playing around with gradio
Browse files
app.py
CHANGED
@@ -1,11 +1,11 @@
|
|
1 |
import functools
|
2 |
-
import matplotlib
|
3 |
-
matplotlib.use('Agg')
|
4 |
-
import matplotlib.pyplot as plt
|
5 |
|
6 |
import gradio as gr
|
|
|
|
|
7 |
|
8 |
|
|
|
9 |
BENCHMARK_DATA = {
|
10 |
"Greedy Search": {
|
11 |
"DistilGPT2": {
|
@@ -29,9 +29,9 @@ BENCHMARK_DATA = {
|
|
29 |
"A100": [],
|
30 |
},
|
31 |
"T5 Small": {
|
32 |
-
"T4": [
|
33 |
-
"3090": [],
|
34 |
-
"A100": [],
|
35 |
},
|
36 |
"T5 Base": {
|
37 |
"T4": [],
|
@@ -137,10 +137,19 @@ BENCHMARK_DATA = {
|
|
137 |
|
138 |
|
139 |
def get_plot(model_name, generate_type):
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
144 |
|
145 |
demo = gr.Blocks()
|
146 |
|
|
|
1 |
import functools
|
|
|
|
|
|
|
2 |
|
3 |
import gradio as gr
|
4 |
+
import seaborn as sns
|
5 |
+
import pandas as pd
|
6 |
|
7 |
|
8 |
+
# benchmark order: pytorch, tf eager, tf xla; units = ms
|
9 |
BENCHMARK_DATA = {
|
10 |
"Greedy Search": {
|
11 |
"DistilGPT2": {
|
|
|
29 |
"A100": [],
|
30 |
},
|
31 |
"T5 Small": {
|
32 |
+
"T4": [99.88, 1527.73, 18.78],
|
33 |
+
"3090": [55.09, 665.70, 9.25],
|
34 |
+
"A100": [124.91, 1642.07, 13.72],
|
35 |
},
|
36 |
"T5 Base": {
|
37 |
"T4": [],
|
|
|
137 |
|
138 |
|
139 |
def get_plot(model_name, generate_type):
|
140 |
+
df = pd.DataFrame(BENCHMARK_DATA[generate_type][model_name])
|
141 |
+
df["framework"] = ["PyTorch", "TF (Eager Execition)", "TF (XLA)"]
|
142 |
+
df = pd.melt(df, id_vars=["framework"], value_vars=["T4", "3090", "A100"])
|
143 |
+
|
144 |
+
g = sns.catplot(
|
145 |
+
data=df, kind="bar",
|
146 |
+
x="variable", y="value", hue="framework",
|
147 |
+
ci="sd", palette="dark", alpha=.6, height=6
|
148 |
+
)
|
149 |
+
g.despine(left=True)
|
150 |
+
# g.set_axis_labels("", "Body mass (g)")
|
151 |
+
# g.legend.set_title("")
|
152 |
+
return g.gcf()
|
153 |
|
154 |
demo = gr.Blocks()
|
155 |
|