Rémi Louf

The `while` case now works on simple examples. However I am currently having difficulties post-processing the storage of each return value, and more specifically only returning the number of values...

The JAX backend now has better support for `Scan`. I have the following remarks/questions: - `mit_mot`s are not currently supported. Should I add this? - The graphs generated by the...

- [ ] Slice with a dynamic index: ```python import aesara import aesara.tensor as at a = at.iscalar("a") x = at.arange(3) out = x[:a] aesara.dprint(out) # Subtensor{:int32:} [id A] #...

> As I recall, the trouble with using that is that it's limited to only the (outermost) graph inputs, and we can't compose `jax.jit`ed functions that use that setting. If...

The following code fails: ```python import aesara import aesara.tensor as at a_at = at.dvector("a") res, updates = aesara.scan( fn=lambda a_t: 2 * a_t, sequences=a_at ) jax_fn = aesara.function((a_at,), res, updates=updates,...

The following code also fails, because of an `Elemwise{add}` that transforms concrete values in a `TracedArray`: ```python import aesara import aesara.tensor as at from aesara.compile.mode import Mode from aesara.graph.rewriting.db import...

I am currently waiting for #1338 to be merged to see what else needs to be fixed in the backend to allow the tests to pass.

Solved in #1284.

Outlines author here. The PR https://github.com/outlines-dev/outlines/pull/366 will allow easy integration into vLLM. Estimated time of completion is next week. See https://github.com/outlines-dev/outlines/issues/163#issuecomment-1820441504 for a diagram that summarizes the new architecture. We...

Yes: https://outlines-dev.github.io/outlines/reference/vllm/ More is coming (soon)!