File size: 7,523 Bytes
7573abd dede9e1 7573abd 6e419a7 f510aac 7573abd dede9e1 7573abd e3716e2 6e419a7 7573abd ed57e85 7573abd c0eaee8 7573abd c0eaee8 7573abd c29119d 7573abd 05cc159 4206e00 36c12a2 7573abd a78ff20 f664b2f 7573abd c936a24 7573abd b38cc6c 7573abd 57659db 7573abd 57659db 7573abd c0eaee8 7573abd |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 |
import json
import os
import shutil
import requests
import gradio as gr
from huggingface_hub import Repository
from text_generation import Client
from share_btn import community_icon_html, loading_icon_html, share_js, share_btn_css
HF_TOKEN = os.environ.get("HF_TOKEN", None)
API_URL = "https://api-inference.huggingface.co/models/codellama/CodeLlama-13b-hf"
FIM_PREFIX = "<PRE> "
FIM_MIDDLE = " <MID>"
FIM_SUFFIX = " <SUF>"
FIM_INDICATOR = "<FILL_ME>"
EOS_STRING = "</s>"
EOT_STRING = "<EOT>"
theme = gr.themes.Monochrome(
primary_hue="indigo",
secondary_hue="blue",
neutral_hue="slate",
radius_size=gr.themes.sizes.radius_sm,
font=[
gr.themes.GoogleFont("Open Sans"),
"ui-sans-serif",
"system-ui",
"sans-serif",
],
)
client = Client(
API_URL,
headers={"Authorization": f"Bearer {HF_TOKEN}"},
)
def generate(
prompt, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,
):
temperature = float(temperature)
if temperature < 1e-2:
temperature = 1e-2
top_p = float(top_p)
fim_mode = False
generate_kwargs = dict(
temperature=temperature,
max_new_tokens=max_new_tokens,
top_p=top_p,
repetition_penalty=repetition_penalty,
do_sample=True,
seed=42,
)
if FIM_INDICATOR in prompt:
fim_mode = True
try:
prefix, suffix = prompt.split(FIM_INDICATOR)
except:
raise ValueError(f"Only one {FIM_INDICATOR} allowed in prompt!")
prompt = f"{FIM_PREFIX}{prefix}{FIM_SUFFIX}{suffix}{FIM_MIDDLE}"
stream = client.generate_stream(prompt, **generate_kwargs)
if fim_mode:
output = prefix
else:
output = prompt
previous_token = ""
for response in stream:
if any([end_token in response.token.text for end_token in [EOS_STRING, EOT_STRING]]):
if fim_mode:
output += suffix
yield output
return output
print("output", output)
else:
return output
else:
output += response.token.text
previous_token = response.token.text
yield output
return output
examples = [
"X_train, y_train, X_test, y_test = train_test_split(X, y, test_size=0.1)\n\n# Train a logistic regression model, predict the labels on the test set and compute the accuracy score",
"// Returns every other value in the array as a new array.\nfunction everyOther(arr) {",
"Poor English: She no went to the market. Corrected English:",
"def alternating(list1, list2):\n results = []\n for i in range(min(len(list1), len(list2))):\n results.append(list1[i])\n results.append(list2[i])\n if len(list1) > len(list2):\n <FILL_ME>\n else:\n results.extend(list2[i+1:])\n return results",
"def remove_non_ascii(s: str) -> str:\n \"\"\" <FILL_ME>\nprint(remove_non_ascii('afkdj$$('))",
]
def process_example(args):
for x in generate(args):
pass
return x
css = ".generating {visibility: hidden}"
monospace_css = """
#q-input textarea {
font-family: monospace, 'Consolas', Courier, monospace;
}
"""
css += share_btn_css + monospace_css + ".gradio-container {color: black}"
description = """
<div style="text-align: center;">
<h1> 🦙 Code Llama Playground</h1>
</div>
<div style="text-align: left;">
<p>This is a demo to generate text and code with the following <a href="https://huggingface.co/codellama/CodeLlama-13b-hf">Code Llama model (13B)</a>. Please note that this model is not designed for instruction purposes but for code completion. If you're looking for instruction or want to chat with a fine-tuned model, you can use [this demo instead](https://huggingface.co/spaces/codellama/codellama-13b-chat). You can learn more about the model in the <a href="https://huggingface.co/blog/codellama/">blog post<\a> or <a href="https://huggingface.co/papers/2308.12950">paper<\a></p>
</div>
"""
with gr.Blocks(theme=theme, analytics_enabled=False, css=css) as demo:
with gr.Column():
gr.Markdown(description)
with gr.Row():
with gr.Column():
instruction = gr.Textbox(
placeholder="Enter your code here",
lines=5,
label="Input",
elem_id="q-input",
)
submit = gr.Button("Generate", variant="primary")
output = gr.Code(elem_id="q-output", lines=30, label="Output")
with gr.Row():
with gr.Column():
with gr.Accordion("Advanced settings", open=False):
with gr.Row():
column_1, column_2 = gr.Column(), gr.Column()
with column_1:
temperature = gr.Slider(
label="Temperature",
value=0.1,
minimum=0.0,
maximum=1.0,
step=0.05,
interactive=True,
info="Higher values produce more diverse outputs",
)
max_new_tokens = gr.Slider(
label="Max new tokens",
value=256,
minimum=0,
maximum=8192,
step=64,
interactive=True,
info="The maximum numbers of new tokens",
)
with column_2:
top_p = gr.Slider(
label="Top-p (nucleus sampling)",
value=0.90,
minimum=0.0,
maximum=1,
step=0.05,
interactive=True,
info="Higher values sample more low-probability tokens",
)
repetition_penalty = gr.Slider(
label="Repetition penalty",
value=1.05,
minimum=1.0,
maximum=2.0,
step=0.05,
interactive=True,
info="Penalize repeated tokens",
)
gr.Examples(
examples=examples,
inputs=[instruction],
cache_examples=False,
fn=process_example,
outputs=[output],
)
submit.click(
generate,
inputs=[instruction, temperature, max_new_tokens, top_p, repetition_penalty],
outputs=[output],
)
demo.queue(concurrency_count=16).launch(debug=True) |