Relax nonzero dimension requirement on HMC (#39)

Removes need for fixed_param

.github/workflows/main.yml

@@ -8,7 +8,7 @@ on:
   workflow_dispatch: {}
 
 env:
-  CACHE_VERSION: 1
+  CACHE_VERSION: 2
 
 # only run one copy per PR
 concurrency:

TODO.md

@@ -4,7 +4,7 @@
 - [x] Add nicer ability to build models from source in the languages
     - [x] download source if needed, similar to bridgestan
     - [x] Version checking
-- [ ] Fixed param sampler for 0 dimension parameters?
+- [-] ~Fixed param sampler for 0 dimension parameters?~
 - [ ] Add wrapper around generate quantities method?
 - [x] Add wraper around laplace sampling?
 - [x] Pathfinder: expose the no lp/no PSIS version

clients/R/R/tinystan.R

@@ -167,9 +167,6 @@ sampler.tinystan_model = function(model, data = "", num_chains = 4, inits = NULL
 
     with_model(model, data, seed, {
         free_params <- get_free_params(model, model_ptr)
-        if (free_params == 0) {
-            stop("Model has no parameters to sample")
-        }
 
         params <- c(HMC_SAMPLER_VARIABLES, get_parameter_names(model, model_ptr))
         num_params <- length(params)

clients/R/tests/testthat/test_sample.R

@@ -200,7 +200,7 @@ test_that("bad num_warmup handled properly", {
 
 test_that("model with no params fails", {
 
-    expect_error(sampler(empty_model), "Model has no parameters")
+    expect_no_error(sampler(empty_model))
 
 })
 

clients/julia/src/model.jl

@@ -250,9 +250,6 @@ function sample(
 
     with_model(model, data, seed) do model_ptr
         free_params = num_free_params(model, model_ptr)
-        if free_params == 0
-            error("Model has no parameters to sample")
-        end
 
         param_names = cat(HMC_SAMPLER_VARIABLES, get_names(model, model_ptr), dims = 1)
         num_params = length(param_names)

clients/julia/test/test_sample.jl

@@ -333,7 +333,9 @@
     end
 
     @testset "Model without parameters" begin
-        @test_throws "Model has no parameters to sample" sample(empty_model)
+        (names, draws, metric) = sample(empty_model; save_metric=true)
+        @test length(names) == 7 # HMC parameters only
+        @test prod(size(metric)) == 0
     end
 
     @testset "Bad num_warmup" begin

clients/python/tests/test_sample.py

@@ -310,9 +310,9 @@ def test_bad_num_warmup(bernoulli_model):
 
 
 def test_model_no_params(empty_model):
-    with pytest.raises(ValueError, match="Model has no parameters to sample"):
-        empty_model.sample()
-
+    fit = empty_model.sample(save_metric=True)
+    assert len(fit.parameters) == 7 # just HMC parameters
+    assert fit.metric.size == 0
 
 @pytest.mark.parametrize(
     "arg, value, match",

clients/python/tinystan/model.py

@@ -557,8 +557,6 @@ def sample(
 
         with self._get_model(data, seed) as model:
             model_params = self._num_free_params(model)
-            if model_params == 0:
-                raise ValueError("Model has no parameters to sample.")
 
             param_names = HMC_SAMPLER_VARIABLES + self._get_parameter_names(model)
 

clients/typescript/src/model.ts

@@ -166,9 +166,6 @@ export default class StanModel {
       const n_params = paramNames.length;
 
       const free_params = this.m._tinystan_model_num_free_params(model);
-      if (free_params === 0) {
-        throw new Error("Model has no parameters to sample.");
-      }
 
       // TODO: allow init_inv_metric to be specified
       const init_inv_metric_ptr = NULL;

clients/typescript/test/model.test.ts

@@ -33,9 +33,11 @@ describe("test tinystan code with a mocked WASM module", () => {
       const { mockedModule, model } = await getMockedModel({
         numFreeParams: 0,
       });
-      expect(() => model.sample({})).toThrow(/no parameters/);
+      const {metric} = model.sample({save_metric:true});
 
-      expect(mockedModule._tinystan_sample).toHaveBeenCalledTimes(0);
+      expect(metric?.[0]?.length).toEqual(0);
+
+      expect(mockedModule._tinystan_sample).toHaveBeenCalledTimes(1);
     });
 
     test("failure in model construction throws", async () => {