Spaces:
Sleeping
Sleeping
prasanna kumar
commited on
Commit
Β·
0d3569b
1
Parent(s):
6e43644
added openai and cohere models support along with token visuvalizations
Browse files- app.py +88 -24
- requirements.txt +4 -0
app.py
CHANGED
|
@@ -4,11 +4,24 @@ import ast
|
|
| 4 |
from collections import Counter
|
| 5 |
import re
|
| 6 |
import plotly.graph_objs as go
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
|
| 8 |
model_path = "models/"
|
| 9 |
|
| 10 |
# Available models
|
| 11 |
-
MODELS = ["Meta-Llama-3.1-8B", "gemma-2b"]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
|
| 13 |
def create_vertical_histogram(data, title):
|
| 14 |
labels, values = zip(*data) if data else ([], [])
|
|
@@ -25,32 +38,80 @@ def create_vertical_histogram(data, title):
|
|
| 25 |
)
|
| 26 |
return fig
|
| 27 |
|
| 28 |
-
def process_text(text:str,model_name):
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
|
| 34 |
-
def process_ids(ids:str,model_name):
|
| 35 |
-
tokenizer = AutoTokenizer.from_pretrained(model_path + model_name)
|
| 36 |
token_ids = ast.literal_eval(ids)
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
|
| 41 |
-
def
|
| 42 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
if input_type == "Text":
|
| 44 |
-
text,tokens,token_ids = process_text(text=input_value,model_name=model_name)
|
| 45 |
elif input_type == "Token IDs":
|
| 46 |
-
text,tokens,token_ids = process_ids(ids=input_value,model_name=model_name)
|
| 47 |
|
| 48 |
character_count = len(text)
|
| 49 |
word_count = len(text.split())
|
| 50 |
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
special_char_count = sum(1 for token in tokens if not token.isalnum() and token != 'β')
|
| 54 |
|
| 55 |
words = re.findall(r'\b\w+\b', text.lower())
|
| 56 |
special_chars = re.findall(r'[^\w\s]', text)
|
|
@@ -71,7 +132,9 @@ def process_input(input_type, input_value, model_name):
|
|
| 71 |
analysis += f"Special character tokens: {special_char_count}\n"
|
| 72 |
analysis += f"Other tokens: {len(tokens) - space_count - special_char_count}"
|
| 73 |
|
| 74 |
-
|
|
|
|
|
|
|
| 75 |
|
| 76 |
def text_example():
|
| 77 |
return "Hello, world! This is an example text input for tokenization."
|
|
@@ -85,8 +148,9 @@ with gr.Blocks() as iface:
|
|
| 85 |
|
| 86 |
with gr.Row():
|
| 87 |
input_type = gr.Radio(["Text", "Token IDs"], label="Input Type", value="Text")
|
| 88 |
-
model_name = gr.Dropdown(choices=MODELS, label="Select Model",value=MODELS[0])
|
| 89 |
|
|
|
|
| 90 |
input_text = gr.Textbox(lines=5, label="Input")
|
| 91 |
|
| 92 |
with gr.Row():
|
|
@@ -96,8 +160,8 @@ with gr.Blocks() as iface:
|
|
| 96 |
submit_button = gr.Button("Process")
|
| 97 |
|
| 98 |
analysis_output = gr.Textbox(label="Analysis", lines=6)
|
| 99 |
-
text_output = gr.Textbox(label="Text",lines=6)
|
| 100 |
-
tokens_output = gr.
|
| 101 |
token_ids_output = gr.Textbox(label="Token IDs", lines=2)
|
| 102 |
|
| 103 |
with gr.Row():
|
|
@@ -117,8 +181,8 @@ with gr.Blocks() as iface:
|
|
| 117 |
|
| 118 |
submit_button.click(
|
| 119 |
process_input,
|
| 120 |
-
inputs=[input_type, input_text, model_name],
|
| 121 |
-
outputs=[analysis_output,text_output
|
| 122 |
)
|
| 123 |
|
| 124 |
if __name__ == "__main__":
|
|
|
|
| 4 |
from collections import Counter
|
| 5 |
import re
|
| 6 |
import plotly.graph_objs as go
|
| 7 |
+
import html
|
| 8 |
+
import random
|
| 9 |
+
import tiktoken
|
| 10 |
+
import anthropic
|
| 11 |
|
| 12 |
model_path = "models/"
|
| 13 |
|
| 14 |
# Available models
|
| 15 |
+
MODELS = ["Meta-Llama-3.1-8B", "gemma-2b", "gpt-3.5-turbo","gpt-4","gpt-4o" "Claude-3-Sonnet"]
|
| 16 |
+
openai_models = ["gpt-3.5-turbo","gpt-4","gpt-4o"]
|
| 17 |
+
# Color palette visible on both light and dark themes
|
| 18 |
+
COLOR_PALETTE = [
|
| 19 |
+
"#e6194B", "#3cb44b", "#ffe119", "#4363d8",
|
| 20 |
+
"#f58231", "#911eb4", "#42d4f4", "#f032e6",
|
| 21 |
+
"#bfef45", "#fabed4", "#469990", "#dcbeff",
|
| 22 |
+
"#9A6324", "#fffac8", "#800000", "#aaffc3",
|
| 23 |
+
"#808000", "#ffd8b1", "#000075", "#a9a9a9"
|
| 24 |
+
]
|
| 25 |
|
| 26 |
def create_vertical_histogram(data, title):
|
| 27 |
labels, values = zip(*data) if data else ([], [])
|
|
|
|
| 38 |
)
|
| 39 |
return fig
|
| 40 |
|
| 41 |
+
def process_text(text: str, model_name: str, api_key: str = None):
|
| 42 |
+
if model_name in ["Meta-Llama-3.1-8B", "gemma-2b"]:
|
| 43 |
+
tokenizer = AutoTokenizer.from_pretrained(model_path + model_name)
|
| 44 |
+
token_ids = tokenizer.encode(text, add_special_tokens=True)
|
| 45 |
+
tokens = tokenizer.convert_ids_to_tokens(token_ids)
|
| 46 |
+
elif model_name in openai_models:
|
| 47 |
+
encoding = tiktoken.encoding_for_model(model_name=model_name)
|
| 48 |
+
token_ids = encoding.encode(text)
|
| 49 |
+
tokens = [encoding.decode([id]) for id in token_ids]
|
| 50 |
+
elif model_name == "Claude-3-Sonnet":
|
| 51 |
+
if not api_key:
|
| 52 |
+
raise ValueError("API key is required for Claude models")
|
| 53 |
+
client = anthropic.Anthropic(api_key=api_key)
|
| 54 |
+
tokenizer = client.get_tokenizer()
|
| 55 |
+
token_ids = tokenizer.encode(text).ids
|
| 56 |
+
tokens = [tokenizer.decode([id]) for id in token_ids]
|
| 57 |
+
else:
|
| 58 |
+
raise ValueError(f"Unsupported model: {model_name}")
|
| 59 |
+
|
| 60 |
+
return text, tokens, token_ids
|
| 61 |
|
| 62 |
+
def process_ids(ids: str, model_name: str, api_key: str = None):
|
|
|
|
| 63 |
token_ids = ast.literal_eval(ids)
|
| 64 |
+
if model_name in ["Meta-Llama-3.1-8B", "gemma-2b"]:
|
| 65 |
+
tokenizer = AutoTokenizer.from_pretrained(model_path + model_name)
|
| 66 |
+
text = tokenizer.decode(token_ids)
|
| 67 |
+
tokens = tokenizer.convert_ids_to_tokens(token_ids)
|
| 68 |
+
elif model_name == openai_models:
|
| 69 |
+
encoding = tiktoken.encoding_for_model(model_name=model_name)
|
| 70 |
+
text = encoding.decode(token_ids)
|
| 71 |
+
tokens = [encoding.decode([id]) for id in token_ids]
|
| 72 |
+
elif model_name == "Claude-3-Sonnet":
|
| 73 |
+
client = anthropic.Anthropic(api_key=api_key)
|
| 74 |
+
tokenizer = client.get_tokenizer()
|
| 75 |
+
text = tokenizer.decode(token_ids)
|
| 76 |
+
tokens = [tokenizer.decode([id]) for id in token_ids]
|
| 77 |
+
else:
|
| 78 |
+
raise ValueError(f"Unsupported model: {model_name}")
|
| 79 |
+
|
| 80 |
+
return text, tokens, token_ids
|
| 81 |
|
| 82 |
+
def get_token_color(token, token_colors):
|
| 83 |
+
if token.startswith('<') and token.endswith('>'):
|
| 84 |
+
return "#42d4f4" # Cyan for special tokens
|
| 85 |
+
elif token == 'β' or token == ' ':
|
| 86 |
+
return "#3cb44b" # Green for space tokens
|
| 87 |
+
elif not token.isalnum():
|
| 88 |
+
return "#f032e6" # Magenta for special characters
|
| 89 |
+
else:
|
| 90 |
+
if token not in token_colors:
|
| 91 |
+
token_colors[token] = random.choice(COLOR_PALETTE)
|
| 92 |
+
return token_colors[token]
|
| 93 |
+
|
| 94 |
+
def create_html_tokens(tokens):
|
| 95 |
+
html_output = '<div style="font-family: monospace; border: 1px solid #ccc; padding: 10px; border-radius: 5px; background-color: #f9f9f9; white-space: pre-wrap; word-break: break-all;">'
|
| 96 |
+
token_colors = {}
|
| 97 |
+
for token in tokens:
|
| 98 |
+
color = get_token_color(token, token_colors)
|
| 99 |
+
escaped_token = html.escape(token)
|
| 100 |
+
html_output += f'<span style="background-color: {color}; color: black; padding: 2px 4px; margin: 1px; border-radius: 3px; display: inline-block;">{escaped_token}</span>'
|
| 101 |
+
html_output += '</div>'
|
| 102 |
+
return html_output
|
| 103 |
+
|
| 104 |
+
def process_input(input_type, input_value, model_name, api_key):
|
| 105 |
if input_type == "Text":
|
| 106 |
+
text, tokens, token_ids = process_text(text=input_value, model_name=model_name, api_key=api_key)
|
| 107 |
elif input_type == "Token IDs":
|
| 108 |
+
text, tokens, token_ids = process_ids(ids=input_value, model_name=model_name, api_key=api_key)
|
| 109 |
|
| 110 |
character_count = len(text)
|
| 111 |
word_count = len(text.split())
|
| 112 |
|
| 113 |
+
space_count = sum(1 for token in tokens if token in ['β', ' '])
|
| 114 |
+
special_char_count = sum(1 for token in tokens if not token.isalnum() and token not in ['β', ' '])
|
|
|
|
| 115 |
|
| 116 |
words = re.findall(r'\b\w+\b', text.lower())
|
| 117 |
special_chars = re.findall(r'[^\w\s]', text)
|
|
|
|
| 132 |
analysis += f"Special character tokens: {special_char_count}\n"
|
| 133 |
analysis += f"Other tokens: {len(tokens) - space_count - special_char_count}"
|
| 134 |
|
| 135 |
+
html_tokens = create_html_tokens(tokens)
|
| 136 |
+
|
| 137 |
+
return analysis, text, html_tokens, str(token_ids), words_hist, special_chars_hist, numbers_hist
|
| 138 |
|
| 139 |
def text_example():
|
| 140 |
return "Hello, world! This is an example text input for tokenization."
|
|
|
|
| 148 |
|
| 149 |
with gr.Row():
|
| 150 |
input_type = gr.Radio(["Text", "Token IDs"], label="Input Type", value="Text")
|
| 151 |
+
model_name = gr.Dropdown(choices=MODELS, label="Select Model", value=MODELS[0])
|
| 152 |
|
| 153 |
+
api_key = gr.Textbox(label="API Key Claude models)", type="password")
|
| 154 |
input_text = gr.Textbox(lines=5, label="Input")
|
| 155 |
|
| 156 |
with gr.Row():
|
|
|
|
| 160 |
submit_button = gr.Button("Process")
|
| 161 |
|
| 162 |
analysis_output = gr.Textbox(label="Analysis", lines=6)
|
| 163 |
+
text_output = gr.Textbox(label="Text", lines=6)
|
| 164 |
+
tokens_output = gr.HTML(label="Tokens")
|
| 165 |
token_ids_output = gr.Textbox(label="Token IDs", lines=2)
|
| 166 |
|
| 167 |
with gr.Row():
|
|
|
|
| 181 |
|
| 182 |
submit_button.click(
|
| 183 |
process_input,
|
| 184 |
+
inputs=[input_type, input_text, model_name, api_key],
|
| 185 |
+
outputs=[analysis_output, text_output, tokens_output, token_ids_output, words_plot, special_chars_plot, numbers_plot]
|
| 186 |
)
|
| 187 |
|
| 188 |
if __name__ == "__main__":
|
requirements.txt
CHANGED
|
@@ -1,11 +1,13 @@
|
|
| 1 |
aiofiles==23.2.1
|
| 2 |
annotated-types==0.7.0
|
|
|
|
| 3 |
anyio==4.4.0
|
| 4 |
certifi==2024.7.4
|
| 5 |
charset-normalizer==3.3.2
|
| 6 |
click==8.1.7
|
| 7 |
contourpy==1.2.1
|
| 8 |
cycler==0.12.1
|
|
|
|
| 9 |
fastapi==0.112.2
|
| 10 |
ffmpy==0.4.0
|
| 11 |
filelock==3.15.4
|
|
@@ -20,6 +22,7 @@ huggingface-hub==0.24.6
|
|
| 20 |
idna==3.8
|
| 21 |
importlib_resources==6.4.4
|
| 22 |
Jinja2==3.1.4
|
|
|
|
| 23 |
kiwisolver==1.4.5
|
| 24 |
markdown-it-py==3.0.0
|
| 25 |
MarkupSafe==2.1.5
|
|
@@ -51,6 +54,7 @@ six==1.16.0
|
|
| 51 |
sniffio==1.3.1
|
| 52 |
starlette==0.38.2
|
| 53 |
tenacity==9.0.0
|
|
|
|
| 54 |
tokenizers==0.19.1
|
| 55 |
tomlkit==0.12.0
|
| 56 |
tqdm==4.66.5
|
|
|
|
| 1 |
aiofiles==23.2.1
|
| 2 |
annotated-types==0.7.0
|
| 3 |
+
anthropic==0.34.1
|
| 4 |
anyio==4.4.0
|
| 5 |
certifi==2024.7.4
|
| 6 |
charset-normalizer==3.3.2
|
| 7 |
click==8.1.7
|
| 8 |
contourpy==1.2.1
|
| 9 |
cycler==0.12.1
|
| 10 |
+
distro==1.9.0
|
| 11 |
fastapi==0.112.2
|
| 12 |
ffmpy==0.4.0
|
| 13 |
filelock==3.15.4
|
|
|
|
| 22 |
idna==3.8
|
| 23 |
importlib_resources==6.4.4
|
| 24 |
Jinja2==3.1.4
|
| 25 |
+
jiter==0.5.0
|
| 26 |
kiwisolver==1.4.5
|
| 27 |
markdown-it-py==3.0.0
|
| 28 |
MarkupSafe==2.1.5
|
|
|
|
| 54 |
sniffio==1.3.1
|
| 55 |
starlette==0.38.2
|
| 56 |
tenacity==9.0.0
|
| 57 |
+
tiktoken==0.7.0
|
| 58 |
tokenizers==0.19.1
|
| 59 |
tomlkit==0.12.0
|
| 60 |
tqdm==4.66.5
|