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

Where is the sum operation of KV?

#82 Yogurt928 closed 3 years ago
3
TransformerDecoderBuilder: decoder only self attention example

#81 patdflynn closed 3 years ago
3
Queries scaling is not consistent for recurrent wrappers

#80 hadaev8 closed 3 years ago
3
Model diff between commits

#79 danieltudosiu closed 3 years ago
2
Rename pyptoject.toml to pyproject.toml

#78 LoicGrobol closed 3 years ago
5
Enable faster CUDA kernels for the causal product

#77 jdemouth closed 3 years ago
1
Add PEP 518 build system/requirements spec (pyproject.toml)

#76 LoicGrobol closed 3 years ago
1
Some errors when I run the Quick-start code

#75 yyysjz1997 closed 3 years ago
2
Add support for half- and double-precision floats to CausalDotProduct (CUDA)

#74 norabelrose closed 2 years ago
5
Implementing `tensorflow` version of `CausalDotProduct`

#73 JamesDeAntonis opened 3 years ago
1
Add event dispatcher to recurrent attention

#72 hadaev8 closed 3 years ago
0
RecurrentCrossFullAttention have no event dispatcher for attention weight

#71 hadaev8 closed 3 years ago
3
Linear Transformers are Fast Weight Memory Systems

#70 angeloskath opened 3 years ago
0
How to install without compiling linear attention?

#69 hadaev8 opened 3 years ago
0
Memory usage: native PyTorch vs. "full"-Attention

#68 GregorKobsik closed 3 years ago
9
TypeError: forward() missing 2 required positional arguments: 'query_lengths' and 'key_lengths'

#67 mHsuann closed 3 years ago
1
Could this be used for faster-training of transformer-based object detection?

#66 nicolasugrinovic closed 3 years ago
2
problems installing on Ubuntu 18.04, pytorch 1.71, CUDA 9.1 / CUDA 10.2

#65 rainwala closed 3 years ago
1
Bug in ExactTopKAttention: "selected index k out of range"

#64 mlw214 closed 3 years ago
1
No module named 'fast_transformers.causal_product.causal_product_cpu'

#63 lonce closed 3 years ago
5
Training with fp16 by DeepSpeed

#62 e4exp opened 3 years ago
0
Dynamically determine extra_compile_args based on operating system.

#61 TariqAHassan closed 3 years ago
1
Feature request: L2 self-attention

#60 ketyi opened 3 years ago
0
allow distinct memory and decoder dimensionalities

#59 konstantinosKokos closed 3 years ago
2
CUDA problems in causal linear product

#58 xyltt closed 3 years ago
8
How to step into linear_attention.py/forward function ?

#57 Yogurt928 closed 3 years ago
1
Should some memory be saved with in place masking?

#56 hadaev8 closed 3 years ago
1
Any chance for pre-built binaries?

#55 umbertov opened 3 years ago
1
Is inconsitent axis order in RecurrentFullAttention intended?

#54 hadaev8 opened 3 years ago
0
Feature request for relative position encoding

#53 hadaev8 opened 3 years ago
9
Image Generation/Completion Training-Code

#52 tibuch closed 4 years ago
3
Local Product CUDA Kernel

#51 AndriyMulyar closed 3 years ago
1
A different windows installation error.

#50 codeninja opened 4 years ago
4
Added feature_redraw_interval feature for FAVOR

#49 norabelrose closed 3 years ago
2
windows installation error linking local_product_cuda.cu

#48 lm-b closed 4 years ago
7
Feature Maps without using builders

#47 adamsolomou closed 3 years ago
6
Expected usage of `length_masks` in `TransformerEncoder.forward`

#46 xvr-hlt closed 4 years ago
2
None type error with local attention

#45 benderama3 closed 4 years ago
3
Do not work with pytorch 1.7

#44 hadaev8 closed 4 years ago
1
Cannot install fast-transformers (Win10, Torch 1.6.0)

#43 FilipAndersson245 opened 4 years ago
13
RuntimeError: CUDA error: invalid argument when running tests/attention/test_improved_clustered_transformer_gpu.py

#42 justimyhxu closed 4 years ago
8
install error

#41 AceCoooool closed 4 years ago
1
Performer integration request

#40 ketyi closed 4 years ago
2
forward() got multiple values for argument 'state'

#39 hadaev8 closed 4 years ago
2
Implementation of random Fourier features

#38 angeloskath closed 4 years ago
11
Local attention returning nan when using mask

#37 bratao closed 4 years ago
1
Replace long with int64_t

#36 bratao closed 4 years ago
1
Very minor typo in Readme

#35 bratao closed 4 years ago
0
Fix build on Windows

#34 bratao closed 4 years ago
1
Huggingface implementation

#33 flozi00 closed 1 year ago
2

Previous Next