Errors during running example script.

[Grigorij-Dudnik]

Hey,

I want to start experimenting with TinyNav, but got that errors when tried to run example script (bash scripts/docker_run.sh):

[junlinp]

The precompiled model file is not compatible with your device. You can open the development environment in Visual Studio Code using the Dev Containers extension, then rebuild the models with the scripts in tinynav/models.

[junlinp]

Could you please provide the following information to help us address the bugs:

1. nvidia-smi
2. ls -l in the models directory
3. the hash id for the tinynav in docker images commands
   Here is our Discord — feel free to continue the discussion there.

[ibvandersluis]

@Grigorij-Dudnik try running xhost +local:docker to address the xcb plugin issue in your bottom right terminal.

@junlinp I also get the 'NoneType' object has no attribute 'create_execution_context' error

1. nvidia-smi

+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.247.01             Driver Version: 535.247.01   CUDA Version: 12.2     |
|-----------------------------------------+----------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |         Memory-Usage | GPU-Util  Compute M. |
|                                         |                      |               MIG M. |
|=========================================+======================+======================|
|   0  NVIDIA GeForce RTX 4070 ...    Off | 00000000:01:00.0 Off |                  N/A |
| N/A   39C    P4              N/A /  35W |      7MiB /  8188MiB |      0%      Default |
|                                         |                      |                  N/A |
+-----------------------------------------+----------------------+----------------------+
                                                                                         
+---------------------------------------------------------------------------------------+
| Processes:                                                                            |
|  GPU   GI   CI        PID   Type   Process name                            GPU Memory |
|        ID   ID                                                             Usage      |
|=======================================================================================|
|    0   N/A  N/A      4129      G   /usr/lib/xorg/Xorg                            4MiB |
+---------------------------------------------------------------------------------------+

2. ls -l tinynav/models

total 678372
-rw-rw-r-- 1 isaac isaac       271 Sep 15 12:43 convert_dinov2.sh
-rw-rw-r-- 1 isaac isaac       887 Sep 15 12:43 convert_lightglue.sh
-rw-rw-r-- 1 isaac isaac       412 Sep 15 12:43 convert_retinify.sh
-rw-rw-r-- 1 isaac isaac       431 Sep 15 12:43 convert_superpoint.sh
-rw-rw-r-- 1 isaac isaac 175607212 Sep 15 12:43 dinov2_base_224x224_fp16_aarch64.plan
-rw-rw-r-- 1 isaac isaac 175551811 Sep 15 12:43 dinov2_base_224x224_fp16.onnx
-rw-rw-r-- 1 isaac isaac 175857932 Sep 15 12:43 dinov2_base_224x224_fp16_x86_64.plan
-rw-rw-r-- 1 isaac isaac  45874708 Sep 15 12:43 lightglue_fp16_aarch64.plan
-rwxrwxr-x 1 isaac isaac  45521576 Sep 15 12:43 lightglue_fp16.onnx
-rw-rw-r-- 1 isaac isaac  43878812 Sep 15 12:43 lightglue_fp16_x86_64.plan
-rw-rw-r-- 1 isaac isaac  12468316 Sep 15 12:43 retinify_0_1_4_480x848_aarch64.plan
-rw-rw-r-- 1 isaac isaac  10446719 Sep 15 12:43 retinify_0_1_4_480x848.onnx
-rw-rw-r-- 1 isaac isaac  14895252 Sep 15 12:43 retinify_0_1_4_480x848_x86_64.plan
-rw-rw-r-- 1 isaac isaac   3214764 Sep 15 12:43 superpoint_240x424_fp16_aarch64.plan
-rw-rw-r-- 1 isaac isaac   5222595 Sep 15 12:43 superpoint_240x424_fp16.onnx
-rw-rw-r-- 1 isaac isaac   4635604 Sep 15 12:43 superpoint_240x424_fp16_x86_64.plan

3. docker images | grep tinynav

uniflexai/tinynav        0c4332e       a328d2e051c8   5 weeks ago     26.1GB

[junlinp]

@Grigorij-Dudnik try running xhost +local:docker to address the xcb plugin issue in your bottom right terminal.

@junlinp I also get the 'NoneType' object has no attribute 'create_execution_context' error

1. `nvidia-smi`

+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.247.01             Driver Version: 535.247.01   CUDA Version: 12.2     |
|-----------------------------------------+----------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |         Memory-Usage | GPU-Util  Compute M. |
|                                         |                      |               MIG M. |
|=========================================+======================+======================|
|   0  NVIDIA GeForce RTX 4070 ...    Off | 00000000:01:00.0 Off |                  N/A |
| N/A   39C    P4              N/A /  35W |      7MiB /  8188MiB |      0%      Default |
|                                         |                      |                  N/A |
+-----------------------------------------+----------------------+----------------------+
                                                                                         
+---------------------------------------------------------------------------------------+
| Processes:                                                                            |
|  GPU   GI   CI        PID   Type   Process name                            GPU Memory |
|        ID   ID                                                             Usage      |
|=======================================================================================|
|    0   N/A  N/A      4129      G   /usr/lib/xorg/Xorg                            4MiB |
+---------------------------------------------------------------------------------------+

2. ls -l tinynav/models

total 678372
-rw-rw-r-- 1 isaac isaac       271 Sep 15 12:43 convert_dinov2.sh
-rw-rw-r-- 1 isaac isaac       887 Sep 15 12:43 convert_lightglue.sh
-rw-rw-r-- 1 isaac isaac       412 Sep 15 12:43 convert_retinify.sh
-rw-rw-r-- 1 isaac isaac       431 Sep 15 12:43 convert_superpoint.sh
-rw-rw-r-- 1 isaac isaac 175607212 Sep 15 12:43 dinov2_base_224x224_fp16_aarch64.plan
-rw-rw-r-- 1 isaac isaac 175551811 Sep 15 12:43 dinov2_base_224x224_fp16.onnx
-rw-rw-r-- 1 isaac isaac 175857932 Sep 15 12:43 dinov2_base_224x224_fp16_x86_64.plan
-rw-rw-r-- 1 isaac isaac  45874708 Sep 15 12:43 lightglue_fp16_aarch64.plan
-rwxrwxr-x 1 isaac isaac  45521576 Sep 15 12:43 lightglue_fp16.onnx
-rw-rw-r-- 1 isaac isaac  43878812 Sep 15 12:43 lightglue_fp16_x86_64.plan
-rw-rw-r-- 1 isaac isaac  12468316 Sep 15 12:43 retinify_0_1_4_480x848_aarch64.plan
-rw-rw-r-- 1 isaac isaac  10446719 Sep 15 12:43 retinify_0_1_4_480x848.onnx
-rw-rw-r-- 1 isaac isaac  14895252 Sep 15 12:43 retinify_0_1_4_480x848_x86_64.plan
-rw-rw-r-- 1 isaac isaac   3214764 Sep 15 12:43 superpoint_240x424_fp16_aarch64.plan
-rw-rw-r-- 1 isaac isaac   5222595 Sep 15 12:43 superpoint_240x424_fp16.onnx
-rw-rw-r-- 1 isaac isaac   4635604 Sep 15 12:43 superpoint_240x424_fp16_x86_64.plan

3. docker images | grep tinynav

uniflexai/tinynav        0c4332e       a328d2e051c8   5 weeks ago     26.1GB

The plan for x86_64 was compiled with Compute Capability 8.6. You can recompile the plan for your device using the scripts in tinynav/models.

[Grigorij-Dudnik]

Hey @junlinp @ibvandersluis ! Sorry for ignoring an issue for some time, now I'm returned to it. I managed to overcome the error - used xhost +local:docker (howewer not sure if it had impact) and with bash scripts/check_env.sh found a root of the problem - my user wasn't in the "docker" group. After adding user to group it started to work.

I mean, window is opened, but it's empty:

[Grigorij-Dudnik]

Hey @ibvandersluis @junlinp, any ideas what could be wrong?

[junlinp]

Hey @ibvandersluis @junlinp, any ideas what could be wrong?

@Grigorij-Dudnik
Are you using the container built from our Docker image?
It looks like you’re running on the host machine instead, which doesn’t include the dependencies required to build Ceres.

[Grigorij-Dudnik]

@junlinp Yes, you right! I missed that documentation line about opening it in devcontainer.

Now I did so, and having such issue now:

[junlinp]

@junlinp Yes, you right! I missed that documentation line about opening it in devcontainer.

Now I did so, and having such issue now:

@Grigorij-Dudnik

1. There are no prebuilt plan files for the Retinify model.
   We generate the plan file from the corresponding ONNX file, since many users have encountered issues where the prebuilt ONNX model isn’t compatible with their NVIDIA GPUs.
   You can build the plan file yourself by running make in the /tinynav/tinynav/core/models directory. The Makefile for this process is located in /tinynav/tinynav/core/models/Makefile.

2. It looks like Hugging Face didn’t download the dataset completely.

3. You can try running xhost + to allow all connections from the host machine, and then check inside the container to make sure the DISPLAY environment variable matches the one on the host.

[Grigorij-Dudnik]

Thanks @junlinp, I'll try.

[Grigorij-Dudnik]

@junlinp
1 - done.
2 - not sure how to do it. Please check my logs if it still needed.
3 - done, for both inside and outside of container variable value is ":0".
I have some progress, but it looks not all problems solved. Please check out the screens:

[junlinp]

@junlinp 1 - done. 2 - not sure how to do it. Please check my logs if it still needed. 3 - done, for both inside and outside of container variable value is ":0". I have some progress, but it looks not all problems solved. Please check out the screens:

@Grigorij-Dudnik You can replace huggingface-cli with hf in the scripts, and everything should work correctly.