Closed cbalkig closed 4 years ago
FYI, in case you're using huggingface transformers, you should be able to load the PyTorch weights into a TF2 model, using the from_pt=True
flag.
(of course, we could also convert the weights pretty easily and upload to S3, but in the meantime, you have this option)
@julien-c I tried to run this
TFBertModel.from_pretrained("dbmdz/bert-base-turkish-cased", from_pt=True) and got the error below.
OSError: Loading a TF model from a PyTorch checkpoint is not supported when using a model identifier name.
I'm trying this in google-colab with tensorflow 2.1.0 and gpu acclerator. Am I missing something?
@balki7 Sure, here are the original TF checkpoints (best checkpoint, incl. config and vocab):
wget https://schweter.eu/cloud/bert-base-turkish-cased/bert-base-turkish-cased-tf.tar.gz
# sha256 bert-base-turkish-cased-tf.tar.gz
# 8113d0aeb32a2e7bcd00027195a13622387ca9e2132d7f9a1b27389a0db26b96
@balki7 Sure, here are the original TF checkpoints (best checkpoint, incl. config and vocab):
wget https://schweter.eu/cloud/bert-base-turkish-cased/bert-base-turkish-cased-tf.tar.gz # sha256 bert-base-turkish-cased-tf.tar.gz # 8113d0aeb32a2e7bcd00027195a13622387ca9e2132d7f9a1b27389a0db26b96
Thank u so much Stefan.
Hi,
Is it possible to share TF checkpoints? Thanx a lot.