>
G
Model Details
Domain:
Model Access:
Open weights (unrestricted)
Citations:
304
Introduction
In this work we present the experiments which lead to the creation of our BERT and ELECTRA based German language models, GBERT and GELECTRA. By varying the input training data, model size, and the presence of Whole Word Masking (WWM) we were able to attain SoTA performance across a set of document classification and named entity recognition (NER) tasks for both models of base and large size. We adopt an evaluation driven approach in training these models and our results indicate that both adding more data and utilizing WWM improve model performance. By benchmarking against existing German models, we show that these models are the best German models to date. Our trained models will be made publicly available to the research community.
Benchmarking
FLOPs
1.43e+21Notes: flops = (64) * (123* 10**12) * (7 * 24 * 3600) * (0.3) = 1.4e21 (num gpu) * (peak flops) * (time in seconds) * (assumed utilization rate) 'large models were trained on pods of 16 TPUs v3 (128 cores).' - from section 4.1 it was trained for 7 days from Table 2 Agrees with 6CN: Tokens seen: 512 (seq len) * 1024 (batch size) * 1 million (steps) = 5.24e11 FLOPs: 6 * 335M * 5.24e11 = 1.05e21
Training
Training Code Accessibility
MIT: https://huggingface.co/deepset/gelectra-largeHardware
Google TPU v3Hardware Quantity
64Size Notes: 163.4GB from Table 1 in the paper assuming 167M words per GB (German Language) we have 163.4 * 167M * 4/3 tokens per word = 36,383,733,333
Parameters
Parameters
335000000Notes: 335M from Table 5
Authors
Branden Chan, Stefan Schweter, Timo Möller