Can't recursively update tensorrt_llm submodule 3rdparty directory links

[BenjaminBraunDev]

When cloning the main branch, and running:

git lfs install
git submodule update --init --recursive

The 3rdparty directory in tensorrt_llm submodule has links that cannot be resolved. I tried manually following the links in the repo and could not resolve them either.

I believe the cause of this issue is that the latest PR #715 (10 hours ago as of writing) changed the tensorrt_llm submodule version to one with a 3rdparty dierctory that references an internal or private GitLab instance within NVIDIA's network, judging from the error saying Could not resolve host: gitlab-master.nvidia.com. These are the 3rdparty links in the in tensorrt_llm submodule of the current tensorrtllm_backend main branch, which I can't resolve in my browser either. Whereas I can resolve the tensorrt_llm main branch 3rdparty submodules just fine (the tensorrtllm_backend:main's tensorrt_llm submodule is not tensorrt_llm:main).

As a workaround I will clone the v0.17.0 release branch, which has a tensorrt_llm version with 3rdparty submodules I can resolve.

Error log:

│ llama-build-job-b7ckg Cloning into '/models/tensorrtllm_backend'...                                                                                                                                                                 │
│ llama-build-job-b7ckg Updated Git hooks.                                                                                                                                                                                            │
│ llama-build-job-b7ckg Git LFS initialized.                                                                                                                                                                                          │
│ llama-build-job-b7ckg Submodule 'tensorrt_llm' (https://github.com/NVIDIA/TensorRT-LLM.git) registered for path 'tensorrt_llm'                                                                                                      │
│ llama-build-job-b7ckg Cloning into '/models/tensorrtllm_backend/tensorrt_llm'...                                                                                                                                                    │
│ llama-build-job-b7ckg Submodule path 'tensorrt_llm': checked out '77d7fe1eb2d89753ae33a0745d1083c133064f92'                                                                                                                         │
│ llama-build-job-b7ckg Submodule '3rdparty/NVTX' (https://gitlab-master.nvidia.com/ftp/GitHubSync/NVTX-mirror.git) registered for path 'tensorrt_llm/3rdparty/NVTX'                                                                  │
│ llama-build-job-b7ckg Submodule '3rdparty/cutlass' (https://gitlab-master.nvidia.com/ftp/GitHubSync/cutlass-mirror.git) registered for path 'tensorrt_llm/3rdparty/cutlass'                                                         │
│ llama-build-job-b7ckg Submodule '3rdparty/cxxopts' (https://gitlab-master.nvidia.com/ftp/GitHubSync/cxxopts-mirror.git) registered for path 'tensorrt_llm/3rdparty/cxxopts'                                                         │
│ llama-build-job-b7ckg Submodule '3rdparty/json' (https://gitlab-master.nvidia.com/ftp/GitHubSync/json-mirror.git) registered for path 'tensorrt_llm/3rdparty/json'                                                                  │
│ llama-build-job-b7ckg Submodule '3rdparty/pybind11' (https://gitlab-master.nvidia.com/ftp/GitHubSync/pybind11.git) registered for path 'tensorrt_llm/3rdparty/pybind11'                                                             │
│ llama-build-job-b7ckg Submodule '3rdparty/ucxx' (https://gitlab-master.nvidia.com/ftp/GitHubSync/ucxx.git) registered for path 'tensorrt_llm/3rdparty/ucxx'                                                                         │
│ llama-build-job-b7ckg Submodule '3rdparty/xgrammar' (https://gitlab-master.nvidia.com/ftp/GitHubSync/xgrammar.git) registered for path 'tensorrt_llm/3rdparty/xgrammar'                                                             │
│ llama-build-job-b7ckg Cloning into '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/NVTX'...                                                                                                                                      │
│ llama-build-job-b7ckg fatal: unable to access 'https://gitlab-master.nvidia.com/ftp/GitHubSync/NVTX-mirror.git/': Could not resolve host: gitlab-master.nvidia.com                                                                  │
│ llama-build-job-b7ckg fatal: clone of 'https://gitlab-master.nvidia.com/ftp/GitHubSync/NVTX-mirror.git' into submodule path '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/NVTX' failed                                         │
│ llama-build-job-b7ckg Failed to clone '3rdparty/NVTX'. Retry scheduled                                                                                                                                                              │
│ llama-build-job-b7ckg Cloning into '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/cutlass'...                                                                                                                                   │
│ llama-build-job-b7ckg fatal: unable to access 'https://gitlab-master.nvidia.com/ftp/GitHubSync/cutlass-mirror.git/': Could not resolve host: gitlab-master.nvidia.com                                                               │
│ llama-build-job-b7ckg fatal: clone of 'https://gitlab-master.nvidia.com/ftp/GitHubSync/cutlass-mirror.git' into submodule path '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/cutlass' failed                                   │
│ llama-build-job-b7ckg Failed to clone '3rdparty/cutlass'. Retry scheduled                                                                                                                                                           │
│ llama-build-job-b7ckg Cloning into '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/cxxopts'...                                                                                                                                   │
│ llama-build-job-b7ckg fatal: unable to access 'https://gitlab-master.nvidia.com/ftp/GitHubSync/cxxopts-mirror.git/': Could not resolve host: gitlab-master.nvidia.com                                                               │
│ llama-build-job-b7ckg fatal: clone of 'https://gitlab-master.nvidia.com/ftp/GitHubSync/cxxopts-mirror.git' into submodule path '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/cxxopts' failed                                   │
│ llama-build-job-b7ckg Failed to clone '3rdparty/cxxopts'. Retry scheduled                                                                                                                                                           │
│ llama-build-job-b7ckg Cloning into '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/json'...                                                                                                                                      │
│ llama-build-job-b7ckg fatal: unable to access 'https://gitlab-master.nvidia.com/ftp/GitHubSync/json-mirror.git/': Could not resolve host: gitlab-master.nvidia.com                                                                  │
│ llama-build-job-b7ckg fatal: clone of 'https://gitlab-master.nvidia.com/ftp/GitHubSync/json-mirror.git' into submodule path '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/json' failed                                         │
│ llama-build-job-b7ckg Failed to clone '3rdparty/json'. Retry scheduled                                                                                                                                                              │
│ llama-build-job-b7ckg Cloning into '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/pybind11'...                                                                                                                                  │
│ llama-build-job-b7ckg fatal: unable to access 'https://gitlab-master.nvidia.com/ftp/GitHubSync/pybind11.git/': Could not resolve host: gitlab-master.nvidia.com                                                                     │
│ llama-build-job-b7ckg fatal: clone of 'https://gitlab-master.nvidia.com/ftp/GitHubSync/pybind11.git' into submodule path '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/pybind11' failed                                        │
│ llama-build-job-b7ckg Failed to clone '3rdparty/pybind11'. Retry scheduled                                                                                                                                                          │
│ llama-build-job-b7ckg Cloning into '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/ucxx'...                                                                                                                                      │
│ llama-build-job-b7ckg fatal: unable to access 'https://gitlab-master.nvidia.com/ftp/GitHubSync/ucxx.git/': Could not resolve host: gitlab-master.nvidia.com                                                                         │
│ llama-build-job-b7ckg fatal: clone of 'https://gitlab-master.nvidia.com/ftp/GitHubSync/ucxx.git' into submodule path '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/ucxx' failed                                                │
│ llama-build-job-b7ckg Failed to clone '3rdparty/ucxx'. Retry scheduled                                                                                                                                                              │
│ llama-build-job-b7ckg Cloning into '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/xgrammar'...                                                                                                                                  │
│ llama-build-job-b7ckg fatal: unable to access 'https://gitlab-master.nvidia.com/ftp/GitHubSync/xgrammar.git/': Could not resolve host: gitlab-master.nvidia.com                                                                     │
│ llama-build-job-b7ckg fatal: clone of 'https://gitlab-master.nvidia.com/ftp/GitHubSync/xgrammar.git' into submodule path '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/xgrammar' failed                                        │
│ llama-build-job-b7ckg Failed to clone '3rdparty/xgrammar'. Retry scheduled                                                                                                                                                          │
│ llama-build-job-b7ckg Cloning into '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/NVTX'...                                                                                                                                      │
│ llama-build-job-b7ckg fatal: unable to access 'https://gitlab-master.nvidia.com/ftp/GitHubSync/NVTX-mirror.git/': Could not resolve host: gitlab-master.nvidia.com                                                                  │
│ llama-build-job-b7ckg fatal: clone of 'https://gitlab-master.nvidia.com/ftp/GitHubSync/NVTX-mirror.git' into submodule path '/models/tensorrtllm_backend/tensorrt_llm/3rdparty/NVTX' failed                                         │
│ llama-build-job-b7ckg Failed to clone '3rdparty/NVTX' a second time, aborting                                                                                                                                                       │
│ llama-build-job-b7ckg fatal: Failed to recurse into submodule path 'tensorrt_llm'

[BenjaminBraunDev]

@kaiyux Could you take a look at this to confirm?

[remusao]

+1, encountering the same issue when trying to build the docker image.

[kaiyux]

Sorry for the late reply, is the issue persist on the latest main branch?

[remusao]

I believe the issue has been resolved on main (I was able to build TensorRT-LLM + tensorrtllm_backend from source a few weeks ago after this issue was posted). Best if @BenjaminBraunDev can also confirm.

[BenjaminBraunDev]

Yeah it's resolved in main now.