dearwolf/aider
1
0
Fork 0
mirror of https://github.com/Aider-AI/aider.git synced 2025-05-29 08:44:59 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

tweaking graph labels

Browse source
This commit is contained in:
Paul Gauthier 2024-02-03 08:25:19 -08:00
parent da14474cbe
commit f5887a5098
1 changed files with 11 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

16
benchmark/benchmark.py
View file

@ -79,8 +79,8 @@ def show_stats(dirnames, graphs):
if "folk" in row.dir_name: if "folk" in row.dir_name:
row.edit_format += "folk" row.edit_format += "folk"
if row.model == "gpt-4-0613": # if row.model == "gpt-4-0613":
row.model += "\n(8k context window is\ntoo small for benchmark)" # row.model += "\n(8k context window is\ntoo small for benchmark)"
if row.completed_tests < 89: if row.completed_tests < 89:
print(f"Warning: {row.dir_name} is incomplete: {row.completed_tests}") print(f"Warning: {row.dir_name} is incomplete: {row.completed_tests}")
@ -126,8 +126,8 @@ def show_stats(dirnames, graphs):
# dump(df) # dump(df)
if graphs: if graphs:
# plot_timing(df) # plot_timing(df)
# plot_outcomes(df, repeats, repeat_hi, repeat_lo, repeat_avg) plot_outcomes(df, repeats, repeat_hi, repeat_lo, repeat_avg)
plot_refactoring(df) # plot_refactoring(df)
def plot_timing(df): def plot_timing(df):
@ -249,7 +249,13 @@ def plot_outcomes(df, repeats, repeat_hi, repeat_lo, repeat_avg):
) )
ax.set_xticks([p + 0.5 * width for p in pos]) ax.set_xticks([p + 0.5 * width for p in pos])
ax.set_xticklabels(models) model_labels = []
for model in models:
pieces = model.split("-")
ml = "-".join(pieces[:2]) + "-\n" + "-".join(pieces[2:])
model_labels.append(ml)
ax.set_xticklabels(model_labels)
top = 95 top = 95
ax.annotate( ax.annotate(