Solved

Dear team, Is it possible to use hardware acceleration provided by the chip for encoding into x264 or x265 video stream on gstreamer. I was able to find plugins here but no specific mention for hardware acceleration.Thank you.

I get this error when I try to deploy Real-ESRGAN. Can anyone tell me why?./deploy.py /home/aetina/Desktop/voyager-sdk-release-v1.4/ax_models/zoo/torch/real-esrgan-x4plus-onnx.yaml --num-cal-images 50INFO    : Detected Metis type as m2INFO    : Compiling network real-esrgan-x4plus-onnx /home/aetina/Desktop/voyager-sdk-release-v1.4/ax_models/zoo/torch/real-esrgan-x4plus-onnx.yamlINFO    : Compile model: real-esrgan-x4plus-onnxINFO    : Using representative images from /home/aetina/.cache/axelera/data/coco2017_repr400 with backend ImageReader.PIL,pipeline input color format ColorFormat.RGBINFO    : Prequantizing real-esrgan-x4plus-onnx: real-esrgan-x4plus-onnxINFO    : Quantizing network real-esrgan-x4plus-onnx /home/aetina/Desktop/voyager-sdk-release-v1.4/ax_models/zoo/torch/real-esrgan-x4plus-onnx.yaml real-esrgan-x4plus-onnxINFO    : Compile model: real-esrgan-x4plus-onnxINFO    : Using representative images from /home/aetina/.cache/axelera/data/coco2017_repr400 with backend ImageRead

Does anyone have any suggestions on how to implement SAHI inference with Yolo via voyager-sdk to run on metis m2?

Our development team requires specialized expertise for Axelera AI platform implementation and edge AI acceleration optimization. We're seeking an experienced consultant with 5-7weeks availability to support us with

Hi, I'm working on an RPI5 and I have the voyager SDK inside a container with ubuntu 22 and it works. I wanted to try running an inference on a stream using GStreamer. First, I tried to access the stream and I managed with the following command:GST_PLUGIN_PATH=/voyager-sdk/operators/lib AXELERA_DEVICE_DIR=../opt/axelera/device-1.2.5-1/omega LD_LIBRARY_PATH=/voyager-sdk/operators/lib:/opt/axelera/runtime-1.2.5-1/lib:$LD_LIBRARY_PATH gst-launch-1.0 rtspsrc location='rtsp://…..' latency=200 ! rtph264depay ! avdec_h264 ! videoconvert ! autovideosink Now I wanted to try the axinferencenet plugin but I'm having problems. I've tested various commands and formats, such as: GST_PLUGIN_PATH=/voyager-sdk/operators/lib AXELERA_DEVICE_DIR=../opt/axelera/device-1.2.5-1/omega LD_LIBRARY_PATH=/voyager-sdk/operators/lib:/opt/axelera/runtime-1.2.5-1/lib:$LD_LIBRARY_PATH gst-launch-1.0 rtspsrc location='rtsp://...' latency=200 ! rtph264depay ! avdec_h264 ! videoconvert ! video/x-raw,format=BGRA !  axtran

Hello everyone,I'm new to the Axelera AI system and am currently working with a Metis PCIe card with 4 GB memory.I want to understand how much of this memory is being used by the neural network model I am deploying. In a way, I'm looking for a tool or command that provides a report similar to what nvidia-smi does for NVIDIA GPUs, but for the Metis core.From my research, I understand that the Metis core's DRAM is a fixed, dedicated resource and that the memory usage is determined at the time of compilation, not at runtime. I've heard that the Voyager SDK's compiler logs might contain this information.My main question is: Where in the compiler output or log files can I find the exact memory footprint and percentage utilization for my model? Is there a specific argument or command I need to use to generate this report?Any guidance on this would be greatly appreciated.Thank you!

Hello,I am having a hard time finding the source code of the metis-dkms.The one git repo I found as vcs.url does not exist publicly (https://github.com/axelera-ai/software-platform) I would love to have the sources because I need a rpm instead of deb-package. Cheers Tim

Hi,I tried the Yolo11-coco-onnx models from Model Zoo and got terrible results. Can anyone give me some advice? ./inference.py yolo11n-coco-onnx dataset --no-display --show-device-fpsINFO    : Model:      yolo11n-coco-onnx                                         INFO    : Dataset:    CocoDataset-COCO2017INFO    : Date:       2025-08-21 12:33:18.028455INFO    : Inference Time: 67178.89msINFO    : Evaluation Time: 36222.56msINFO    : Evaluation Metrics:INFO    : =========================INFO    : | mAP_box       | 0.03% |INFO    : | mAP50_box     | 0.05% |INFO    : | precision_box | 0.08% |INFO    : | recall_box    | 1.70% |INFO    : =========================INFO    : Key Metric (mAP_box): 0.03%INFO    : Core Temp : 49.0°C                                                    INFO    : Metis : 804.0fpsINFO    : CPU % : 31.4%INFO    : End-to-end : 74.7fps./inference.py yolo11s-coco-onnx dataset --no-display --show-device-fpsINFO    : Model:      yolo11s-coco-onnx                            

I’m doing some testing with the PCI-Express based AI card. After some initial hiccups due to motherboard compatibility which required me to swap place for GPU and AI-card and then exposing the card into a ubuntu-22.04 docker container it’s working fine. I’m impressed by the performance of the card compared to using --pipe torch and having a CUDA based backend for the benchmark problems identifying traffic. I want to do inference with not only custom weights, but also custom nets and see how it performs. My wish is to have a sample that uses numpy f32 arrays as input and output and does inference on the card given a model from a ONNX file. To keep it simple, I made a PyTorch net [3, 64][ Relu ][64, 64][ Relu ][64, 1] and trained iton a mathematical function f(x1, x2, x3) = y.This toy net was exported to ONNX. From there I wanted to test inference on the AI card before moving to more complex things.I could not see that my wanted usecase was compatible with `deploy.py` and the required it

Putting this up for folk who may hit a similar issue.If you have a machine which has a single root PCIe device, where the output of lspci -tv looks something like this;alsutton@svr204:~$ lspci -tv-[0000:00]-+-00.0 Intel Corporation 8th Gen Core Processor Host Bridge/DRAM Registers +-02.0 Intel Corporation CoffeeLake-S GT2 [UHD Graphics 630] +-08.0 Intel Corporation Xeon E3-1200 v5/v6 / E3-1500 v5 / 6th/7th/8th Gen Core Processor Gaussian Mixture Model +-14.0 Intel Corporation Cannon Lake PCH USB 3.1 xHCI Host Controller +-14.2 Intel Corporation Cannon Lake PCH Shared SRAM +-16.0 Intel Corporation Cannon Lake PCH HECI Controller +-1b.0-[01]----00.0 Micron/Crucial Technology P2 [Nick P2] / P3 / P3 Plus NVMe PCIe SSD (DRAM-less) +-1f.0 Intel Corporation B360 Chipset LPC/eSPI Controller +-1f.3 Intel Corporation Cannon Lake PCH cAVS +-1f.4 Intel Corporation Cannon Lake PCH SMBus Controller

Hi, I’m trying to deploy and inference with a yolov8 segmentation model. I tried to follow the tutorial, but I just get:INFO: deploying model yolov8sseg-coco for 4 cores. This may take a while… How long should this run? I left it on for 3.5 hours and nothing happened (except the loading bar). This seems to be a model included in your library: yolov8sseg-coco.Any idea if I can turn on more extensive logging to see what is happening? Thank you in advance!  

When trying to run the application.py example, I get this error. The vent is activated. Per this thread I thought I could solve with a removal of the venv and reinstall but it’s being persistent now. Has anybody else found and overcome this?I’m using the Aetina Eval Board that came with the Project challenge kit.

Hi all,not sure if my issue is related to this one, but symptoms are at least similar, so I hope it matches best here. Otherwise sorry for derailing.I am trying to get my hands on the PCIe metis card on my AMD64 based system driven by an ASUS PRIME X570-P mainboard under Ubuntu-24.04 (6.8.0-58-generic x86_64). First issue was that the card did no show up under lspci, no matter which kernel boot options I tried. Then I looked into the mainboard BIOS to double-check the peripheral config and noticed that the PCIe slots are not fully independent. In fact, the board has two PCIex16 slots and three PCIex1, of whose the latter are meant for expansion cards and the former for graphic cards. They are pre-configured for ‘auto’ mode, which means that the BIOS detects how the slots are populated and splits available lanes between the two x16 slots.I changed this ‘auto’ setting to manually split the lanes to ‘x8+x8’ and after rebooting the card is now visible to the kernel:zefir@zefir-PC:~$ lspci

Hi,I get an error when I want to deploy the WASB model for sports ball detection and tracking, converted to onnx with opset 15:ERROR   : Traceback (most recent call last):ERROR   :   File "<frozen compiler.top_level>", line 584, in quantizeERROR   :   File "<frozen qtools_tvm_interface.graph_exporter_v2.graph_exporter>", line 254, in exportERROR   :   File "<frozen qtools_tvm_interface.graph_exporter_v2.graph_exporter>", line 209, in _convert_operatorsERROR   :   File "<frozen qtools_tvm_interface.graph_exporter_v2.replacement_functions>", line 242, in closureERROR   : RuntimeError: Multiple consecutive arithmetic operations by runtime ops found in the graph. This is not allowed presently.ERROR   : ERROR   : The above exception was the direct cause of the following exception:ERROR   : ERROR   : Traceback (most recent call last):ERROR   :   File "/home/aetina/Desktop/voyager-sdk/axelera/app/compile.py", line 456, in compileERROR   :     the_manifest = top_level.c

Using the Aetina Eval kit that came with the project challenge.Trying to run some of the standard examples, I get this error message:(venv) aetina@aetina:~/voyager-sdk$ ../ll-tests/ll-application.pyWARNING : This model is restricted to deploy for single-core (but can be run using multiple cores).INFO : Deploying model yolov8n-coco-onnx for 1 core. This may take a while...|████████████████████████████████████████| 10:18.7INFO : Deploying model yolov8s-coco-onnx for 1 core. This may take a while...|████████████████████████████████████████| 12:44.9arm_release_ver: g13p0-01eac0, rk_so_ver: 9WARNING : pyglet could not access the display, OpenGL is not available: No standard config is available.[libtriton_linux.c:629] DMA_GET_XFER_SYNC_STATUS failed: Connection timed out[AxeleraDmaBuf.cpp:271] DMABUF_METIS_WAIT failed: Connection timed out[ERROR][waitForAsyncKernelExecOrMemTransfer]: Wait host to dev failed[ERROR][axeCommandQueueExecuteCommandListsAsync]: Waiting for command lists failed: 0x

Hi,When trying to deploy yolo11 with custom weights, I noticed a significant performance drop due to the model conversion to onnx. Can I ask what method and parameters were used to convert the axelera yolo11s-coco-onnx model?Thank you!

Hi ​Axelera  team,I’m running parallel and cascading pipelines together with inference.py—the pipelines initialize and generally work. The issue is input routing: during inference I’m unable to reliably select/bind the correct video (file/RTSP/USB) to each network. For example, launching two independent detectors plus one cascade with multiple media sources (e.g., ./inference.py modelA modelB media/a.mp4 media/b.mp4 --pipe=gst --verbose) often results in the streams attaching to the wrong model or failing to open, while one stream runs as expected.Is there a supported way to explicitly map inputs to models/pipelines when using inference.py? For instance: CLI flags to pin inputs by index/name (e.g., --input[0]=..., --input[1]=...) YAML fields that define per-network input URIs A recommended PipeManager pattern for deterministic source‑to‑network routing when combining parallel and cascading graphs   the command i wrote:./inference.py parellel5_yv11pose_scpmyv8_bcfyv8_yv11_ppeyv8 med

Hey!PCIe card did run out of the box with a first running demo in minutes with no issues!Question: Is there a way to control the fan speed to reduce the noise of the card?Thanks!

Hi! I'm testing the Metis M2 on a Rapsberry Pi 5 using the Ubuntu 22.04 container. I can successfully get results using the inference test suggested by the article:./inference.py yolov8s-coco-onnx ./media/traffic1_480p.mp4 Unfortunately, I get around 4FPS. Is this expected? If I profile the device and the host separately (using --show-host-fps and --show-device-fps) I see that the device is running at ~800FPS while the host is the one bottlenecking at ~4FPS.I tried enabling GLES processing using export AXELERA_OPENGL_BACKEND=gles,3,1but, unfortunately, it makes not much of a difference. I also tried setting the PCI to gen3 by setting the following in the /boot/firmware/config.txt, without any luck:dtparam=pciex1_gen=3As a reference: these are the instructions I'm following:- https://support.axelera.ai/hc/en-us/articles/26362016484114-Bring-up-Voyager-SDK-in-Raspberry-Pi-5- https://support.axelera.ai/hc/en-us/articles/25953148201362-Install-Voyager-SDK-in-a-Docker-ContainerIs this perfo

I am using zsh as my shell application. While activating the environment, I encountered this error:   voyager-sdk git:(release/v1.3) ✗ source venv/bin/activate _AX_set_env:4: bad substitutionDo you think providing support for shells other than bash would help? 

I’ve used the yolov5s.pt file provided from the official yolov5 repository and converted it to onnx with this command:  python3 export.py --weights yolov5s.pt --imgsz 640 --batch-size 1 --include onnx --opset 17  . I’ve then tried to compile this with this command:  compile -i /home/axelera/Vision.AxeleraTesting/data/yolov5s.onnx -o /home/axelera/Vision.AxeleraTesting/data/compile --overwriteI got this output:09:49:45 [INFO] Dump used CLI arguments to: /home/vintecc/axelera/Vision.AxeleraTesting/data/compile/cli_args.json09:49:45 [INFO] Dump used compiler configuration to: /home/vintecc/axelera/Vision.AxeleraTesting/data/compile/conf.json09:49:45 [INFO] Input model has static input shape(s): ((1, 3, 640, 640),). Use it for quantization.09:49:45 [INFO] Data layout of the input model: NCHW09:49:45 [INFO] Using dataset of size 100 for calibration.09:49:45 [INFO] In case of compilation failures, turn on 'save_error_artifact' and share the archive with Axelera AI.09:49:45 [INFO] Quantizing

Hi All,I’ve got a PoC running using OpenCV+M1 Mac while I wait for my Metis hardware. In preparation for the kit arrival, I want to plan the porting process and get ahead on refracturing so I can compare the performance. I plan to use inferenceStream with Python to keep it really simple.Any experience here going from OpenCV to Voyager? Is it a drop in replacement or are they used in parallel?If there is a better development environment than OpenCV i’d be glad to hear it - my application is taking shape so I want to make sure I do not stray to far down the wrong path!

Hi again ​@Spanner ! 😊I need to infer an image obtained from byte data stored in a memory space (https://docs.python.org/3/library/mmap.html) and after converting the bytes I get an image.I've seen the sources available for create_inference_stream and I know that I can make inferences to images (https://github.com/axelera-ai-hub/voyager-sdk/blob/release/v1.2.5/docs/tutorials/application.md), but I have to specify the/path/to/file. But in this case I won't have that information.In the code I'll have this line where I read the bytes in the memory space:        frame_size=image_width*image_height*3 //RGB        frame_bytes = mm.read(frame_size)And then I get the new image:      frame = np.frombuffer(frame_bytes,dtype=np.uint8).reshape((image_height,image_width, 3))Is there any way to infer “frame”?Thank you!

Hi, I’m getting very pool performance on Yolo11l custom weigts after the conversion in onnx.I’ve used opset 17 and 16, do you have any suggestions?Thanks.

How can I deploy WASB model on metis m2?