Hello, Support team.
I’ve already run NVIDIA VSS and tested it with a sample traffic video. But in both the summary and Q&A, it’s not detecting any collisions or accidents. Could you help me figure this out?
Sincerely
Taopik H.
Sorry for the late reply. We will use the video you provided and try that on our side first.
Thank you very much.
I’ve already rented 8 * A100 (80 GB) GPUs and deployed everything using Helm.
When I deployed, I used an overrides.yaml file like this:
nim-llm:
env:
- name: NVIDIA_VISIBLE_DEVICES
value: “0,1,2,3”
resources:
limits:
nvidia.com/gpu: 0 # no limit
vss:
applicationSpecs:
vss-deployment:
containers:
vss:
env:
- name: VLM_MODEL_TO_USE
value: nvila
- name: MODEL_PATH
value: “git:Efficient-Large-Model/NVILA-15B · Hugging Face”
- name: NVIDIA_VISIBLE_DEVICES
value: “4,5”
resources:
limits:
nvidia.com/gpu: 0 # no limit
nemo-embedding:
applicationSpecs:
embedding-deployment:
containers:
embedding-container:
env:
- name: NGC_API_KEY
valueFrom:
secretKeyRef:
key: NGC_API_KEY
name: ngc-api-key-secret
- name: NVIDIA_VISIBLE_DEVICES
value: ‘6’
resources:
limits:
nvidia.com/gpu: 0 # no limit
nemo-rerank:
applicationSpecs:
ranking-deployment:
containers:
ranking-container:
env:
- name: NGC_API_KEY
valueFrom:
secretKeyRef:
key: NGC_API_KEY
name: ngc-api-key-secret
- name: NVIDIA_VISIBLE_DEVICES
value: ‘7’
resources:
limits:
nvidia.com/gpu: 0 # no limit
I used the NVILA-15B model, but when I test it, it doesn’t detect any accidents accurately.
This is a problem. Please help me with this.
Sincerely,
Taopik H.
Could you please provide me with the detailed overrides.yaml file? Also, let me know which GPU(s) you rented.
And according to the documentation, I can only see the alert API for live streaming. I also need alert APIs for both video and picture summarization. Could you assist with this as well?
Sincerely,
Taopik H
override.yaml (1.4 KB)
You can create alerts for any sources.
Thanks for providing detail scripts.
But regarding alert api, what I want to know is when I summarize video or image, how can I get alert result if I use which apis? could you please detail guide for getting alert using api?
I attached alert and summarization apis. But I am not clear how can get alert result.
Please help me.
Sincerely
Taopik H.
Hello,
Please help me with detecting accidents on Q&A.
Sincerely,
Taopik H.
Thanks for responding.
I have deployed highres nvila model.
Summary shows accident but Q&A sometimes doesnt show accidents. It is inconsistent.
Could you help this problem?
And is it possible to deploy VILA1.5-40B model not highres in 8 * A100 GPU?
There are multiple accidents but it doesn’t show.
Accuracy is missing.
Could you tell me the reason and please give me the solution.
Hello,
I’ve deployed the highres nvila model, but it’s not detecting all accidents—only a few are being recognized, and the overall accuracy is lacking. I believe the issue might be related to the model itself.
Could you please advise me on how to improve detection accuracy or how I can get more complete accident detection, on Q&A? I’d really appreciate your help.
Taopik H.
There are two methods that can be tried.
1.You need to finetune your model according to your specific scenario yourself.
2. .You can try more detailed prompts for your scenario.
Thanks for your guidance. Following your advice, I’m going to fine-tune the HighRes model. According to the NVIDIA VSS documentation, I’m trying to find the scripts, APIs, and tutorial book for fine-tuning.
I’ve already downloaded the fine-tuning microservice container using this command:
docker pull nvcr.io/nvidia/tao/vlm-lita-finetuning-ea:0.3.0-ea
However, I don’t see any files related to fine-tuning inside the container.
Could you help me with the fine-tuning process?
Looking forward to your response.
Sincerely
Taopik H.
Could you refer to the fine-tuning microservice notebook first to learn that from the notebook first?
Thanks for responding.
I reviewed (NVILA Finetuning Microservice Getting Started - Early Access | NVIDIA NGC).
I think it is not detail. Could you please give me more detail?
According to that url, it mentioned, NVILA FTMS EA package includes the following:
- Finetuning Microservice Container containing scripts and APIs to finetune an NVILA-15B-Lite High Res LITA model
- Sample tutorial notebook to walk-through the end-to-end finetuning workflow for the NVILA-15B-Lite High Res LITA model
- A pre-trained NVILA-15B-Lite High Res VLM with LITA
how can I get this? Please help with this.
Sincerely
Taopik H.
Hello Support Team,
I have a question about creating an API key in the NVIDIA organization.
According to the documentation, I created an API key, but it has limited access.
Here are doc and my generate personal key interfaces.
On my side, I can’t see all the options for “services included”, like in the documentation.
Could you please let me know the reason?
Sincerely
Taopok H.
Could you file a new topic for the API key issue?