PyMC4 thoughts that can be on top of pytorch
Come on it desire a threat of its own
We’re still working on verifying and refactoring the sequential monte carlo sampler (courtesy of some hard work from hvasbath!) – I think @junpenglao has some good ideas on extending that
Some ideas yes
Mainly trying to refactor SMC and extent it to better integrate with Approximate Bayesian Computation methods.
Also super excited about getting a machine to run benchmarks in various languages!
Would be very cool to have some mix of speed comparisons (a la Julia benchmarks) and style comparisons (a la Bob Carpenter’s Hello, World! essay).