
Maria Chen
Maria Chen
About
- Username
- Maria Chen
- Joined
- Visits
- 1,092
- Last Active
- Roles
- Member, ModeratorsKL520, ModeratorsKL720, ModeratorsAIModel
Comments
-
Hi, Besides using 100 images related to the ones you used in training and inferencing (the images should have variety as well) for quantization, there are other ways to debug when the onnx and bie results are greatly different: -Check the channel or…
-
Hi, According to the optimized onnx model, it seems like the correct input name for the onnx model is "images" https://www.kneron.com/forum/uploads/490/FXAIJC3DV9I0.png https://www.kneron.com/forum/uploads/210/M8OWCBP7CS4R.png Please chang…
-
Hi Jin, Thank you for providing the files. We'll take a look at them and get back to you. How many images did you use to quantize your onnx model into bie model? We recommend you to use around 100 images, relevant to the ones you used to train and i…
-
Hi Jin, Could you provide the following files so that we could check them for you? Thanks! -Optimized onnx model -Python script for converting your model in docker -Input images used in the python script -Postprocess code for optimized onnx model -L…
-
Hi, It seems like the KL720 device failed to reconnect when the firmware is updating on it. Just to make sure, have you installed the following dependencies on your Ubuntu 24.04? Also, please check if your OS is verified to build and run Kneron PLUS…
-
Hi, According to your screenshot, it seems like you've built the DFUT console, but your firmware is outdated (KDP). To update your firmware, please run DFUT_console at kneron_plus/build/bin. For example, to update KL520 to USB boot mode, you could e…
-
您好, 理論上,Ubuntu23.10以上也可以build C的Kneron PLUS examples以及DFUT console (DFUT console是在Kneron PLUS裡面build,而不是Kneron DFUT本身)。目前尚未聽說會更新建置環境
-
您好, 請問一下您說的環境是指Kneron PLUS嗎? 還是Kneron DFUT,或是其他的環境呢?
-
Hi, What are the specs for your Raspberry Pi4? Kneron DFUT supports Raspberry Pi OS - Buster (armv7l 32-bit). https://www.kneron.com/forum/uploads/353/QAG9THT9MWQ1.png If you're not using that, you may use DFUT by building the DFUT console on your K…
-
Hi, Sometimes, when you're using a virtual machine, you might need to make sure that your KL720 device (dongle/board) is connected to your virtual machine instead of your computer. What virtual machine software are you using? Are you using a KL720 d…
-
您好, 不好意思,我們也轉換失敗了。請嘗試在剛才的toolchain.py將datapath_bitwidth_mode設定成"all int8" datapath_bitwidth_mode="all int8" 我有試試看,這個有轉換成功。v0.24.0的算法有做過提升,雖然設定值是int8,有一些operator還是會使…
-
您好, 關於toolchain版本,請在v0.24.0的設定確認analysis (workspace/miniconda/lib/python3.7/site-packages/ktc/toolchain.py)裡面的這些設定是8-bit (int8): datapath_bitwidth_mode weight_bitwidth_mode model_in_bitwidth_mode model_out_bitwidth_mode https…
-
您好, 以下是toolchain建議的系統要求,會建議用quad-core CPU喔 https://www.kneron.com/forum/uploads/703/A7DW4FFIRHGK.png
-
您好, 感謝您的測試。以防萬一,可以請您確認您的toolchain版本嗎? 因為就算是pull latest的,不一定是最新的版本。 請使用: cat /workspace/version.txt 來檢查
-
您好, 轉換模型時,我們建議您將這些code寫在一個Python檔裡面之後再一次執行。
-
Hi Pradeep, Could you change the accessibility on your Google Drive so I could access it? Thank you for your help.
-
Hi Youngjun, Your binary_inf_results and fixed_point_inf_results should be almost the same, so it seems successful. https://www.kneron.com/forum/uploads/738/8S64MPMDQPD3.png According to your screenshot, it seems like your nef file is generated. It …
-
Hi, It seems like something is off with the nef file. Could you provide us the nef file and jpg file you used, and the python file if you edited it, so we could look into them? Thanks!
-
您好, 雖然QAT是toolchain v0.24.0剛剛支持,我們目前沒有詳細的文檔說明QAT,也沒有範例程式,不好意思。
-
您好, 是的,這個範例是舊的,如果要使用的話必須要對齊以前的pytorch等等的版本。
-
Hi, Thank you for testing the code. Since it would fail when loading the model for the first time, we've taken that out. Could you try running this code and see if it works? Thanks! https://www.kneron.com/forum/uploads/889/8H7LST4WXEP0.zip
-
您好, KneronDFUT支持Ubuntu 18.04。如果您是用20.04的話,請用kneron_plus裡面的DFUT console,謝謝! https://www.kneron.com/forum/uploads/807/E6G98BPC9K1Q.png
-
您好, 請問一下您有安裝以下的dependency嗎? Install Dependency - Document Center (kneron.com) https://www.kneron.com/forum/uploads/771/RVPB77MRSUG7.png 也可以請您試試看用C來build,然後跑generic_command看看會不會成功。 Build C code的時候,可以build wi…
-
Hi, If that's the case, could you test our Python code without editing anything? You could place it inside kneron_plus/python/example folder and run it. Thanks! https://www.kneron.com/forum/uploads/121/98ZVJ3RUVJR9.zip By the way, when you print out…
-
Hi, This should solve the error. Could you try it out with different sleep_secs and see which ones would work? (1 sec, 5 secs, 10 secs, etc.) Also, could you also try using connect_devices() instead of init_device()? Thanks!
-
Hi, This could be a timing issue. When a model is already inside the KL720, the device would reset and reconnect. This reconnection would depend on the host to detect the KL720 device. At this time, if it takes too long, there's a possibility that t…
-
Hi, Thank you for answering the questions and providing data. We can upload the same model onto 2 KL720 devices/chips by connecting different ports (in my case, port 61 and 53) at the same time with connect_devices(), then use load_model_from_file()…
-
Hi, According to your first post, it seems like the error appears sometimes and not every time. Do you know exactly where you encounter the error? Is it when you're uploading model onto the first KL720 chip or the second KL720 chip? Does your code l…
-
Hi, Thank you for answering the questions. I have some more questions: Are you using KL720 dongles? Or are you using KL720 development board? Also, have you followed the instructions in our documentation center? Install Dependency - Document Center …
-
Hi, Could you describe the environment/OS you're using the chips on? Are they both KL720 chips? How are they connected to your device? Which program on PLUS are you running? Thanks!