#======================== """ LOAD DATA """ list_of_data = [] for path in data_paths: list_of_data += glob.glob(path) list_of_data.sort(key=os.path.getmtime) if last_nfiles > 0: list_of_data = list_of_data[-last_nfiles:] if len(list_of_data) == 0: exit() loader = DataLoader(list_of_data) if td: if eligibility_trace: _child_stats = loader.child_stats n = _child_stats[:, 0] q = _child_stats[:, 3] _episode = loader.episode _score = loader.score _v = np.sum(n * q, axis=1) / np.sum(n, axis=1) values = np.zeros(_v.shape) for idx, ep in enumerate(_episode): idx_r = idx weight = 1.0 _sum = 0
list_of_data = [] for path in data_paths: list_of_data += glob.glob(path) list_of_data = sorted(list_of_data, key=keyFile) x = [] y_s = [] y_s_err = [] y_l = [] y_l_err = [] for f in list_of_data: print('Processing: ', f) try: _tmp = DataLoader([f,]) except: pass _idx = np.ediff1d(_tmp.episode, to_end=-1) == -1 _cycle = _tmp.cycle[0] _scores = _tmp.score[_idx] s_mean = _scores.mean() s_stddev = _scores.std() / np.sqrt(np.sum(np.absolute(_idx))) _lines = _tmp.lines[_idx] l_mean = _lines.mean() l_stddev = _lines.std() / np.sqrt(np.sum(np.absolute(_idx))) x.append(_cycle) y_s.append(s_mean) y_s_err.append(s_stddev) y_l.append(l_mean)
y_f = offset_y + b_h color = 'gray' + str(int(100*v)) canvas.create_rectangle(x_i, y_i, x_f, y_f, fill=color) index = 0 if __name__ == '__main__': master = Tk() master.geometry('%dx%d'%(width, height)) master.resizable(False, False) master.title('Replay') list_of_data = [] for path in data_paths: list_of_data += glob.glob(path) data = DataLoader(data_paths) canvas_frame = Frame(master) canvas_frame.grid(row=0,column=0,rowspan=10,columnspan=5) canvas_frame_2 = Frame(master) canvas_frame_2.grid(row=5,column=5,rowspan=5,columnspan=5) info_frame = Frame(master) info_frame.grid(row=0,column=5,rowspan=3,columnspan=5) control_frame = Frame(master) control_frame.grid(row=3,column=5,rowspan=1,columnspan=5) control_frame_2 = Frame(master) control_frame_2.grid(row=4,column=5,rowspan=1,columnspan=5)