Spaces:
Build error
Build error
Update app.py
Browse files
app.py
CHANGED
@@ -10,8 +10,10 @@ from text_generation import Client
|
|
10 |
from share_btn import community_icon_html, loading_icon_html, share_js, share_btn_css
|
11 |
|
12 |
HF_TOKEN = os.environ.get("HF_TOKEN", None)
|
|
|
13 |
API_URL = "https://api-inference.huggingface.co/models/bigcode/starcoder"
|
14 |
API_URL_BASE ="https://api-inference.huggingface.co/models/bigcode/starcoderbase"
|
|
|
15 |
|
16 |
FIM_PREFIX = "<fim_prefix>"
|
17 |
FIM_MIDDLE = "<fim_middle>"
|
@@ -20,30 +22,39 @@ FIM_SUFFIX = "<fim_suffix>"
|
|
20 |
FIM_INDICATOR = "<FILL_HERE>"
|
21 |
|
22 |
FORMATS = """## Model Formats
|
|
|
23 |
The model is pretrained on code and is formatted with special tokens in addition to the pure code data,\
|
24 |
such as prefixes specifying the source of the file or tokens separating code from a commit message.\
|
25 |
Use these templates to explore the model's capacities:
|
|
|
26 |
### 1. Prefixes 🏷️
|
27 |
For pure code files, use any combination of the following prefixes:
|
|
|
28 |
```
|
29 |
<reponame>REPONAME<filename>FILENAME<gh_stars>STARS\ncode<|endoftext|>
|
30 |
```
|
31 |
STARS can be one of: 0, 1-10, 10-100, 100-1000, 1000+
|
|
|
32 |
### 2. Commits 💾
|
33 |
The commits data is formatted as follows:
|
|
|
34 |
```
|
35 |
<commit_before>code<commit_msg>text<commit_after>code<|endoftext|>
|
36 |
```
|
|
|
37 |
### 3. Jupyter Notebooks 📓
|
38 |
The model is trained on Jupyter notebooks as Python scripts and structured formats like:
|
|
|
39 |
```
|
40 |
<start_jupyter><jupyter_text>text<jupyter_code>code<jupyter_output>output<jupyter_text>
|
41 |
```
|
|
|
42 |
### 4. Issues 🐛
|
43 |
We also trained on GitHub issues using the following formatting:
|
44 |
```
|
45 |
<issue_start><issue_comment>text<issue_comment>...<issue_closed>
|
46 |
```
|
|
|
47 |
### 5. Fill-in-the-middle 🧩
|
48 |
Fill in the middle requires rearranging the model inputs. The playground handles this for you - all you need is to specify where to fill:
|
49 |
```
|
@@ -71,6 +82,9 @@ client = Client(
|
|
71 |
client_base = Client(
|
72 |
API_URL_BASE, headers={"Authorization": f"Bearer {HF_TOKEN}"},
|
73 |
)
|
|
|
|
|
|
|
74 |
|
75 |
def generate(
|
76 |
prompt, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0, version="StarCoder",
|
@@ -101,6 +115,8 @@ def generate(
|
|
101 |
|
102 |
if version == "StarCoder":
|
103 |
stream = client.generate_stream(prompt, **generate_kwargs)
|
|
|
|
|
104 |
else:
|
105 |
stream = client_base.generate_stream(prompt, **generate_kwargs)
|
106 |
|
@@ -126,6 +142,7 @@ def generate(
|
|
126 |
examples = [
|
127 |
"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",
|
128 |
"// Returns every other value in the array as a new array.\nfunction everyOther(arr) {",
|
|
|
129 |
"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_HERE>\n else:\n results.extend(list2[i+1:])\n return results",
|
130 |
]
|
131 |
|
@@ -150,9 +167,16 @@ css += share_btn_css + monospace_css + ".gradio-container {color: black}"
|
|
150 |
|
151 |
description = """
|
152 |
<div style="text-align: center;">
|
153 |
-
<h1>
|
154 |
-
|
155 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
156 |
</div>
|
157 |
"""
|
158 |
disclaimer = """⚠️<b>Any use or sharing of this demo constitues your acceptance of the BigCode [OpenRAIL-M](https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement) License Agreement and the use restrictions included within.</b>\
|
@@ -161,15 +185,23 @@ disclaimer = """⚠️<b>Any use or sharing of this demo constitues your accepta
|
|
161 |
with gr.Blocks(theme=theme, analytics_enabled=False, css=css) as demo:
|
162 |
with gr.Column():
|
163 |
gr.Markdown(description)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
164 |
with gr.Row():
|
165 |
with gr.Column():
|
166 |
instruction = gr.Textbox(
|
167 |
placeholder="Enter your code here",
|
168 |
-
|
|
|
169 |
elem_id="q-input",
|
170 |
)
|
171 |
submit = gr.Button("Generate", variant="primary")
|
172 |
-
output = gr.Code(elem_id="q-output", lines=30)
|
173 |
with gr.Row():
|
174 |
with gr.Column():
|
175 |
with gr.Accordion("Advanced settings", open=False):
|
@@ -213,13 +245,7 @@ with gr.Blocks(theme=theme, analytics_enabled=False, css=css) as demo:
|
|
213 |
interactive=True,
|
214 |
info="Penalize repeated tokens",
|
215 |
)
|
216 |
-
|
217 |
-
version = gr.Dropdown(
|
218 |
-
["StarCoderBase", "StarCoder"],
|
219 |
-
value="StarCoder",
|
220 |
-
label="Version",
|
221 |
-
info="",
|
222 |
-
)
|
223 |
gr.Markdown(disclaimer)
|
224 |
with gr.Group(elem_id="share-btn-container"):
|
225 |
community_icon = gr.HTML(community_icon_html, visible=True)
|
|
|
10 |
from share_btn import community_icon_html, loading_icon_html, share_js, share_btn_css
|
11 |
|
12 |
HF_TOKEN = os.environ.get("HF_TOKEN", None)
|
13 |
+
|
14 |
API_URL = "https://api-inference.huggingface.co/models/bigcode/starcoder"
|
15 |
API_URL_BASE ="https://api-inference.huggingface.co/models/bigcode/starcoderbase"
|
16 |
+
API_URL_PLUS = "https://api-inference.huggingface.co/models/bigcode/starcoderplus"
|
17 |
|
18 |
FIM_PREFIX = "<fim_prefix>"
|
19 |
FIM_MIDDLE = "<fim_middle>"
|
|
|
22 |
FIM_INDICATOR = "<FILL_HERE>"
|
23 |
|
24 |
FORMATS = """## Model Formats
|
25 |
+
|
26 |
The model is pretrained on code and is formatted with special tokens in addition to the pure code data,\
|
27 |
such as prefixes specifying the source of the file or tokens separating code from a commit message.\
|
28 |
Use these templates to explore the model's capacities:
|
29 |
+
|
30 |
### 1. Prefixes 🏷️
|
31 |
For pure code files, use any combination of the following prefixes:
|
32 |
+
|
33 |
```
|
34 |
<reponame>REPONAME<filename>FILENAME<gh_stars>STARS\ncode<|endoftext|>
|
35 |
```
|
36 |
STARS can be one of: 0, 1-10, 10-100, 100-1000, 1000+
|
37 |
+
|
38 |
### 2. Commits 💾
|
39 |
The commits data is formatted as follows:
|
40 |
+
|
41 |
```
|
42 |
<commit_before>code<commit_msg>text<commit_after>code<|endoftext|>
|
43 |
```
|
44 |
+
|
45 |
### 3. Jupyter Notebooks 📓
|
46 |
The model is trained on Jupyter notebooks as Python scripts and structured formats like:
|
47 |
+
|
48 |
```
|
49 |
<start_jupyter><jupyter_text>text<jupyter_code>code<jupyter_output>output<jupyter_text>
|
50 |
```
|
51 |
+
|
52 |
### 4. Issues 🐛
|
53 |
We also trained on GitHub issues using the following formatting:
|
54 |
```
|
55 |
<issue_start><issue_comment>text<issue_comment>...<issue_closed>
|
56 |
```
|
57 |
+
|
58 |
### 5. Fill-in-the-middle 🧩
|
59 |
Fill in the middle requires rearranging the model inputs. The playground handles this for you - all you need is to specify where to fill:
|
60 |
```
|
|
|
82 |
client_base = Client(
|
83 |
API_URL_BASE, headers={"Authorization": f"Bearer {HF_TOKEN}"},
|
84 |
)
|
85 |
+
client_plus = Client(
|
86 |
+
API_URL_PLUS, headers={"Authorization": f"Bearer {HF_TOKEN}"},
|
87 |
+
)
|
88 |
|
89 |
def generate(
|
90 |
prompt, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0, version="StarCoder",
|
|
|
115 |
|
116 |
if version == "StarCoder":
|
117 |
stream = client.generate_stream(prompt, **generate_kwargs)
|
118 |
+
elif version == "StarCoderPlus":
|
119 |
+
stream = client_plus.generate_stream(prompt, **generate_kwargs)
|
120 |
else:
|
121 |
stream = client_base.generate_stream(prompt, **generate_kwargs)
|
122 |
|
|
|
142 |
examples = [
|
143 |
"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",
|
144 |
"// Returns every other value in the array as a new array.\nfunction everyOther(arr) {",
|
145 |
+
"Poor English: She no went to the market. Corrected English:",
|
146 |
"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_HERE>\n else:\n results.extend(list2[i+1:])\n return results",
|
147 |
]
|
148 |
|
|
|
167 |
|
168 |
description = """
|
169 |
<div style="text-align: center;">
|
170 |
+
<h1> ⭐ StarCoder <span style='color: #e6b800;'>Models</span> Playground</h1>
|
171 |
+
</div>
|
172 |
+
<div style="text-align: left;">
|
173 |
+
<p>This is a demo to generate text and code with the following StarCoder models:</p>
|
174 |
+
<ul>
|
175 |
+
<li><a href="https://huggingface.co/bigcode/starcoderplus" style='color: #e6b800;'>StarCoderPlus</a>: A finetuned version of StarCoderBase on English web data, making it strong in both English text and code generation.</li>
|
176 |
+
<li><a href="https://huggingface.co/bigcode/starcoderbase" style='color: #e6b800;'>StarCoderBase</a>: A code generation model trained on 80+ programming languages, providing broad language coverage for code generation tasks.</li>
|
177 |
+
<li><a href="https://huggingface.co/bigcode/starcoder" style='color: #e6b800;'>StarCoder</a>: A finetuned version of StarCoderBase specifically focused on Python, while also maintaining strong performance on other programming languages.</li>
|
178 |
+
</ul>
|
179 |
+
<p><b>Please note:</b> These models are not designed for instruction purposes. If you're looking for instruction or want to chat with a fine-tuned model, you can visit the <a href="https://huggingface.co/spaces/HuggingFaceH4/starchat-playground">StarChat Playground</a>.</p>
|
180 |
</div>
|
181 |
"""
|
182 |
disclaimer = """⚠️<b>Any use or sharing of this demo constitues your acceptance of the BigCode [OpenRAIL-M](https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement) License Agreement and the use restrictions included within.</b>\
|
|
|
185 |
with gr.Blocks(theme=theme, analytics_enabled=False, css=css) as demo:
|
186 |
with gr.Column():
|
187 |
gr.Markdown(description)
|
188 |
+
with gr.Row():
|
189 |
+
version = gr.Dropdown(
|
190 |
+
["StarCoderPlus", "StarCoderBase", "StarCoder"],
|
191 |
+
value="StarCoderPlus",
|
192 |
+
label="Model",
|
193 |
+
info="Choose a model from the list",
|
194 |
+
)
|
195 |
with gr.Row():
|
196 |
with gr.Column():
|
197 |
instruction = gr.Textbox(
|
198 |
placeholder="Enter your code here",
|
199 |
+
lines=5,
|
200 |
+
label="Input",
|
201 |
elem_id="q-input",
|
202 |
)
|
203 |
submit = gr.Button("Generate", variant="primary")
|
204 |
+
output = gr.Code(elem_id="q-output", lines=30, label="Output")
|
205 |
with gr.Row():
|
206 |
with gr.Column():
|
207 |
with gr.Accordion("Advanced settings", open=False):
|
|
|
245 |
interactive=True,
|
246 |
info="Penalize repeated tokens",
|
247 |
)
|
248 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
249 |
gr.Markdown(disclaimer)
|
250 |
with gr.Group(elem_id="share-btn-container"):
|
251 |
community_icon = gr.HTML(community_icon_html, visible=True)
|