tensorflow-yolov4-tflite VS tensorflow-yolo-v3

Hello all, I've been desperate for help in the tensorflow yolo objection detection framework. Basically, in https://github.com/hunglc007/tensorflow-yolov4-tflite 's implementation, there is problem whereby with GPU, it only detects the first frame in the whole video while CPU works fine but absolutely slow. There are issues opened at https://github.com/hunglc007/tensorflow-yolov4-tflite/issues/282 but no proper solution is found. I've also opened a question stackoverflow: https://stackoverflow.com/questions/68333281/tensorflow-yolov4-detect-video

I used yolov4 and trained model on my classes, then saved weights in file.weights. Now, I want to integrate that model into my flutter app. On firebase, there is an option to use a custom model but that requires uploading .tflite file. My question is how can I convert my trained model weights and upload them as .tflite so could be used in my app. I have tried following this https://github.com/mystic123/tensorflow-yolo-v3 but not success. I would appreciate your help in the conversion of .weights to .tflite or suggest of there is any other way round