Clémentine commited on
Commit
1886745
1 Parent(s): ca4a86f

10 samples viewer

Browse files
Files changed (2) hide show
  1. app.py +50 -0
  2. samples.json +0 -0
app.py ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import gradio as gr
2
+ import datasets
3
+ import random
4
+ import json
5
+
6
+ def load_from_ds():
7
+ name_to_dataset = {
8
+ "bbh": ("SaylorTwift/bbh", "test", ["sports_understanding", "tracking_shuffled_objects_three_objects", "navigate", "snarks", "date_understanding", "reasoning_about_colored_objects", "object_counting", "logical_deduction_seven_objects", "geometric_shapes", "web_of_lies", "movie_recommendation", "logical_deduction_five_objects", "salient_translation_error_detection", "disambiguation_qa", "temporal_sequences", "hyperbaton", "logical_deduction_three_objects", "causal_judgement", "formal_fallacies", "tracking_shuffled_objects_seven_objects", "ruin_names", "penguins_in_a_table", "boolean_expressions", "tracking_shuffled_objects_five_objects"]),
9
+ #"GPQA": "Idavidrein/gpqa",
10
+ "ifeval": ("wis-k/instruction-following-eval", "train", None),
11
+ "math": ("lighteval/MATH-Hard", "test", None),
12
+ "mmlu_pro": ("TIGER-Lab/MMLU-Pro", "test", None),
13
+ "musr": ("TAUR-Lab/MuSR", "murder_mysteries", ["default"]),
14
+ "musr": ("TAUR-Lab/MuSR", "object_placements", ["default"]),
15
+ "musr": ("TAUR-Lab/MuSR", "team_allocation", ["default"])
16
+ }
17
+
18
+ examples = {}
19
+ for name, (path, split, subsets) in name_to_dataset.items():
20
+ if subsets is None:
21
+ ds = datasets.load_dataset(path, split=split)
22
+ examples[name] = {"default": random.sample(list(ds), 10)}
23
+ else:
24
+ examples[name] = {}
25
+ for subset in subsets:
26
+ ds = datasets.load_dataset(path, subset, split=split)
27
+ examples[name][subset] = random.sample(list(ds), 10)
28
+
29
+ examples = json.load(open("samples.json"))
30
+
31
+ with gr.Blocks() as demo:
32
+ header = gr.Markdown("# Samples explorer")
33
+ with gr.Row():
34
+ name = gr.Dropdown(choices=examples.keys(), label="Dataset")
35
+ subset = gr.Dropdown(choices=[], label="Subset")
36
+ index = gr.Dropdown(choices=range(10), value=0, label="Choice index")
37
+
38
+ with gr.Row():
39
+ text_box = gr.JSON(value={})
40
+
41
+ def change_subset_dropdown(name):
42
+ keys = sorted(examples[name].keys())
43
+ return gr.Dropdown(choices=keys, value=keys[0], label="Subset")
44
+
45
+ name.change(fn=change_subset_dropdown, inputs=name, outputs=subset)
46
+
47
+ for dropdown in [name, subset, index]:
48
+ dropdown.change(fn= lambda name, subset, index: examples[name][subset][index], inputs=[name, subset, index], outputs=text_box)
49
+
50
+ demo.launch()
samples.json ADDED
The diff for this file is too large to render. See raw diff