georgematheos
georgematheos
Here is a MWE to illustrate the bug: ```julia julia> using Gen julia> @gen (static) function foo() {:x => :y} ~ normal(0, 1) end var"##StaticGenFunction_foo#269"(Dict{Symbol, Any}(), Dict{Symbol, Any}()) julia> @load_generated_functions()...
Here's some progress on #129 from the hackathon. Some more debugging work is needed, as well as writing a bit of documentation. The tests appeared to be passing when I...
I find I get `not defined` errors when using the kernel DSL to choose which kernel to use depending on a bernoulli sample. As an example: ``` julia> @kern function...
Caching some work from the hackathon today. To enable propagating change hints when constructing objects which hold several values and unpacking values from these objects, we introduce a `Construct(type)` generative...
Currently, it appears that the dynamic DSL does not throw a method error when variable types are used which do not match the definition. Presumably, in the parser, in the...
Currently, if we declare a dist DSL function like ``` @dist foo(x::Int) = normal(0, 1) ``` and then call `foo("hello")`, this does not throw an error. A type error _is_...
This recursive function in the `@dist` DSL file worries me: ```julia eval_arg(x::TransformedArg, args) = x.arg_passer(x.orig_f, [eval_arg(a, args) for a in x.f_args]...) ``` This is a classic "implement `fib` with vs...
Currently debugging static DSL generative functions can be quite difficult, since the compiled static DSL code don't include any line number markers from the original code, which makes it difficult...
I've been finding that I've needed a vector-shaped choicemap for several custom generative functions I've had to implement recently. I figure we could add a simple implementation of a generic...
This PR builds upon #279 to change the signature of `update` to the following: ```julia (new_tr, weight, retdiff, reverse_update_spec) = update(trace, args, argdiffs, update_spec::UpdateSpec, externally_constrained_addresses::Selection) ``` `update_spec` is a specific...