixing in generate Jul 3
fix words Jul 3
Fix words generate Jul 3
fix generate blog in markdown Jul 3
Fix how to generate Jul 3
Add reformer notebook Jul 3

[Benchmark] Fix models without `architectures` param in config Jul 16
Ensure OpenAI GPT position_ids is correctly initialized and registered at init. Jul 15
Addition of a DialoguePipeline Jul 15
[cleanup] T5 test, warnings Jul 15
[Reformer] - Cache hidden states and buckets to speed up inference Jul 14
doc: fix apparent copy-paste error in docstring Jul 14
Cleanup bart caching logic Jul 14
[fix] mbart_en_ro_generate test now identical to fairseq Jul 14
[Reformer classification head] Implement the reformer model classification head for text classification Jul 13
[Longformer] fix longformer global attention output Jul 13
FlaubertForTokenClassification Jul 13
Fixed use of memories in XLNet (caching for language generation + warning when loading improper memoryless model) Jul 10
Improvements to PretrainedConfig documentation Jul 10
Fixed TextGenerationPipeline on torch + GPU Jul 9
[Benchmark] Add benchmarks for TF Training Jul 8
Add mbart-large-cc25, support translation finetuning Jul 7
[Reformer] Adapt Reformer MaskedLM Attn mask Jul 7
Added data collator for permutation (XLNet) language modeling and related calls Jul 6
The `add_space_before_punct_symbol` is only for TransfoXL Jul 6
Refactor generation sampling parameters (e.g. top k, temperature) into "Sampling" classes Jul 4
Fix saved model creation Jul 3
[Bart] enable test_torchscript, update test_tie_weights Jul 3
QA pipeline BART compatible Jul 3
Change model outputs types to self-document outputs Jul 3
[Almost all TF models] TF clean up: add missing CLM / MLM loss; fix T5 naming and keras compile Jul 3
Some pull request reviews not shown.

Update dataset loading and features - Add TREC dataset Jul 13
Add text dataset Jul 10

Created an issue in google/trax that received 1 comment

Jul 6

Question/Potential Issue - Attention mask in SelfAttention Layer

Description It seems to me that it is not possible to mask the first position of a tensor when using the layer SelfAttention. Looking at this line:

1 comment

Create your own GitHub profile