Kwinqaku elilandelayo siza kujonga iNetron. Yi le inkqubo yokujonga imifuziselo ye uthungelwano lwe-neural. Esi sicelo esisebenzisa i-Electron / NodeJS kwaye sipapashwe phantsi kwelayisenisi ye-MIT, sinokuyiqhuba kwi-Gnu / Linux, macOS, kwiinkqubo zeWindows kunye nakwi-browser yewebhu.
Le nkqubo yenziwa nguLutz Roeder. INetron sisixhobo somthombo ovulekileyo esikuvumela ukuba ubone iimodeli zenethiwekhi ye-neural, nayo iya kusivumela ukuba sihlalutye isakhiwo somzekelo kwaye ngaloo ndlela siqinisekise ukuba ihambelana noyilo olulindelekileyo. Yisoftware ehambelana nezakhelo ezahlukeneyo kunye neefomathi zemodeli.
Iifomati ezixhaswayo zeNetron
inetron inenkxaso yeefomathi njengoko zinjalo:
- ONX(.onnx, .pb, .pbtxt)
- Keras (.h5, .keras)
- TensorFlow lite (.tflite)
- Ikofu (.icaffemodel, .prototxt)
- Umnatha omnyama (.cfg)
- Undoqo ML (.mlmodel)
- MNN (.mnn)
- MXNet (.imodeli, -symbol.json)
- NONE.iparam)
- Ukubheqa (.zip, __imodeli__)
- ICaffe2 (predict_net.pb)
- barracuda.nn)
- Injini (.tmfile)
- TNN (.tnnproto)
- I-RKNN (.rknn)
- Mind Spore Lite (.Nksk)
- Ewe (.uff)
Ukongeza, iNetron nayo unenkxaso yovavanyo ye; TensorFlow (.pb, .meta, .pbtxt, .ckpt, .index), PyTorch (.pt, .pth), TorchScript (.pt, .pth), OpenVINO (.xml), Torch (.t7), Arm NN (.armnn), BigDL (.bigdl, .model), Chainer (.npz, .h5), CNTK (.model, .cntk), Deeplearning4j (.zip), MediaPipe (.pbtxt), ML.NET (.zip ), scikit-learn (.pkl), TensorFlow.js (model.json, .pb).
Faka iNetron Neural Network Viewer kwi-Ubuntu
Uvavanyo olusuka kwibhrawuza yewebhu
Ngaphambi kokuthatha isigqibo sokufakela le nkqubo, sinokukhetha yivavanye kwisikhangeli sewebhu. Ukuba awunayo imodeli onokuthi uyilayishe ukuze uyivavanye, ungasebenzisa imizekelo yemizekelo enokufumaneka kwi indawo yokugcina kwiGitHub yeprojekthi, ukukhuphela okanye ukuvula ngolu guqulelo lomkhangeli zincwadi.
Njengepakethe ye-snap
Ukuba uthatha isigqibo sokuyifaka le software kwikhompyuter yakho, Ungayifaka le nkqubo ngephakheji yayo ye-snap, enokufumaneka kwi Ukutshintsha.
Njengoko benditshilo, iNetron Neural Network Viewer inokufakwa kwi-Ubuntu nge-Snap ngokwenza oku kulandelayo. Ukuqala kuya kufuneka sivule i-terminal (Ctrl + Alt + T) kwaye emva koko siya kwenza faka uguqulelo oluzinzileyo lwenkqubo usebenzisa umyalelo:
sudo snap install netron
Emva kofakelo, ukuba ufuna hlaziya inkqubo, kwi-terminal kufuneka uphumeze:
sudo snap refresh netron
Emva kwakho konke oku ngasentla, singakwazi qala inkqubo ukusuka kusetyenziso lwemenyu okanye nakwesiphi na esinye isiqalisi esinaso kunikezelo lwethu. Ukongeza, sinokuyiqala ngokuchwetheza kwi-terminal (Ctrl + Alt + T):
netron
Khupha ngaphandle
ukuba Khupha iNetron Neural Network Viewer efakwe ngephakheji ye-Snap, kuya kufuneka siphumeze kuphela kwi-terminal (Ctrl + Alt + T) umyalelo:
sudo snap remove netron
Khuphela iAppImage
Sinokusebenzisa le nkqubo ngokusebenzisa Iphakheji ye-AppImage enokukhutshelwa kwi iphepha lokukhupha iprojekthi. Ukongeza kokukwazi ukukhuphela le phakheji kwisikhangeli sewebhu, siya kuba nethuba lokusebenzisa wget ukubamba ifayile.
ukuba khuphela ingxelo yokugqibela epapashwe namhlanjeKuya kufuneka sivule kuphela i-terminal (Ctrl + Alt + T) kwaye siyenze kuyo:
wget https://github.com/lutzroeder/netron/releases/download/v5.3.4/Netron-5.3.4.AppImage
Xa ukhuphelo lugqityiwe, siye nika imvume yokwenza ifayile ukuba sisanda kukhuphela. Kule nto, ukuba sifudukela kwifolda apho ifayile igcinwe khona, kuya kufuneka siphumeze lo myalelo kuphela:
sudo chmod +x Netron-5.3.4.AppImage
Emva komyalelo wangaphambili, singakwazi qala inkqubo ngokunqakraza kabini kwifayile, okanye ngokuchwetheza kwi-terminal efanayo:
./Netron-5.3.4.AppImage
INetron yindlela elula yokubona uthungelwano lwe-neural. Le nkqubo izakusivumela ukuba sisebenzise uluhlu olubanzi lwezakhelo kunye neentlobo zemodeli ehambelanayo. Iyakaleka ngokwenene kwaye iyasebenziseka kubantu abaninzi kuluntu lokufunda. Imizobo ingade ithunyelwe ngaphandle, nangona unokufuna ukusebenzisa indlela eyahlukileyo ukuba injongo yakho kukwenza imizobo yoshicilelo, ngakumbi xa inzulu kakhulu.
Abasebenzisi abafuna, banako fumana ulwazi oluninzi malunga nale nkqubo kwi iwebhusayithi yeprojekthi okanye kweyakho Indawo yokugcina izinto zeGitHub.