Issues
- 0
Problem with serialization
#236 opened by keyvan-amiri - 2
OOM when try to fit LA on ViT
#232 opened by ruili-pml - 0
Confusing behavior for one-dimensional targets
#235 opened by joemrt - 0
Strict type checking in the CI
#234 opened by wiseodd - 0
- 0
Rewrite documentation site
#226 opened by wiseodd - 3
Optimization of prior noise
#222 opened by cbuelt - 7
- 3
`test_gp_kernel` error "fixture 'mocker' not found"
#214 opened by elcorto - 2
`FullSubnetLaplace.sample()` doesn't use `generator`
#213 opened by elcorto - 2
- 2
Efficient, universal, standalone Jacobian backend
#203 opened by wiseodd - 0
Generic wrapper for Huggingface transformers
#206 opened by wiseodd - 0
Implement KronLLLaplace fast functional variance
#201 opened by wiseodd - 1
Support for structured Kronecker factors
#174 opened by wiseodd - 0
Fast `KronLLLaplace` joint predictive
#175 opened by wiseodd - 0
- 0
Implement heteroscedastic Laplace
#153 opened by wiseodd - 0
Support for `torchtune`'s models and datasets
#162 opened by wiseodd - 1
Conform to PyTorch convention in the loss
#163 opened by wiseodd - 0
Optimize `torch.func` full/diag GGN/Fisher-MC
#149 opened by wiseodd - 2
`torch.func` jacobian OOM
#186 opened by wiseodd - 2
Fix defaults and suggestions
#147 opened by aleximmer - 8
Fit function of FullLLLaplace runs slower in Github version in comparison to pip version
#193 opened by MJordahn - 3
- 1
- 3
Add "Contributing" section to README
#196 opened by wiseodd - 1
FeatureExtractor and Python <3.11
#195 opened by Ludvins - 1
Add linting and formatting via `ruff`
#152 opened by runame - 0
- 0
Handle feature reduction properly in LLLA
#169 opened by wiseodd - 0
Adding prior precision in `KronLaplace` is broken
#171 opened by wiseodd - 0
Efficient LLLA predictive samples
#161 opened by wiseodd - 5
- 0
Generalized dict-input mechanism
#165 opened by wiseodd - 0
Add example & doc for reward modeling functionality
#166 opened by wiseodd - 2
- 6
subnetwork with Kronecker covariance
#189 opened by ruili-pml - 2
- 0
- 0
Remove default `loss` for `optimize_prior_precision`
#156 opened by runame - 1
Cache features in LLLA with MC predictive
#164 opened by wiseodd - 2
Serialization: `torch.load(..., map_location="cpu")` doesn't map all tensors in `Laplace` instance
#158 opened by elcorto - 2
- 9
- 4
- 1
How can I use pretrained models with inplace ReLUs?
#139 opened by arunpatro - 9
- 2
underconfident predictions
#141 opened by Snowmanda - 2
Saving a laplace approximation module after `fit`
#140 opened by PaulScemama