Issues
- 0
- 4
Fast_jl undefined symbol
#67 opened by Bas-2k - 2
RuntimeError: CUDA error: the provided PTX was compiled with an unsupported toolchain.
#68 opened by simplelifetime - 2
- 1
Request for a ChatBot Example
#66 opened by Taha-Bahadori - 1
Support the LDS metric
#55 opened by Jiaxin-Wen - 2
Error when execute `finalize_scores`
#53 opened by Jiaxin-Wen - 5
- 4
- 4
Allow taking gradients with respect to a subset of the model's parameters
#1 opened by kristian-georgiev - 4
Parallelized score finalization?
#45 opened by nickk124 - 2
Order of Arguments in LLM can mismatch
#46 opened by awe2 - 1
RuntimeError: input.dtype() == torch::kFloat16 || input.dtype() == torch::kFloat32 error on QNLI
#48 opened by rishabluthra - 2
- 1
- 6
all NaN or all zero scores?
#41 opened by nickk124 - 0
Can trak be applied to diffusion models?
#37 opened by xszheng2020 - 1
- 3
- 2
- 1
Save scores in memory-mapped files and allow loading existing scores from the `TRAKer` class
#33 opened by kristian-georgiev - 5
Support dataset sharding
#29 opened by scoutsaachi - 5
CIFAR-10 Trak Scores Reproduction
#22 opened by vasusingla - 1
Add a colab where one can download pre-computed TRAK scores and visualize them.
#31 opened by kristian-georgiev - 3
- 1
- 1
- 1
Add `allow_skip` to `finalize_scores`
#19 opened by kristian-georgiev - 10
[cifar_quickstart.ipynb] _LinAlgError: linalg.inv: The diagonal element 1 is zero, the inversion could not be completed because the input matrix is singular.
#24 opened by xszheng2020 - 1
CudaProjector gives `CUDA error: too many resources requested for launch` on V100
#26 opened by kristian-georgiev - 0
Implement `zarr` saver
#2 opened by kristian-georgiev - 1
FAILED tests/test_rademacher.py::test_odd - AssertionError: Expected <True>, but was not.
#23 opened by xszheng2020 - 1
Tutorial for CLIP model output
#16 opened by kristian-georgiev - 1
NLP model output
#15 opened by kristian-georgiev - 1
Concurrent TRAKer init (with diff model id) in same save dir may lead to I/O errors
#14 opened by kristian-georgiev - 1
- 1
Make `fast_jl` a separate package, add it as an optional dependency to `trak`
#11 opened by kristian-georgiev - 2
Implement soft thresholding
#10 opened by kristian-georgiev - 6
Proper docstrings
#3 opened by kristian-georgiev - 2
- 0
- 1