Hi,
I am having problems sampling use the most simplest model:
x <- normal(0,1)
m <- model(x)
mcmc(m)
My environment/installation:
jupyther data science notebook running in docker
installed greta according to the getting started tutorial.
Getting this error message:
Error in py_call_impl(callable, dots$args, dots$keywords): RuntimeError: Evaluation error: ValueError: Tensor(“leapfrog_integrate_one_step/maybe_call_fn_and_grads/value_and_gradients/strided_slice:0”, shape=(), dtype=int32) must be from the same graph as Tensor(“mcmc_sample_chain/trace_scan/while/Const:0”, shape=(1,), dtype=float64).
Detailed traceback:
File “/home/jovyan/.local/lib/python3.7/site-packages/tensorflow/python/ops/gen_array_ops.py”, line 11055, in tile
“Tile”, input=input, multiples=multiples, name=name)
File “/home/jovyan/.local/lib/python3.7/site-packages/tensorflow/python/framework/op_def_library.py”, line 309, in _apply_op_helper
g = ops._get_graph_from_inputs(_Flatten(keywords.values()))
File “/home/jovyan/.local/lib/python3.7/site-packages/tensorflow/python/framework/ops.py”, line 5921, in _get_graph_from_inputs
_assert_same_graph(original_graph_element, graph_element)
File “/home/jovyan/.local/lib/python3.7/site-packages/tensorflow/python/framework/ops.py”, line 5856, in _assert_same_graph
(item, original_item))
Traceback:
- mcmc(m)
- lapply(initial_values_split, build_sampler, sampler, model)
- FUN(X[[i]], …)
- sampler$class$new(initial_values, model, sampler$parameters,
. seed = seed) - .subset2(public_bind_env, “initialize”)(…)
- self$define_tf_draws()
- dag$tf_run(sampler_batch <- tfp$mcmc$sample_chain(num_results = tf$math$floordiv(sampler_burst_length,
. sampler_thin), current_state = free_state, kernel = sampler_kernel,
. trace_fn = function(current_state, kernel_results) {
. kernel_results
. }, num_burnin_steps = tf$constant(0L, dtype = tf$int32),
. num_steps_between_results = sampler_thin, parallel_iterations = 1L)) - self$on_graph(with(tfe, eval(expr)))
- with(self$tf_graph$as_default(), expr)
- with.python.builtin.object(self$tf_graph$as_default(), expr)
- tryCatch(force(expr), finally = {
. data$__exit__(NULL, NULL, NULL)
. }) - tryCatchList(expr, classes, parentenv, handlers)
- force(expr)
- with(tfe, eval(expr))
- with.default(tfe, eval(expr))
- eval(substitute(expr), data, enclos = parent.frame())
- eval(substitute(expr), data, enclos = parent.frame())
- eval(expr)
- eval(expr)
- tfp$mcmc$sample_chain(num_results = tf$math$floordiv(sampler_burst_length,
. sampler_thin), current_state = free_state, kernel = sampler_kernel,
. trace_fn = function(current_state, kernel_results) {
. kernel_results
. }, num_burnin_steps = tf$constant(0L, dtype = tf$int32),
. num_steps_between_results = sampler_thin, parallel_iterations = 1L) - py_call_impl(callable, dots$args, dots$keywords)