iicd/autodl-projects
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Update vis

Browse Source
This commit is contained in:
D-X-Y
2021-04-27 00:10:39 +08:00
parent d3371296a7
commit 8ae9237ef0
1 changed files with 27 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

44
exps/LFNA/vis-synthetic.py
View File

@@ -37,7 +37,9 @@ def draw_multi_fig(save_dir, timestamp, scatter_list, wh, fig_title=None):
fig = plt.figure(figsize=figsize) fig = plt.figure(figsize=figsize)
if fig_title is not None: if fig_title is not None:
fig.suptitle(fig_title, fontsize=LegendFontsize) fig.suptitle(
fig_title, fontsize=LegendFontsize, fontweight="bold", x=0.5, y=0.92
)
for idx, scatter_dict in enumerate(scatter_list): for idx, scatter_dict in enumerate(scatter_list):
cur_ax = fig.add_subplot(len(scatter_list), 1, idx + 1) cur_ax = fig.add_subplot(len(scatter_list), 1, idx + 1)
@@ -83,7 +85,7 @@ def compare_cl(save_dir):
save_dir.mkdir(parents=True, exist_ok=True) save_dir.mkdir(parents=True, exist_ok=True)
dynamic_env, function = create_example_v1( dynamic_env, function = create_example_v1(
# timestamp_config=dict(num=200, min_timestamp=-1, max_timestamp=1.0), # timestamp_config=dict(num=200, min_timestamp=-1, max_timestamp=1.0),
timestamp_config=None, timestamp_config=dict(num=200),
num_per_task=1000, num_per_task=1000,
) )
@@ -107,16 +109,18 @@ def compare_cl(save_dir):
# compute cl-min # compute cl-min
cl_xaxis_min = find_min(cl_xaxis_min, xaxis_all.mean() - xaxis_all.std()) cl_xaxis_min = find_min(cl_xaxis_min, xaxis_all.mean() - xaxis_all.std())
cl_xaxis_max = ( cl_xaxis_max = find_max(cl_xaxis_max, xaxis_all.mean() + xaxis_all.std())
find_max(cl_xaxis_max, xaxis_all.mean() + xaxis_all.std()) + idx * 0.1 """
)
cl_xaxis_all = np.arange(cl_xaxis_min, cl_xaxis_max, step=0.05) cl_xaxis_all = np.arange(cl_xaxis_min, cl_xaxis_max, step=0.05)
cl_yaxis_all = cl_function.noise_call(cl_xaxis_all) cl_yaxis_all = cl_function.noise_call(cl_xaxis_all)
current_data["cl_xaxis_all"] = cl_xaxis_all current_data["cl_xaxis_all"] = cl_xaxis_all
current_data["cl_yaxis_all"] = cl_yaxis_all current_data["cl_yaxis_all"] = cl_yaxis_all
"""
all_data[timestamp] = current_data all_data[timestamp] = current_data
global_cl_xaxis_all = np.arange(cl_xaxis_min, cl_xaxis_max, step=0.1)
global_cl_yaxis_all = cl_function.noise_call(global_cl_xaxis_all)
for idx, (timestamp, xdata) in enumerate(tqdm(all_data.items(), ncols=50)): for idx, (timestamp, xdata) in enumerate(tqdm(all_data.items(), ncols=50)):
scatter_list = [] scatter_list = []
scatter_list.append( scatter_list.append(
@@ -124,7 +128,7 @@ def compare_cl(save_dir):
"xaxis": xdata["lfna_xaxis_all"], "xaxis": xdata["lfna_xaxis_all"],
"yaxis": xdata["lfna_yaxis_all"], "yaxis": xdata["lfna_yaxis_all"],
"color": "k", "color": "k",
"s": 10, "s": 12,
"alpha": 0.99, "alpha": 0.99,
"xlim": (-6, 6), "xlim": (-6, 6),
"ylim": (-40, 40), "ylim": (-40, 40),
@@ -132,17 +136,21 @@ def compare_cl(save_dir):
} }
) )
cl_xaxis_all = current_data["cl_xaxis_all"] cur_cl_xaxis_min = cl_xaxis_min
cl_yaxis_all = current_data["cl_yaxis_all"] cur_cl_xaxis_max = cl_xaxis_min + (cl_xaxis_max - cl_xaxis_min) * (
idx + 1
) / len(all_data)
cl_xaxis_all = np.arange(cur_cl_xaxis_min, cur_cl_xaxis_max, step=0.01)
cl_yaxis_all = cl_function.noise_call(cl_xaxis_all, std=0.2)
scatter_list.append( scatter_list.append(
{ {
"xaxis": cl_xaxis_all, "xaxis": cl_xaxis_all,
"yaxis": cl_yaxis_all, "yaxis": cl_yaxis_all,
"color": "r", "color": "k",
"s": 10, "s": 12,
"xlim": (round(cl_xaxis_all.min(), 1), round(cl_xaxis_all.max(), 1)), "xlim": (round(cl_xaxis_min, 1), round(cl_xaxis_max, 1)),
"ylim": (round(cl_xaxis_all.min(), 1), round(cl_yaxis_all.max(), 1)), "ylim": (-18, 2),
"alpha": 0.99, "alpha": 0.99,
"label": "Continual Learning", "label": "Continual Learning",
} }
@@ -152,18 +160,20 @@ def compare_cl(save_dir):
save_dir, save_dir,
idx, idx,
scatter_list, scatter_list,
wh=(2000, 1300), wh=(2200, 1800),
fig_title="Timestamp={:03d}".format(idx), fig_title="Timestamp={:03d}".format(idx),
) )
print("Save all figures into {:}".format(save_dir)) print("Save all figures into {:}".format(save_dir))
save_dir = save_dir.resolve() save_dir = save_dir.resolve()
base_cmd = ( base_cmd = (
"ffmpeg -y -i {xdir}/%04d.png -vf fps=2 -vf scale=2000:1300 -vb 5000k".format( "ffmpeg -y -i {xdir}/%04d.png -vf fps=1 -vf scale=2200:1800 -vb 5000k".format(
xdir=save_dir xdir=save_dir
) )
) )
os.system("{:} -pix_fmt yuv420p {xdir}/vis.mp4".format(base_cmd, xdir=save_dir)) video_cmd = "{:} -pix_fmt yuv420p {xdir}/vis.mp4".format(base_cmd, xdir=save_dir)
os.system("{:} -c:a libvorbis {xdir}/vis.webm".format(base_cmd, xdir=save_dir)) print(video_cmd + "\n")
os.system(video_cmd)
# os.system("{:} {xdir}/vis.webm".format(base_cmd, xdir=save_dir))
if __name__ == "__main__": if __name__ == "__main__":