mattearllongshot
mattearllongshot
> N.B. Your example is a little strange, because `shared_val`'s value is actually zero-dimensional, but its `TensorType` is a vector. Nevertheless, if you use a different value and set `shared_val...
Never mind, I see that #1122 is tracking this, thanks. Should this issue (#1119) be closed since the two particular issues that it refers to are being tracked independently?
Confirmed that #1124 does indeed speed things up for us: https://github.com/aesara-devs/aesara/pull/1124#issuecomment-1220687554
Hello, I've just tried numba, and it seems to work well. The first iteration takes a long time (presumably it is compiling) but afterwards it is about 5% faster than...
Hello, I just tested this, but I don't see any output on stderr: ```bash $ cat profile_simple.py import aesara import aesara.tensor as at import numpy as np params = at.vector()...
Hello, `PYTHONHASHSEED` does indeed fix this, although it's not ideal since we want to be able to keep the benefits of randomized hashing (eg. avoid dictionary keying attacks).
Here's what the graph looks like for the minimal repro:
> Does it keep increasing in every run, or just from run 1 to run 2? This minimal example only increases by 1 although larger graphs increase on subsequent runs...