idiap fast-transformers issues

idiap / fast-transformers

Pytorch library for fast transformer implementations

1.65k stars 179 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Support for 16-bit Floats

#32 anti4m opened 4 years ago
7
What is best way to perform recurrent sampling while training?

#31 hadaev8 closed 3 years ago
27
Relative Position Representations

#30 burcehan closed 4 years ago
5
Masking extension

#29 angeloskath opened 4 years ago
1
feature map function φ (x) = elu(x) + 1,

#28 burcehan closed 4 years ago
4
Segment-Level Recurrence with State Reuse

#27 burcehan closed 4 years ago
3
n_heads param in TransformerEncoderLayer unused and inconsistent with TransformerDecoderLayer

#26 hadaev8 closed 4 years ago
2
Error with recurrent attention ValueError: too many values to unpack (expected 2)

#25 hadaev8 closed 4 years ago
11
causal_product_cuda.cu

#24 burcehan closed 4 years ago
0
causal_product_cuda.cu，Error compiling objects for extension

#23 burcehan closed 4 years ago
4
Regarding arbitrary mask

#22 netw0rkf10w closed 4 years ago
6
Its not convient (or possible) to use checkpointing in pytorch because masks is not tensor type objects

#21 hadaev8 opened 4 years ago
4
How to visualize attention weights?

#20 hadaev8 closed 4 years ago
17
RuntimeError: CUDA error: an illegal memory access was encountered

#19 annahung31 closed 4 years ago
1
Cant install on colab

#18 hadaev8 closed 4 years ago
3
Positional embedding

#17 horiacristescu closed 4 years ago
2
Problems Installing on Debian GNU/Linux 10 (buster) using Python 3.7.3

#16 maximilianreimer closed 4 years ago
5
Please Add Pytorch to Be Automatically Installed

#15 maximilianreimer closed 4 years ago
2
Image generation/completion

#14 lorinczszabolcs closed 4 years ago
2
Ensure all forward() methods have a proper description of input tensors

#13 angeloskath opened 4 years ago
0
No module named 'fast_transformers.causal_product.causal_product_cpu' (solved: needed to at CUDA to the PATH)

#12 ghost closed 4 years ago
8
TO DO in Causal Linear denomenator

#11 eelxpeng closed 4 years ago
1
Replace inplace operations in recurrent linear attention

#10 TariqAHassan closed 4 years ago
1
fix extremely minor typo

#9 bionicles closed 4 years ago
0
please add comments with meaning of einsum dimensions

#8 bionicles closed 4 years ago
2
n_heads parameter of RecurrentTransformerEncoderLayer is not currently used

#7 TariqAHassan closed 4 years ago
5
Unable to install extensions

#6 TariqAHassan closed 3 years ago
3
Experimental Code - CTC Loss issue.

#5 iiSeymour closed 4 years ago
4
Training

#4 Uglj closed 4 years ago
3
Encoder-decoder setup?

#3 ghost closed 4 years ago
17
hash_cuda.cu: No such file or directory

#2 pathway closed 4 years ago
5
Why this implementation is fast and how fast is it?

#1 jiqiujia closed 4 years ago
2