Hugging Face's logo

ch-min
/
Fine-tuning-data

Model card Files
xet
 Community
Fine-tuning-data
File size: 6,660 Bytes
849ca03
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import seaborn as sns
import os # Import os for path joining

# Provided data (shortened for brevity, full data used in execution)
# data = {
#   "SYNTHETIC": {
#     "status": "Success",
#     "category_counts": {
#       "point_prediction_prism": 32.85, "simple_direction_sat": 11.1, "yes_no": 8.51, "short_answer_other": 8.31,
#       "number_only": 5.37, "json_answer:small": 3.15, "json_answer:large": 3.14, "json_answer:no": 2.68,
#       "json_answer:yes": 2.6, "multiple_choice_letter_spar7m": 1.85, "object_movement_sat": 1.51,
#       "json_answer:1": 1.49, "json_answer:0": 1.2, "json_answer:2": 0.95, # ... other json_answers omitted
#       "long_descriptive_other": 0.82
#       # ... other categories omitted for brevity in this display
#     }
#   },
#   "REAL": {
#     "status": "Success",
#     "category_counts": {
#       "yes_no": 33.33, "multiple_choice_letter_spar7m": 22.7, "point_list_robospatial": 16.67,
#       "number_only": 12.29, "long_descriptive_other": 10.35, "descriptive_sentence_spar7m": 4.66
#     }
#   },
#   "STATIC": {
#     "status": "Success",
#     "category_counts": {
#       "point_prediction_prism": 20.0, "yes_no": 17.12, "multiple_choice_letter_spar7m": 13.26,
#       "simple_direction_sat": 7.77, "point_list_robospatial": 6.67, "long_descriptive_other": 6.14,
#       "number_only": 3.87, "short_answer_other": 2.68, "descriptive_sentence_spar7m": 2.51,
#       "json_answer:large": 2.15, "json_answer:small": 2.15, "json_answer:no": 1.81, "json_answer:yes": 1.78,
#       "json_answer:1": 1.01 # ... other categories omitted
#     }
#   },
#   "DYNAMIC": {
#     "status": "Success",
#     "category_counts": {
#       "number_only": 48.02, "short_answer_other": 23.16, "yes_no": 16.92, "object_movement_sat": 7.87,
#       "json_answer:large": 0.62, "json_answer:small": 0.57 # ... other categories omitted
#     }
#   },
#   "PERCEPTION": {
#     "status": "Success",
#     "category_counts": {
#       "number_only": 98.89, "short_answer_other": 0.27, "json_answer:no": 0.11, "json_answer:small": 0.1,
#       "json_answer:large": 0.1, "json_answer:yes": 0.1 # ... other categories omitted
#     }
#   },
#   "REASONING": {
#     "status": "Success",
#     "category_counts": {
#       "point_prediction_prism": 20.0, "yes_no": 18.64, "multiple_choice_letter_spar7m": 13.11,
#       "simple_direction_sat": 6.93, "point_list_robospatial": 6.67, "long_descriptive_other": 6.07,
#       "short_answer_other": 5.15, "descriptive_sentence_spar7m": 2.51, "json_answer:small": 2.17,
#       "json_answer:large": 2.13, "json_answer:no": 1.78, "json_answer:yes": 1.71, "json_answer:1": 1.03,
#       "object_movement_sat": 0.94 # ... other categories omitted
#     }
#   },
#   "_2D": {
#     "status": "Success",
#     "category_counts": {
#       "number_only": 56.27, "simple_direction_sat": 25.85, "yes_no": 4.43, "json_answer:no": 2.68,
#       "json_answer:yes": 2.6, "json_answer:1": 1.49, "json_answer:0": 1.2, "json_answer:small": 0.72,
#       "json_answer:large": 0.7, "json_answer:2": 0.58 # ... other categories omitted
#     }
#   },
#   "_3D": {
#     "status": "Success",
#     "category_counts": {
#       "yes_no": 21.13, "point_prediction_prism": 20.91, "multiple_choice_letter_spar7m": 14.48,
#       "short_answer_other": 8.87, "point_list_robospatial": 6.97, "long_descriptive_other": 6.68,
#       "descriptive_sentence_spar7m": 2.62, "json_answer:large": 2.44, "json_answer:small": 2.44,
#       "object_movement_sat": 1.62 # ... other categories omitted
#     }
#   }
# }

# --- Data Preparation ---
import json
with open('__finetuning_data_analysis.json', 'r') as f:
    data = json.load(f)
# Simplify 'json_answer:...' categories into one 'json_answer'
simplified_data = {}
all_categories = set()

for dataset, info in data.items():
    if info['status'] == 'Success':
        counts = info['category_counts']
        new_counts = {}
        json_sum = 0
        for cat, perc in counts.items():
            if cat.startswith("json_answer:"):
                json_sum += perc
            else:
                new_counts[cat] = perc
                all_categories.add(cat)
        if json_sum > 0:
            new_counts['json_answer'] = round(json_sum, 2)
            all_categories.add('json_answer')
        simplified_data[dataset] = new_counts

# Convert to DataFrame
df = pd.DataFrame(simplified_data).fillna(0).T # Transpose for plotting

# --- Select Top Categories and Group Others ---
# Calculate total percentage for each category across all datasets
category_totals = df.sum().sort_values(ascending=False)

# Keep top N categories (adjust N as needed)
top_n = 10
top_categories = category_totals.head(top_n).index.tolist()
other_categories = category_totals.iloc[top_n:].index.tolist()

# Group remaining categories into 'Other'
if other_categories:
    df['Other'] = df[other_categories].sum(axis=1)
    df = df[top_categories + ['Other']] # Keep only top N + Other
else:
    df = df[top_categories] # Keep only top N if no 'Other' needed

# Ensure DataFrame columns are sorted for consistent legend order
df = df[sorted(df.columns)]


# --- Plotting ---
plt.style.use('seaborn-v0_8-whitegrid') # Use a clean style
fig, ax = plt.subplots(figsize=(14, 8))

# Define a color palette
colors = sns.color_palette("tab20", n_colors=len(df.columns))

# Create the stacked bar chart
df.plot(kind='bar', stacked=True, ax=ax, color=colors, width=0.8)

# --- Formatting ---
ax.set_xlabel("Dataset Category", fontsize=12, fontweight='bold')
ax.set_ylabel("Percentage of Answer Formats (%)", fontsize=12, fontweight='bold')
ax.set_title("Distribution of Answer Formats Across Fine-tuning Datasets", fontsize=16, fontweight='bold', pad=20)
ax.tick_params(axis='x', rotation=45, labelsize=11)
ax.tick_params(axis='y', labelsize=11)
ax.set_ylim(0, 100)
ax.yaxis.grid(True, linestyle='--', alpha=0.7)
ax.xaxis.grid(False) # Remove vertical grid lines

# Add percentage labels inside the bars (only for segments > 5%)
# Labels removed for clarity in this version

# --- Legend ---
# Place legend outside the plot area
ax.legend(title="Answer Format Categories", bbox_to_anchor=(1.02, 1), loc='upper left', borderaxespad=0., fontsize=10, title_fontsize=11)

plt.tight_layout(rect=[0, 0, 0.85, 1]) # Adjust layout to make space for legend

# --- Save the plot instead of showing it ---
output_filename = "answer_format_distribution.png"
plt.savefig(output_filename, dpi=300, bbox_inches='tight') # Use bbox_inches='tight' to include legend
print(f"Plot saved as {output_filename}")
# plt.show() # Replaced with savefig