Janusz Lisiecki

Hi @JosselinSomervilleRoberts, Thank you for reaching out. I'm afraid DALI doesn't support the sampling patterns you ask for. What it can do is sample video with constant steps and stride,...

Hi @qihang720, Can you provide more details about the environment you are using to run your tests? Can you reproduce a similar number using DALI as a standalone library? Just...

Hi @qihang720, `nvcr.io/nvidia/tritonserver:23.05-py3` uses DALI 1.25. In DALI 1.30 we did a couple of optimizations for the crop_mirror_normalize operator. Please stay tuned for TRITON 23.10 which should include this DALI...

Hi @qihang720, > For batch of data, if my every input is different, the base64 length is also different, so how can I batch it together. Please check if [this](https://github.com/triton-inference-server/dali_backend/blob/main/docs/tutorials/training_to_inference.md#batching-data-in-the-request)...

Hi @zmasih, Can you tell me how do you run the example and what kind of error do you observe? I cannot rule out that the code itself is not...

Hi @zmasih, Indeed, the code is an example that was not prepared for more than one node in mind for the CPU. In your case, if you want to run...

> So, you are saying that if more than one device is available, with no explicit request, when running the code, DALI will use multiple CPU cores to load, decode,...

Hi @zmasih, As I mentioned, the example is not prepared to run without the GPU even if the pipeline can run on the CPU. In this case, each DALI pipeline...

Hi @zmasih, You can start with this toy example: ``` import os.path test_data_root = os.environ["DALI_EXTRA_PATH"] # MXNet RecordIO base = os.path.join(test_data_root, "db", "recordio") idx_files = [base + "/train.idx"] rec_files =...

Can you please also add this to the documentation?