Ethon Lin
Ethon Lin
About
- Username
- Ethon Lin
- Joined
- Visits
- 1,325
- Last Active
- Roles
- Member, ModeratorsKL520, ModeratorsKL720, ModeratorsAIModel
Comments
-
sudo ./scan_devices指令表示以sudo權限來執行scan_devices這個執行檔，請先確認您執行的路徑下，透過指令 ls 來列出所有檔案後，看的到scan_devices這個執行檔再來執行 scan_devices為PLUS編譯後生成的執行檔，在完成整個build流程後(最後一部make -j 成功後)，會在您…
-
Hello, 模型後段看起來有些layer維度超過NPU的限制了，NPU只能接受4維 batch x channel x height x width，且batch size必須要是1 建議照紅線處裁剪掉模型後段的layers，並將裁剪掉的部分改在後處理內實現(請務必記得加入此部分運算，否則模型結果會錯誤) https://www…
-
Hello, bash: install_libusb.sh:沒有此一檔案或目錄 表示沒有在install_libusb.sh這個路徑下執行，所以系統找不到檔案.. 檔案在PLUS中有提供，路徑為 /kneron_plus_v3.0.0/kneron_plus/python/install_libusb.sh 另外 樹梅派本身有許多的版本，其OS也有許多版本，…
-
Hello, 這是耐能內部訓練來提供給商務客戶使用的版本，而mmpose則為mmlab所提供的公開模型，所以兩者效果有差。 若對於耐能內部的模型有興趣的話歡迎與我們業務洽詢。
-
https://www.kneron.com/forum/discussion/comment/2081#Comment_2081 這是壓縮分割檔，請將所有檔案下載後放在同一個資料夾內再進行解壓縮 https://www.kneron.com/forum/discussion/comment/2082#Comment_2082 是的，可以照文件中的網址至github上下載，也可以直接…
-
您好, 這類問題通常是由前/後處理有差異所造成，建議先在toolchain上透過simulator確認可以拿到正確的推論答案後再移到630上推論 simulator的使用可以參考https://doc.kneron.com/docs/#toolchain/manual_1_overview/#14-floating-point-model-preparation 參考1.4.3.…
-
你好， 依據你的應用我們這邊可能沒有現成符合你需求的模型，但有關於模型的編修耐能這邊有針對幾個公開模型做實作範例，可以供你參考 Yolov3: https://doc.kneron.com/docs/#toolchain/appendix/yolo_example/ 其他模型: https://doc.kneron.com/docs/#model_traini…
-
您好， 可以參考耐能model zoo中所提供的模型，當中也有一些classification模型可以供您評估是否符合您的需求 https://doc.kneron.com/docs/#model_training/#_top
-
Hi FK, Actually, both Python and C methods mentioned above are for running inference on the KL520 and are not related to model conversion. To convert a model from the original framework (e.g., Keras, PyTorch, etc.), you should use the corresponding …
-
Hi FK, To run model inference on Kneron's NPU, you need to perform quantization and compilation in Kneron's toolchain beforehand. The quantization process will reduce the model's file size. Although there might be a slight loss of precision during q…
-
Hi FK, Yes, if you program the firmware and model file into flash beforehand, there is no need to load the firmware and model before running the application. However, this method is typically used for mass production. You should verify that your fir…
-
Hi FK, The available memory depends on how you upload the model. Upload Model via USB: 35 MB for KL520 Load Model from Flash: 32 MB for KL520 Please note that the capacity is not only for the model size but also includes memory for input data, outpu…
-
Hi FK, It appears that you are using an outdated version (host_lib) that is no longer supported. Please uninstall the previous version and install the latest one. The new development tool is PLUS. You can download it from the following link: https…
-
Greetings, It's possible to run an audio AI model on the KL520 if the model consists of operators supported by the KL520. Here is the information on supported operators: https://doc.kneron.com/docs/#toolchain/appendix/operators/ Additionally, it is…
-
Hello, 若使用的是KL520 usb dongle的話，可以參考我們網頁的說明，透過我們的開發程式PLUS來完成你的應用 https://doc.kneron.com/docs/#plus_c/getting_started/ 設定好PLUS編譯環境後可以參考文件透過generic_inference API來進行推論 https://doc.kneron.com/docs/…
-
請參考文件的流程說明來進行量化(quantization)、編譯等流程 https://doc.kneron.com/docs/#toolchain/manual_1_overview/#12-workflow-overview
-
Hello, Could you please provide a specific example to clarify where the issue is? For instance, provide an ONNX file and explain the steps you took, then describe the differences you found abnormal. These steps will help us assist you in analyzing t…
-
Hello, Thanks for feedback. We did find some issues with the latest version 0.25.0 and are currently addressing them. Please hold on and wait for the next version.
-
https://www.kneron.com/forum/discussion/comment/1828/#Comment_1828 如Maria所貼的規格，Toolchain最低會需要quad-core CPU，因為你使用的是 Intel Duo可能導致不相容，還請理解。
-
Hello, It's recommended to build the DFUT tool by yourself instead of using the pre-built .sh file. Here is a reference link for building a console version of DFUT. https://www.kneron.com/forum/discussion/comment/1614/#Comment_1614
-
Hello, 有關您的問題如下 是的，因為是耐能的NPU是INT8，建議運算的數值介於 -2^n ~ 2^n這個範圍才不會犧牲精確度，耐能本身都是採用-0.5~0.5這個範圍 因為投影片是依據過去修改的經驗整理出來的，所以行數可能與現在最新的版本有所差異。但最後一頁的概念就是要將訓練…
-
Hello, 看訊息中有提到ArgMax，推測是模型中含有這個不支援的operator所造成 argmax是基於多個機率取最大值的運算，一般來說我們會放在後處理中再執行，所以要請你將onnx模型的最末端做調整，我們自己內部會是到Sigmoid的地方做結尾，剩餘的移到後處理來執行 https://w…
-
Hi Nguyen, It's possible to burn the KL720 FW image to blank flash through UART and JTag (JLink), but you need to set the HW setup to change bootstrap to UART mode before burning by UART. The Jlink would be the easier way to access flash. To use USB…
-
Hello, 若使用的是KL720的話，Mul與Sigmoid都要是可以轉換的才對，會方便請您提供onnx模型讓我們這邊分析後再跟您回報嗎?
-
您好，使用轉接頭將TypeC轉成TypeA是可行的，但從您的截圖來看，KL720依舊被降速成USB2.0 (High-Speed) 還請針對KL720的接線、轉接頭以及連接電腦時使用的usb port進行檢查，當中只要有一部分不符合usb3.0的規範整體就會被降速成USB2.0 若成功以USB3.0連接上的話，畫面…
-
Hello, 相關資訊可能還是要請您向耐能業務或我們的代理商-全科聯絡，以便確認您的需求並回覆您的問題。
-
Hello, 目前GPT相關功能只限商務客戶使用，若有相關需求請直接與耐能業務洽群。
-
Hi Tri, The model "yolov5_pd/models_720.nef" is trained by Kneron. We provided the NEF file as an example for users, but we will not release the original ONNX model. If you want to train the yolov5/v7 model yourself, you can refer to the …
-
Hello, Please refer to the following description for your question. It's recommended to combine different models into an NEF file before inference. You can then select the inference model by its ID. The issue with NEF combining occurred previously d…
-
Hi Haresh, Maybe you can try to build a console DFUT instead of pre-built DFUT GUI tool we provided. step1: Download the latest PLUS here https://www.kneron.com/tw/support/developers/ Or download PLUS v2.2.0 directly by following link. https://www.…