Grok for DcpOMatic CLI

[kristophm]

Hello,

I have obtained a GROK license for GPU encoding with DoM. My question is, if I'm using the CLI interface, where do I input the license information, and do I need to specify a flag to explicitly use grok?

Lastly, I've downloaded the package but it's unclear where I should be putting this such that DoM can find it and use it.

[boxerab]

Hi @kristophm,

Is this is a headless instance ? For desktop system there is a GPU tab in preferences
where you can configure the add-on.

Cheers,
Aaron

[kristophm]

Hi @boxerab,

If by headless you mean CLI only (no desktop GUI), yes correct.

So I run using https://dcpomatic.com/manual/html/ch15.html#idp34 and https://dcpomatic.com/manual/html/ch15s02.html. Those docs don't seem to have a flag to indicate GPU use or a Grok license key.

Please let me know if you need more info.

[boxerab]

Thanks, we are working on a fix for this.
Aaron

[kristophm]

Thanks Aaron and team! I'll look forward to the update!

[carl]

Hi @kristophm,

This build of DCP-o-matic adds some features to the command line client (dcpomatic2_cli):

Code: Select all

dcpomatic2_cli config grok-licence foo
dcpomatic2_cli list-gpus
dcpomatic2_cli config grok-gpu-index 1
dcpomatic2_cli config grok-binary-location /home/bob/grok_install

It would be really helpful if you could try it out, and see if it works for you... let us know here how you get on, if you do try it!

[kristophm]

Hi Carl and team, thanks!

It was able to successfully render the DCP but unfortunately it doesn't look like it used the GPU.

I ran the following command during the rendering and saw "No running processes found":
```
root@localhost:~# nvidia-smi
Wed Mar 5 14:32:27 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.124.06 Driver Version: 570.124.06 CUDA Version: 12.8 |
|-----------------------------------------+------------------------+----------------------+
| 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 Quadro RTX 6000 Off | 00000000:00:02.0 Off | Off |
| 34% 32C P8 8W / 260W | 1MiB / 24576MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| No running processes found |
+-----------------------------------------------------------------------------------------+
```

I followed the steps nearly exactly as you prescribed, but because the output of `dcpomatic2_cli list-gpus` was "0: Device 0: Quadro RTX 6000", I set grok-gpu-index to "0" instead of "1" assuming it should match the output of list-gpus. Note, I set it to "1" also just in case, and noticed the same result.


Hardware/OS specs:
Ubuntu 24.04 LTS | 1x GPU (RTX 6000) | 8 CPU cores | 32 GB RAM | Nvidia Driver version: 570.124.06

Please let me know if there's anything else I can try! I'm motivated to get this working . Thanks in advance

[carl]

Ah sorry I think I missed a config step

Code: Select all

dcpomatic2_cli config grok-enable 1

[kristophm]

Thanks! Ok this time I can see that it's using the GPU, but I get an error "synch failed on file" and it just hangs at 9%.

```
root@localhost:~# dcpomatic2_cli xavier
Making DCP for CocoFilm
[2025-03-06 03:52:33.607] [warning] Cinema 4K profile activated. Other options specified may be overridden
[2025-03-06 03:52:33.612] [info] Plugin loaded
[2025-03-06 03:52:33.613] [info] Connecting to server https://grokcompression.com/api/register
[2025-03-06 03:52:33.876] [info] Received response: 200
GPU Device 0: "Turing" with compute capability 7.5
[2025-03-06 03:52:34.228] [info] Initializing shared memory buffers: num frames 16, uncompressed frame size 51788160, compressed frame size 51788160
[2025-03-06 03:52:34.228] [info] Initialized shared uncompressed memory buffers: success
[2025-03-06 03:52:34.228] [info] Initialized shared compressed memory buffers: success
[2025-03-06 03:52:34.229] [info] The desired maximum codestream size has limited at least one of the desired quality layers.
[2025-03-06 03:52:34.674] [info] Device memory allocated: 8426 MB
[2025-03-06 03:52:34.674] [info] Initializing encoders
[2025-03-06 03:52:35.097] [info] Sending shared memory batch compress init message
[2025-03-06 03:52:36.072] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:36.207] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:36.271] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:36.385] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:36.509] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:36.672] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:36.768] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:36.834] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:36.912] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:37.019] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:37.111] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:37.268] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:37.357] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:37.421] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:37.501] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
[2025-03-06 03:52:37.647] [info] Populate for compressed input channel 0 : numPasses reached max of 67

synch failed on file
Transcoding CocoFilm; Encoding: 9%; 139/1440 frames
```

The nvidia-smi output is as follows. It shows that it's definitely using the GPU now.
```
Thu Mar 6 03:54:46 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.124.06 Driver Version: 570.124.06 CUDA Version: 12.8 |
|-----------------------------------------+------------------------+----------------------+
| 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 Quadro RTX 6000 Off | 00000000:00:02.0 Off | Off |
| 33% 44C P2 67W / 260W | 3464MiB / 24576MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 8266 C ./grk_compress 3460MiB |
+-----------------------------------------------------------------------------------------+
```

[carl]

@boxerab do you have an idea about this?