PAB
The goal of this PR is to add the axis option to the `np.linalg.norm` function in NoPython mode. Some issues request the support of this feature: #2181 #2558 #3116 Roadmap:...
The goal of this PR is to add multi-task ElasticNet to the `elasticnet` crate. A quick roadmap: - [x] Write block coordinate descent - [x] Write dual gap for multi-task...
I can certainly implement MultiTaskLasso, as well as GroupLasso, in the ElasticNet package. Would you consider a valuable addition to the crate?
In high dimensional statistics, we usually rely on CSC and CSR arrays when matrices are very sparse. It is particularly for speeding up the coordinate descent procedure. I can implement...
I tried solving a simple linear system with generics `T` having the `Float` trait, but I couldn't make it work. Here is a short snippet of code: ``` // A...
I'm running into an issue when generating the HTML report: ``` Writing results to outputs/benchmark_multi_task_lasso_benchopt_run_2021-12-22_15h42m48.html Traceback (most recent call last): File "/Users/pierre-antoine/opt/miniconda3/envs/benchopt/bin/benchopt", line 33, in sys.exit(load_entry_point('benchopt', 'console_scripts', 'benchopt')()) File "/Users/pierre-antoine/opt/miniconda3/envs/benchopt/lib/python3.8/site-packages/click/core.py",...
As discussed [here](https://github.com/ggerganov/llama.cpp/issues/1328#issuecomment-1547007119), this PR adds to the example section the BioGPT model. Apart from minor architectural changes, the model remains very similar to GPT-2 (medium). The main difference is...
Hi! I just released [encodec.cpp](https://github.com/PABannier/encodec.cpp/tree/main) which features a handful of new operations, like transposed 1d convolution, argmax and activation functions like `elu` and `tanh`. Is it something you'd like implemented...
This PR adds the `tanh` and `elu` activations required in the `Encodec` model as well as the `argmax` function. Closes #281
A quick proof-of-concept of a function that checks if the combination `(solver, datafit, penalty)` is supported. Currently we have some edge cases where one can pass `ProxNewton` solver with `L0_5`...