lvwerra HF staff commited on
Commit
fee0cbb
·
1 Parent(s): 5bc0a7e

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +5 -54
app.py CHANGED
@@ -19,38 +19,6 @@ FIM_SUFFIX = "<fim_suffix>"
19
 
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
- ```
50
- code before<FILL_HERE>code after
51
- ```
52
- """
53
-
54
  theme = gr.themes.Monochrome(
55
  primary_hue="indigo",
56
  secondary_hue="blue",
@@ -284,19 +252,17 @@ css += share_btn_css + monospace_css + ".gradio-container {color: black}"
284
 
285
  description = """
286
  <div style="text-align: center;">
287
- <h1> TRL + TextEnvironment <span style='color: #e6b800;'>Models</span> Playground</h1>
 
288
  </div>
289
  <div style="text-align: left;">
290
  <p>This is a demo to generate text and code with the following StarCoderBase models:</p>
291
  <ul>
292
- <li><a href="https://huggingface.co/bigcode/starcoderplus" style='color: #e6b800;'>StarCoderBase TriviaQA</a>: A finetuned version of StarCoderBase on on the TriviaQA dataset using reinforcement learning via TRL's TextEnvironment (https://github.com/huggingface/trl/pull/424)</li>
293
- <li><a href="https://huggingface.co/bigcode/starcoderbase" style='color: #e6b800;'>StarCoderBase GSM8K</a>: A finetuned version of StarCoderBase on on the GSM8K dataset using reinforcement learning via TRL's TextEnvironment (https://github.com/huggingface/trl/pull/424).</li>
294
  </ul>
295
- <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>
296
  </div>
297
  """
298
- 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>\
299
- <br>**Intended Use**: this app and its [supporting model](https://huggingface.co/bigcode) are provided for demonstration purposes; not to serve as replacement for human expertise. For more details on the model's limitations in terms of factuality and biases, see the [model card.](hf.co/bigcode)"""
300
 
301
  with gr.Blocks(theme=theme, analytics_enabled=False, css=css) as demo:
302
  with gr.Column():
@@ -379,33 +345,18 @@ with gr.Blocks(theme=theme, analytics_enabled=False, css=css) as demo:
379
  )
380
 
381
  gr.Markdown(disclaimer)
382
- with gr.Group(elem_id="share-btn-container"):
383
- community_icon = gr.HTML(community_icon_html, visible=True)
384
- loading_icon = gr.HTML(loading_icon_html, visible=True)
385
- share_button = gr.Button(
386
- "Share to community", elem_id="share-btn", visible=True
387
- )
388
  gr_examples = gr.Examples(
389
  examples=[example for client in clients.values() for example in client[3]],
390
  inputs=[instruction],
391
  cache_examples=False,
392
  )
393
 
394
- # def update(version):
395
- # return clients[version][2],
396
-
397
- # version.select(
398
- # lambda x: (clients[x][2], clients[x][3]),
399
- # inputs=[version],
400
- # outputs=[system_prompt, gr_examples],
401
- # )
402
 
403
- # gr.Markdown(FORMATS)
404
 
405
  submit.click(
406
  generate,
407
  inputs=[instruction, system_prompt, version, temperature, max_new_tokens, top_p, repetition_penalty],
408
  outputs=[output, output2],
409
  )
410
- share_button.click(None, [], [], _js=share_js)
411
  demo.queue(concurrency_count=16).launch(debug=True)
 
19
 
20
  FIM_INDICATOR = "<FILL_HERE>"
21
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22
  theme = gr.themes.Monochrome(
23
  primary_hue="indigo",
24
  secondary_hue="blue",
 
252
 
253
  description = """
254
  <div style="text-align: center;">
255
+ <h1> TRL + TextEnvironment </h1>
256
+ <h2> Teaching Language Models to use tools. </h2>
257
  </div>
258
  <div style="text-align: left;">
259
  <p>This is a demo to generate text and code with the following StarCoderBase models:</p>
260
  <ul>
261
+ <li><a href="https://huggingface.co/vwxyzjn/starcoderbase-triviaqa" style='color: #e6b800;'>StarCoderBase TriviaQA</a>: A finetuned version of StarCoderBase on on the TriviaQA dataset using reinforcement learning via [TRL's TextEnvironment](https://github.com/huggingface/trl/pull/424)</li>
262
+ <li><a href="https://huggingface.co/lvwerra/starcoderbase-gsm8k" style='color: #e6b800;'>StarCoderBase GSM8K</a>: A finetuned version of StarCoderBase on on the GSM8K dataset using reinforcement learning via TRL's [TextEnvironment](https://github.com/huggingface/trl/pull/424).</li>
263
  </ul>
 
264
  </div>
265
  """
 
 
266
 
267
  with gr.Blocks(theme=theme, analytics_enabled=False, css=css) as demo:
268
  with gr.Column():
 
345
  )
346
 
347
  gr.Markdown(disclaimer)
348
+
 
 
 
 
 
349
  gr_examples = gr.Examples(
350
  examples=[example for client in clients.values() for example in client[3]],
351
  inputs=[instruction],
352
  cache_examples=False,
353
  )
354
 
 
 
 
 
 
 
 
 
355
 
 
356
 
357
  submit.click(
358
  generate,
359
  inputs=[instruction, system_prompt, version, temperature, max_new_tokens, top_p, repetition_penalty],
360
  outputs=[output, output2],
361
  )
 
362
  demo.queue(concurrency_count=16).launch(debug=True)