Last active
April 24, 2021 09:56
-
-
Save apsdehal/1d348bb5c1a179c430f9676d4c2972f0 to your computer and use it in GitHub Desktop.
This can be used to plot a train.log from MMF using plotly. Update `JOBS_BASEPATH` to point to your save folders and `METRIC` variable to point to metric that you want to plot.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import sys | |
import os | |
import numpy as np | |
import json | |
from collections import defaultdict | |
import seaborn | |
import glob | |
import random | |
import plotly | |
import plotly.graph_objs as go | |
from plotly.offline import download_plotlyjs, init_notebook_mode, plot, iplot | |
import plotly.io as pio | |
JOBS_BASEPATH = ["<path_to_save_folder>"] | |
METRIC = "val/vqa2/vqa_accuracy" | |
def read_logline(line): | |
line = line.strip() | |
if "{\"" in line: | |
info_index = line.find("{\"") | |
line = line[info_index:] | |
res = json.loads(line) | |
else: | |
info_index = line.find("progress") | |
line = line[info_index:] | |
res = { | |
split[0].strip(): split[1].strip() | |
for item in line.split(",") | |
for split in [item.strip().split(":")] | |
} | |
if "val/total_loss" in res: | |
new_res = {} | |
for k, v in res.items(): | |
if "val" not in k: | |
k = "val/" + k | |
new_res[k] = v | |
res = new_res | |
return res | |
def process_log(exp_out, label=None): | |
res = defaultdict(list) | |
filename = exp_out | |
if label is None: | |
label = filename.split(os.path.sep)[-2] | |
with open(filename) as f: | |
for line in f: | |
if "epoch" in line and "loss" in line: | |
r = read_logline(line) | |
for k, v in r.items(): | |
res[k].append(v) | |
return { | |
"res": res, | |
"label": label, | |
} | |
return res | |
files = [] | |
for path in JOBS_BASEPATH: | |
path = os.path.join(path, "train.log") | |
files += glob.glob(path, recursive=True) | |
results = {e + str(i): process_log(e) for i, e in enumerate(files)} | |
plt_data = [] | |
for i, r in results.items(): | |
g = go.Scatter( | |
x=r["res"]["num_updates"], | |
y=r["res"][METRIC], | |
name=r["label"] | |
) | |
plt_data.append(g) | |
layout = dict( | |
xaxis=dict(title="num_updates"), | |
yaxis=dict(title=METRIC), | |
font=dict( | |
size=8 | |
), | |
legend=dict(x=-.1, y=1.3), | |
hoverlabel = dict(namelength = -1) | |
) | |
fig = dict(data=plt_data, layout=layout) | |
iplot(fig, show_link=False) |
@dinhanhx I will edit this in the morning (PST) to support normal format as well.
@dinhanhx Can you try now?
@dinhanhx Can you try now?
It works perfectly. Thank you.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Sir, I try to run it with this log file, and I got this error