[BUG] Failed to compile tile_cholesky

[arpit15]

Bug Description

I am trying to run an example in mujoco_warp. I am getting error related to tile_choleksy function. I tried to build warp from scratch with libmathdx and still keep getting this error.
Is their a way to debug this further?

Error log

libmathdx cuSOLVER error: 3 on /home/user/projects/warp/warp/native/warp.cu:3272
Module mujoco_warp._src.solver._update_gradient 83bb4cc load on device 'cuda:0' took 343.87 ms
Traceback (most recent call last):
  File "/home/user/projects/warp/warp/codegen.py", line 1020, in build
    adj.eval(adj.tree.body[0])
  File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval
    return emit_node(adj, node)
           ^^^^^^^^^^^^^^^^^^^^
  File "/home/user/projects/warp/warp/codegen.py", line 1674, in emit_FunctionDef
    adj.eval(f)
  File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval
    return emit_node(adj, node)
           ^^^^^^^^^^^^^^^^^^^^
  File "/home/user/projects/warp/warp/codegen.py", line 2566, in emit_Assign
    rhs = adj.eval(node.value)
          ^^^^^^^^^^^^^^^^^^^^
  File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval
    return emit_node(adj, node)
           ^^^^^^^^^^^^^^^^^^^^
  File "/home/user/projects/warp/warp/codegen.py", line 2298, in emit_Call
    out = adj.add_call(func, args, kwargs, type_args, min_outputs=min_outputs)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/projects/warp/warp/codegen.py", line 1409, in add_call
    func_args, template_args, ltoirs, extra_shared_memory = func.lto_dispatch_func(
                                                            ^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/projects/warp/warp/builtins.py", line 6605, in tile_cholesky_generic_lto_dispatch_func
    lto_symbol, lto_code_data = warp.build.build_lto_solver(
                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/user/projects/warp/warp/build.py", line 423, in build_lto_solver
    raise RuntimeError("Failed to compile tile_cholesky")
RuntimeError: Failed to compile tile_cholesky
During handling of the above exception, another exception occurred:
Traceback (most recent call last):

File "/home/user/projects/mujoco_warp/.venv/bin/mjwarp-testspeed", line 8, in 

sys.exit(main())

^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/testspeed.py", line 207, in main

app.run(_main)

File "/home/user/projects/mujoco_warp/.venv/lib/python3.11/site-packages/absl/app.py", line 316, in run

_run_main(main, args)

File "/home/user/projects/mujoco_warp/.venv/lib/python3.11/site-packages/absl/app.py", line 261, in _run_main

sys.exit(main(argv))

^^^^^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/testspeed.py", line 131, in _main

jit_time, run_time, trace, ncon, nefc = mjwarp.benchmark(

^^^^^^^^^^^^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/test_util.py", line 69, in benchmark

fn(m, d)

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/warp_util.py", line 101, in wrapper

return fn(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/forward.py", line 684, in step

forward(m, d)

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/warp_util.py", line 101, in wrapper

return fn(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/forward.py", line 678, in forward

solver.solve(m, d)

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/warp_util.py", line 101, in wrapper

return fn(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/solver.py", line 1002, in solve

_create_context(m, d, grad=True)

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/solver.py", line 69, in _create_context

_update_gradient(m, d)

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/solver.py", line 296, in _update_gradient

wp.launch(_zero_grad_dot, dim=(d.nworld), inputs=[d])

File "/home/user/projects/warp/warp/context.py", line 5723, in launch

module_exec = kernel.module.load(device, block_dim)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/context.py", line 2186, in load

builder = ModuleBuilder(self, builder_options, hasher=self.hashers[active_block_dim])

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/context.py", line 1692, in init

self.build_kernel(kernel)

File "/home/user/projects/warp/warp/context.py", line 1721, in build_kernel

kernel.adj.build(self)

File "/home/user/projects/warp/warp/codegen.py", line 1031, in build

raise e

File "/home/user/projects/warp/warp/codegen.py", line 1020, in build

adj.eval(adj.tree.body[0])

File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval

return emit_node(adj, node)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 1674, in emit_FunctionDef

adj.eval(f)

File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval

return emit_node(adj, node)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 2566, in emit_Assign

rhs = adj.eval(node.value)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval

return emit_node(adj, node)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 2298, in emit_Call

out = adj.add_call(func, args, kwargs, type_args, min_outputs=min_outputs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 1409, in add_call

func_args, template_args, ltoirs, extra_shared_memory = func.lto_dispatch_func(

^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/builtins.py", line 6605, in tile_cholesky_generic_lto_dispatch_func

lto_symbol, lto_code_data = warp.build.build_lto_solver(

^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/build.py", line 423, in build_lto_solver

raise RuntimeError("Failed to compile tile_cholesky")

RuntimeError: Error while parsing function "_cholesky" at /home/user/projects/mujoco_warp/mujoco_warp/_src/solver.py:290:

fact_tile = wp.tile_cholesky(mat_tile)

;Failed to compile tile_cholesky

System Information

OS: Ubuntu 22.04
CUDA: 12.8
Nvidia Driver: 560.35.05
python: 3.11.9
g++ - 11.4.0
Warp - 1.7.0-rc.4 (d7e9b9e641f7b2ad4c4056e9dfb30ad9f4c40c0a)
GPU: RTX 4500 Ada

[shi-eric]

What GPU are you running with? We should be able to get more information if you run with LIBMATHDX_LOG_LEVEL=5 set in your environment.

[arpit15]

I am using RTX 4500 Ada. Is this an env flag or a runtime flag or a compile define?

[shi-eric]

I just do something like LIBMATHDX_LOG_LEVEL=5 <your run command> at runtime.

[arpit15]

Thanks. Shared mem issue as shown below. Maybe will post in mujoco_warp

Error log

[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverCreate] handle=0x7ffc078ab980
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverSetOperatorInt64Array] handle=0 op=CUSOLVER_OPERATOR_SIZE count=2 array=570,570
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverSetOperatorInt64Array] handle=0 op=CUSOLVER_OPERATOR_BLOCK_DIM count=3 array=256,1,1
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverSetOperatorInt64] handle=0 op=CUSOLVER_OPERATOR_TYPE value=0
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverSetOperatorInt64] handle=0 op=CUSOLVER_OPERATOR_API value=0
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverSetOperatorInt64] handle=0 op=CUSOLVER_OPERATOR_FUNCTION value=2
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverSetOperatorInt64] handle=0 op=CUSOLVER_OPERATOR_EXECUTION value=1
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverSetOperatorInt64] handle=0 op=CUSOLVER_OPERATOR_PRECISION value=5
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverSetOperatorInt64] handle=0 op=CUSOLVER_OPERATOR_FILL_MODE value=0
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverSetOperatorInt64] handle=0 op=CUSOLVER_OPERATOR_SM value=890
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverSetOptionStr] handle=0 opt=0 value=potrf_570_570_89_5
[2025-03-27 18:02:37][libmathdx][3010371][Api][cusolverGetLTOIRSize] handle=0 lto_size=0x7ffc078ab988
[2025-03-27 18:02:38][libmathdx][3010371][Info][cusolverGetLTOIRSize] Decompression done in 59.573 ms.
[2025-03-27 18:02:38][libmathdx][3010371][Info][cusolverGetLTOIRSize] NVRTC 12.8 compiling for SM 8.9
[2025-03-27 18:02:38][libmathdx][3010371][Info][cusolverGetLTOIRSize] NVRTC options --relocatable-device-code=true,--device-as-default-execution-space,--std=c++17,-DCUFFTDX_DETAIL_USE_CUDA_STL=1,-dlto,--gpu-architecture=compute_89
[2025-03-27 18:02:38][libmathdx][3010371][Info][cusolverGetLTOIRSize] NVRTC compilation finished in 250.019 ms.
[2025-03-27 18:02:38][libmathdx][3010371][Error][cusolverGetLTOIRSize] NVRTC compilation failed, error message:
cusolverdx/detail/solver_description.hpp(183): error: static assertion failed with "Provided combination of data type and sizes makes this problem not fit into shared memory available on the specified architecture"
              static_assert((not has_sm) or valid_shared_size,
              ^
          detected during instantiation of class "cusolverdx::detail::solver_description [with Operators=, cusolverdx::Precision, cusolverdx::Type<(cusolverdx::type)0>, cusolverdx::Function, cusolverdx::SM<890U>>]" at line 3 of 68765bc8664acee4.cu
cusolverdx/detail/solver_description.hpp(183): error: static assertion failed with "Provided combination of data type and sizes makes this problem not fit into shared memory available on the specified architecture"

static_assert((not has_sm) or valid_shared_size,

^

detected during instantiation of class "cusolverdx::detail::solver_description<Operators...> [with Operators=<cusolverdx::Size<570U, 570U, 1U>, cusolverdx::Precision<float, float, float>, cusolverdx::Type<(cusolverdx::type)0>, cusolverdx::Functioncusolverdx::function::potrf, cusolverdx::SM<890U>, cusolverdx::FillModecusolverdx::fill_mode::upper>]" at line 3 of 68765bc8664acee4.cu
cusolverdx/detail/solver_description.hpp(183): error: static assertion failed with "Provided combination of data type and sizes makes this problem not fit into shared memory available on the specified architecture"

static_assert((not has_sm) or valid_shared_size,

^

detected during instantiation of class "cusolverdx::detail::solver_description<Operators...> [with Operators=<cusolverdx::Size<570U, 570U, 1U>, cusolverdx::Precision<float, float, float>, cusolverdx::Type<(cusolverdx::type)0>, cusolverdx::Functioncusolverdx::function::potrf, cusolverdx::SM<890U>, cusolverdx::FillModecusolverdx::fill_mode::upper, cusolverdx::BlockDim<256U, 1U, 1U>>]" at line 3 of 68765bc8664acee4.cu
cusolverdx/detail/solver_description.hpp(183): error: static assertion failed with "Provided combination of data type and sizes makes this problem not fit into shared memory available on the specified architecture"

static_assert((not has_sm) or valid_shared_size,

^

detected during:

instantiation of class "cusolverdx::detail::solver_description<Operators...> [with Operators=<cusolverdx::Size<570U, 570U, 1U>, cusolverdx::Precision<float, float, float>, cusolverdx::Type<(cusolverdx::type)0>, cusolverdx::Functioncusolverdx::function::potrf, cusolverdx::SM<890U>, cusolverdx::FillModecusolverdx::fill_mode::upper, cusolverdx::BlockDim<256U, 1U, 1U>, cusolverdx::Block>]" at line 29 of cusolverdx/detail/solver_execution.hpp

instantiation of class "cusolverdx::detail::solver_execution<Operators...> [with Operators=<cusolverdx::Size<570U, 570U, 1U>, cusolverdx::Precision<float, float, float>, cusolverdx::Type<(cusolverdx::type)0>, cusolverdx::Functioncusolverdx::function::potrf, cusolverdx::SM<890U>, cusolverdx::FillModecusolverdx::fill_mode::upper, cusolverdx::BlockDim<256U, 1U, 1U>, cusolverdx::Block>]" at line 67 of cusolverdx/detail/solver_execution.hpp

instantiation of class "cusolverdx::detail::block_execution<Operators...> [with Operators=<cusolverdx::Size<570U, 570U, 1U>, cusolverdx::Precision<float, float, float>, cusolverdx::Type<(cusolverdx::type)0>, cusolverdx::Functioncusolverdx::function::potrf, cusolverdx::SM<890U>, cusolverdx::FillModecusolverdx::fill_mode::upper, cusolverdx::BlockDim<256U, 1U, 1U>, cusolverdx::Block>]" at line 3 of 68765bc8664acee4.cu
4 errors detected in the compilation of "68765bc8664acee4.cu".
[2025-03-27 18:02:38][libmathdx][3010371][Error][cusolverGetLTOIRSize] Generated sources passed to NVRTC:

#include <cusolverdx.hpp>

using namespace cusolverdx;

using Solver = decltype(Size<570,570 >() + Precision() + Type<type::real >() + Function() + SM<890 >() + FillMode<fill_mode::upper >() + BlockDim<256,1,1 >() + Block());

extern "C" { device void potrf_570_570_89_5(Solver::a_data_type* smem, unsigned int ld) {

Solver::status_type info;

Solver().execute(smem, ld, &info);

};

}

[2025-03-27 18:02:38][libmathdx][3010371][Error][cusolverGetLTOIRSize] NVRTC error NVRTC_ERROR_COMPILATION at /home/jenkins/agent/workspace/mathdx/helpers/libmathdx/release-0.1.2/L1_Nightly/build/libmathdx/src/nvrtc_helper.cpp:230

[2025-03-27 18:02:38][libmathdx][3010371][Error][cusolverGetLTOIRSize] ERROR at /home/jenkins/agent/workspace/mathdx/helpers/libmathdx/release-0.1.2/L1_Nightly/build/libmathdx/src/libcommondx.cpp:49

[2025-03-27 18:02:38][libmathdx][3010371][Error][cusolverGetLTOIRSize] ERROR at /home/jenkins/agent/workspace/mathdx/helpers/libmathdx/release-0.1.2/L1_Nightly/build/libmathdx/src/libcommondx.cpp:74

libmathdx cuSOLVER error: 3 on /home/user/projects/warp/warp/native/warp.cu:3272

Module mujoco_warp._src.solver._update_gradient 83bb4cc load on device 'cuda:0' took 338.02 ms

Traceback (most recent call last):

File "/home/user/projects/warp/warp/codegen.py", line 1020, in build

adj.eval(adj.tree.body[0])

File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval

return emit_node(adj, node)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 1674, in emit_FunctionDef

adj.eval(f)

File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval

return emit_node(adj, node)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 2566, in emit_Assign

rhs = adj.eval(node.value)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval

return emit_node(adj, node)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 2298, in emit_Call

out = adj.add_call(func, args, kwargs, type_args, min_outputs=min_outputs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 1409, in add_call

func_args, template_args, ltoirs, extra_shared_memory = func.lto_dispatch_func(

^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/builtins.py", line 6605, in tile_cholesky_generic_lto_dispatch_func

lto_symbol, lto_code_data = warp.build.build_lto_solver(

^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/build.py", line 423, in build_lto_solver

raise RuntimeError("Failed to compile tile_cholesky")

RuntimeError: Failed to compile tile_cholesky
During handling of the above exception, another exception occurred:
Traceback (most recent call last):

File "/home/user/projects/mujoco_warp/.venv/bin/mjwarp-testspeed", line 8, in 

sys.exit(main())

^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/testspeed.py", line 207, in main

app.run(_main)

File "/home/user/projects/mujoco_warp/.venv/lib/python3.11/site-packages/absl/app.py", line 316, in run

_run_main(main, args)

File "/home/user/projects/mujoco_warp/.venv/lib/python3.11/site-packages/absl/app.py", line 261, in _run_main

sys.exit(main(argv))

^^^^^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/testspeed.py", line 131, in _main

jit_time, run_time, trace, ncon, nefc = mjwarp.benchmark(

^^^^^^^^^^^^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/test_util.py", line 69, in benchmark

fn(m, d)

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/warp_util.py", line 101, in wrapper

return fn(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/forward.py", line 684, in step

forward(m, d)

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/warp_util.py", line 101, in wrapper

return fn(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/forward.py", line 678, in forward

solver.solve(m, d)

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/warp_util.py", line 101, in wrapper

return fn(*args, **kwargs)

^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/solver.py", line 1002, in solve

_create_context(m, d, grad=True)

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/solver.py", line 69, in _create_context

_update_gradient(m, d)

File "/home/user/projects/mujoco_warp/mujoco_warp/_src/solver.py", line 296, in _update_gradient

wp.launch(_zero_grad_dot, dim=(d.nworld), inputs=[d])

File "/home/user/projects/warp/warp/context.py", line 5723, in launch

module_exec = kernel.module.load(device, block_dim)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/context.py", line 2186, in load

builder = ModuleBuilder(self, builder_options, hasher=self.hashers[active_block_dim])

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/context.py", line 1692, in init

self.build_kernel(kernel)

File "/home/user/projects/warp/warp/context.py", line 1721, in build_kernel

kernel.adj.build(self)

File "/home/user/projects/warp/warp/codegen.py", line 1031, in build

raise e

File "/home/user/projects/warp/warp/codegen.py", line 1020, in build

adj.eval(adj.tree.body[0])

File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval

return emit_node(adj, node)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 1674, in emit_FunctionDef

adj.eval(f)

File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval

return emit_node(adj, node)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 2566, in emit_Assign

rhs = adj.eval(node.value)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 2777, in eval

return emit_node(adj, node)

^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 2298, in emit_Call

out = adj.add_call(func, args, kwargs, type_args, min_outputs=min_outputs)

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/codegen.py", line 1409, in add_call

func_args, template_args, ltoirs, extra_shared_memory = func.lto_dispatch_func(

^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/builtins.py", line 6605, in tile_cholesky_generic_lto_dispatch_func

lto_symbol, lto_code_data = warp.build.build_lto_solver(

^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "/home/user/projects/warp/warp/build.py", line 423, in build_lto_solver

raise RuntimeError("Failed to compile tile_cholesky")

RuntimeError: Error while parsing function "_cholesky" at /home/user/projects/mujoco_warp/mujoco_warp/_src/solver.py:290:

fact_tile = wp.tile_cholesky(mat_tile)

;Failed to compile tile_cholesky

[arpit15]

I am trying a scene which has only 98 boxes. My VRAM 24GB. I was expecting this to work on my GPU. Can it be a GPU issue or a parameter which I can set in warp?

[shi-eric]

Thanks for the information, will ask around next week, but please raise this issue in https://github.com/google-deepmind/mujoco_warp so the team can track it there. Were you running an example straight from the repo, or did you modify it? It might help to have the command you're running.

[arpit15]

I am running mjwarp-testspeed with my test scene which only has a bunch of boxes.

[shi-eric]

Got it. Shared memory is a limited resource, so there will be a limit on the tile sizes that can be used depending on what's going on inside the kernel. The way it's being determined in mujoco_warp might be pretty naive right now and without bound (can't tell exactly how much from a quick look at your logs).

[arpit15]

I reduced the number of boxes to 18 and things seem to work fine. I guess I will read the code to see if I can fix something to make it work.

[arpit15]

I have posted the issue in (mujoco_warp repo)[https://github.com/google-deepmind/mujoco_warp/issues/47]. Please let me know if there is a quick hack to this issue.

[shi-eric]

I think we determined that this isn't a bug with the Warp library, so I'm closing the issue. I think we can improve the error reporting, so I opened #608 to track this work.