Default Avatar

Ethon Lin

Ethon Lin

About

Username
Ethon Lin
Joined
Visits
1,261
Last Active
Roles
Member, ModeratorsKL520, ModeratorsKL720, ModeratorsAIModel

Comments

  • Hello, 32.768KHzk27 Xtal需依照使用挑選的用料來設計 如果時間要很精準, ppm越小越好 (eg. ±20 ppm@EVB);  不要求精準, 可以是幾百ppm..
    in Requstment of 32.768KHz crystal matched with KL520 Comment by Ethon Lin April 2022
  • Now KNEO utility (registering tool) is only compatible with Windows and Ubuntu, so there is no method to register device on Raspberry. But you can still register your device on another PC (Win, Ubnutu) and take KL520 back to raspberry pi after regis…
    in Register Raspberry PI device Comment by Ethon Lin April 2022
  • Or you can build a console DFUT instead of pre-built DFUT GUI tool we provided. step1: Download plus v1.3 here https://www.kneron.com/en/support/developers/?folder=Kneron%20PLUS%20(incompatible%20successor%20to%20host_lib)/Kneron%20PLUS/&downl…
    in Segmentation fault running KneronDFUT Comment by Ethon Lin April 2022
  • Here is the PRi image created by Kneron. To make sure your environment, maybe you can find a new SD card to install with it and try again. https://www.kneron.com/en/support/developers/ https://www.kneron.com/forum/uploads/751/CV4Y4Q5Z70XB.png
    in Segmentation fault running KneronDFUT Comment by Ethon Lin April 2022
  • Hello, For the error you posted before "pthread_create", you can try to delete the file /Kneron_DFUT_rsapberry_pi_1.3.0/Kneron_DFUT/lib/libpthread.so.0 and execute again. The application will load system default library instead. https://w…
    in Segmentation fault running KneronDFUT Comment by Ethon Lin April 2022
  • Hello, Are you using RaspberryPi4? The DFUT tool only supports RPi4 now.
    in Segmentation fault running KneronDFUT Comment by Ethon Lin April 2022
  • Hello, Please feel free to contact our sales < tony.lu@kneron.us > to get more detail.
    in KL520 38x38 board availability Comment by Ethon Lin March 2022
  • Hello, 錯誤看起來是沒有燒錄到fw_ncpu.bin或是所燒錄的fw_scpu與fw_ncpu的版本不合所導致。若要使用companion mode的話，目前已經更新置1.7版(log顯示所燒錄的是1.4版)，還請參考文件來更新 http://doc.kneron.com/docs/#plus_c/introduction/upgrade_ai_device_to_kd…
    in KL520 board ncpu error Comment by Ethon Lin March 2022
  • Hello, 如果減少calibration用的圖數量就可以成功轉換的話，也有可能是docker的記憶體不足所導致，可以試著調大virtualbox中的ram size來試試是否會改善
    in YoloV3 Tiny Model migrate Error Comment by Ethon Lin March 2022
  • Hello, 請問一下，原始模型是使用哪一種framework? 模型看起來沒有執行過3.1.5 ONNX Optimization，各種framework的模型在轉成onnx後都還需要再執行onnx_optimizer.onnx2onnx_flow()來確保符合NPU架構。
    in 再將model轉換成nef出錯 Comment by Ethon Lin March 2022
  • Hello, 從錯誤訊息來看，裡面似乎有不符合NPU架構的operator (unsupported cpu node in setup bin) 還請參考文件章節2.3的支援列表(http://doc.kneron.com/docs/#toolchain/manual/#2-toolchain-docker-overview )，移除模型中不被支援的節點後再試
    in 再將model轉換成nef出錯 Comment by Ethon Lin March 2022
  • Hello, 在不同的平台上DFUT需要不同的編譯器，因此，目前僅提供Windows10, ubuntu18.04以及RaspberryPi4的環境，沒有jetson nano相關的支援。
    in 在Jetson nano上開啟 KneronDFUT 失敗 Comment by Ethon Lin February 2022
  • Hello, KL720在使用上為了確保傳輸速度與供電穩定度，會要求連接USB3.0的環境。從DFUT上USB speed欄位顯示為high speed來看，KL720在你的樹梅派上僅連接到USB2.0，還請確認連接的USB孔與連接線是否都支援USB3.0 ，成功連接USB3.0的話將會顯示為super speed。
    in 樹莓派連接720問題 Comment by Ethon Lin February 2022
  • Hello, Yes, the input_names=["data_out"] means the input layer name. You can check it in https://netron.app/. https://www.kneron.com/forum/uploads/788/W462AIRVWYA3.jpg Can you show the error message you encountered to us for debugging? 2…
    in How do I infer with the ONNX model? Comment by Ethon Lin February 2022
  • Hello, 目前沒有音訊前處理的範例，影像前處理的部分依據模型需求會有所不同，連結的preprocess()為 tiny yolo v3這個模型的前處理 http://doc.kneron.com/docs/#toolchain/yolo_example/#step-4-check-onnx-model-and-preprocess-and-postprocess-are-good KL720加速…
    in 整合問題 Comment by Ethon Lin February 2022
  • Hello, 以KL720 CPU的算力來做頻譜轉換可能會相當的慢，導致整體速度表現不佳。建議還是在前端就先處理好之後再送入KL720做模型推論。
    in 整合問題 Comment by Ethon Lin February 2022
  • 因為kneron plus的library(kp)並沒有登錄到pip的資料庫中，所以直接透過pip install kp是找不到的，以Windows為例，需透過安裝kneron_plus_v1.3.0\python\package\windows\KneronPLUS-1.3.0-py3-none-any.whl 才行，詳細安裝步驟可以參考下面網頁的說明。 http://doc.…
    in pip3 install kp失敗 Comment by Ethon Lin January 2022
  • 還請協助描述一下所使用的環境(Win? Linux? )，python的版本等環境資訊 截圖時也請包含安裝指令，方便我們了解是在哪個步驟發生錯誤
    in pip3 install kp失敗 Comment by Ethon Lin January 2022
  • kdp_host_api為舊架構host_lib的library，新架構PLUS已全面改用kp，故兩者是不能互通的 若想要使用resnet的話可以參考PLUS專案中的classification範例 (/kneron_plus_v1.3.0/python/example_model_zoo/KL520KnModelZooGenericInferenceClassification.py ) PLUS 下載連…
    in Kneron PLUS是否可以使用kdp_host_api ? Comment by Ethon Lin January 2022
  • For processing with model yolo v3, please refer to the our document center. http://doc.kneron.com/docs/#toolchain/yolo_example/
    in yolov3-tiny model convert problem Comment by Ethon Lin December 2021
  • In latest version of toolchain, we provided a python api for model inference. You can compare your onnx model and nef model through the api "kneron_inference()". For more detail, please refer to the link below. http://doc.kneron.com/docs/#…
    in Does KL520 support the darknet backbone? Comment by Ethon Lin December 2021
  • Hello, Here are version condition of tensorflow and tflite, please make sure the version of your model and try again. Tensorflow = 1.14.0 (tensorflow 2 is not supported yet) tflite: unquantized TF Lite models
    in about the example of toolchain Comment by Ethon Lin December 2021
  • Yes, please refer to our document center. http://doc.kneron.com/docs/#model_training/classification/
    in Resnet classification Sample Comment by Ethon Lin December 2021
  • Hello, The python api of kneron's toolchain is ktc, please try "import ktc" instead of "import kte". https://www.kneron.com/forum/uploads/119/CNDSFG3F67SF.jpg https://www.kneron.com/forum/uploads/788/HJA2R068Q7N5.png
    in Kneron toolchain Docker image does not have ktc? Comment by Ethon Lin December 2021
  • Hello, 可以的話是否方便提供截圖呢? 畫面中希望可以包含輸入的指令以及完整的錯誤訊息，但從描述來看不太能確認你遇到了什麼問題
    in windows 使用 MSYS2 執行Kneron_DFUT產生錯誤 Comment by Ethon Lin December 2021
  • 還是可以的，但DFUT的UI介面僅支援kdp -> kdp2，若要燒回則需透過command line執行KneronDFUT.exe 是的，PLUS上就不應該有這樣的錯誤。 若是使用的是Kneron提供的YOLO模型則可直接使用KL520DemoCamAppYoloInference，或是參考kl520_demo_generic_inference_post_yol…
    in KL520 dongle 出現異常問題 Comment by Ethon Lin November 2021
  • Hi Fred, 目前在kdp架構下(host_lib)的usb driver沒有特別處理這類操作，所以必須重新插拔或reset才能解決。 PLUS (kdp2)架構下有額外處理，且使用fw_loader的架構也更能確保系統穩定性，目前PLUS針對RPi4也有提供對應的工具，建議更新體驗。
    in KL520 dongle 出現異常問題 Comment by Ethon Lin November 2021
  • Hi Fred, 確認一下，是原本應用程式與KL520的dongle都可以正常使用，但執行中直接重啟主機後發生這個錯誤現象嗎? 從錯誤訊息推測，錯誤是在執行api.kdp_start_dme_ext 時所發生，表示傳送模型NEF資料給KL520失敗，懷疑是主機重新啟動時dongle沒有完全對應到主機的動作 …
    in KL520 dongle 出現異常問題 Comment by Ethon Lin November 2021
  • Hello, 模型的轉換會有三個階段 ONNX -> BIE (量化模型) -> NEF (編譯後的量化模型，供KL520 / KL720使用) 依據不同的script可以逐步轉換也可以直接ONNX到NEF，從你使用fpAnalyserBatchCompile_520.py表示你希望從ONNX直接轉換成NEF，這樣的話參數中也需要有量化…
    in 如果要用其他模型轉換的onnx檔，該如何設定input_params.json? Comment by Ethon Lin November 2021
  • Hello, 目前examples內的yolo範例主要是針對耐能所提供的yolo NEF做處理，包含model id, post process function 都有相當的限制 若使用的是我們所提供的training script，建議改使用範例/kneron_plus/python/example_model_zoo/KL720KnModelZooGenericInferenceYolov5.…
    in YOLOv5 on KL720 toolchain Comment by Ethon Lin November 2021

© Kneron, Inc.