laplace-torch v0.3 aims to bring extensive support to diffusion models. These models are challenging due to their (often images) output dimensionality. Considering that the linearized Laplace requires the (n_outputs, n_params)-shaped Jacobians, this can be very challenging (e.g. images have n_outputs = width * height).
All features we aim for are:
- Efficient, universal, ultimate Jacobian backend
- Functional Laplace with subset-of-data support
- Heteroskedastic Laplace
- Fast last-layer predictive
List view
0 issues of 24 selected
- Status: Open.#75 In aleximmer/Laplace;
- Status: Open.#81 In aleximmer/Laplace;
- Status: Open.#111 In aleximmer/Laplace;
- Status: Open.#118 In aleximmer/Laplace;
- Status: Open.#119 In aleximmer/Laplace;
- Status: Open.#130 In aleximmer/Laplace;
- Status: Open.#149 In aleximmer/Laplace;
- Status: Open.#153 In aleximmer/Laplace;
- Status: Open.#162 In aleximmer/Laplace;
- Status: Open.#163 In aleximmer/Laplace;
- Status: Open.#174 In aleximmer/Laplace;
- Status: Open.#175 In aleximmer/Laplace;
- Status: Open.#178 In aleximmer/Laplace;
- Status: Open.#201 In aleximmer/Laplace;
- Status: Open.#203 In aleximmer/Laplace;
- Status: Open.#206 In aleximmer/Laplace;
- Status: Open.#20 In aleximmer/Laplace;
- Status: Open.#24 In aleximmer/Laplace;
- Status: Open.#29 In aleximmer/Laplace;
- Status: Open.#31 In aleximmer/Laplace;
- Status: Open.#43 In aleximmer/Laplace;
- Status: Open.#50 In aleximmer/Laplace;
- Status: Open.#70 In aleximmer/Laplace;
- Status: Open (in progress).aleximmer/Laplacenumber 177#177 In aleximmer/Laplace;