File size: 17,748 Bytes
c68892d
 
 
 
 
 
 
f59f99a
 
c68892d
f59f99a
 
 
 
 
c68892d
 
 
 
 
 
 
 
 
 
f59f99a
 
 
 
 
 
 
19e7e1f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
aa98be2
19e7e1f
 
 
 
 
 
 
 
 
 
 
20f48e5
19e7e1f
 
 
 
 
 
 
f59f99a
19e7e1f
20f48e5
19e7e1f
 
 
 
 
 
f59f99a
19e7e1f
 
 
 
 
 
 
f59f99a
19e7e1f
f59f99a
19e7e1f
 
 
 
 
 
 
aa98be2
19e7e1f
 
 
b26fc16
19e7e1f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c68892d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
009b11d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c68892d
19e7e1f
 
 
 
 
 
 
 
 
 
 
 
 
c68892d
19e7e1f
 
 
 
 
 
 
 
c68892d
19e7e1f
 
 
 
 
 
c68892d
19e7e1f
 
 
 
 
 
 
c68892d
19e7e1f
c68892d
19e7e1f
 
 
 
 
 
 
 
 
 
 
c68892d
19e7e1f
 
 
 
 
 
 
c68892d
19e7e1f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25d8974
19e7e1f
 
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
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
from pdfminer.high_level import extract_pages
from pdfminer.layout import LTTextContainer
from tqdm import tqdm
import re
import gradio as gr
import os
import accelerate
import spaces
import subprocess
from huggingface_hub import hf_hub_download
from llama_cpp import Llama
from llama_cpp_agent import LlamaCppAgent, MessagesFormatterType
from llama_cpp_agent.providers import LlamaCppPythonProvider
from llama_cpp_agent.chat_history import BasicChatHistory
from llama_cpp_agent.chat_history.messages import Roles

# subprocess.run('pip install llama-cpp-python==0.2.75 --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cu124', shell=True)
# subprocess.run('pip install llama-cpp-agent==0.2.10', shell=True)


# hf_hub_download(
#     repo_id="QuantFactory/Meta-Llama-3-8B-Instruct-GGUF",
#     filename="Meta-Llama-3-8B-Instruct.Q8_0.gguf",
#     local_dir = "./models"
# )

hf_hub_download(
    repo_id="bartowski/Meta-Llama-3-70B-Instruct-GGUF",
    filename="Meta-Llama-3-70B-Instruct-Q3_K_M.gguf",
    local_dir = "./models"
)

# def process_document(pdf_path, page_ids=None):
#     extracted_pages = extract_pages(pdf_path, page_numbers=page_ids)

#     page2content = {}

#     for extracted_page in tqdm(extracted_pages):
#         page_id = extracted_page.pageid
#         content = process_page(extracted_page)
#         page2content[page_id] = content

#     return page2content


# def process_page(extracted_page):
#     content = []
#     elements = [element for element in extracted_page._objs]
#     elements.sort(key=lambda a: a.y1, reverse=True)
#     for i, element in enumerate(elements):
#         if isinstance(element, LTTextContainer):
#             line_text = extract_text_and_normalize(element)
#             content.append(line_text)
#     content = re.sub('\n+', ' ', ''.join(content))
#     return content


# def extract_text_and_normalize(element):
#     # Extract text from line and split it with new lines
#     line_texts = element.get_text().split('\n')
#     norm_text = ''
#     for line_text in line_texts:
#         line_text = line_text.strip()
#         if not line_text:
#             line_text = '\n'
#         else:
#             line_text = re.sub('\s+', ' ', line_text)
#             if not re.search('[\w\d\,\-]', line_text[-1]):
#                 line_text += '\n'
#             else:
#                 line_text += ' '
#         norm_text += line_text
#     return norm_text


# def txt_to_html(text):
#     html_content = "<html><body>"
#     for line in text.split('\n'):
#         html_content += "<p>{}</p>".format(line.strip())
#     html_content += "</body></html>"
#     return html_content

# @spaces.GPU(duration=60)
# def deidentify_doc(pdftext, maxtokens, temperature, top_probability):
#     # prompt = "In the following text replace any person name and any address with term [redacted], replace any Date of Birth and NHS number with term [redacted]"
#     prompt = "Perform the following actions on given report: 1. Replace any person names and person date of birth and person age and person gender with term [redacted] 2. Replace any person addresses with term [redacted] 3. Replace any person age with term [redacted] 4. DO NOT REPLACE ANY MEDICAL MEASUREMENTS 5. Replace only the CALENDAR DATES of format 'day/month/year' with term [redacted]"
    
#     # model_id = "models/Meta-Llama-3-70B-Instruct-Q3_K_M.gguf"
#     # # model = Llama(model_path=model_id, n_ctx=2048, n_threads=8, n_gpu_layers=-1, n_batch=128)
#     # model = Llama(
#     #             model_path=model_id,
#     #             flash_attn=True,
#     #             n_gpu_layers=81,
#     #             n_batch=1024,
#     #             n_ctx=8192,
#     #             )

#     chat_template = MessagesFormatterType.LLAMA_3
    
#     llm = Llama(
#         model_path="models/Meta-Llama-3-70B-Instruct-Q3_K_M.gguf",
#         flash_attn=True,
#         n_gpu_layers=81,
#         n_batch=1024,
#         n_ctx=8192,
#     )

#     provider = LlamaCppPythonProvider(llm)
    
#     agent = LlamaCppAgent(
#         provider,
#         system_prompt="You are a helpful assistant.",
#         predefined_messages_formatter_type=chat_template,
#         debug_output=True
#     )
    
#     settings = provider.get_provider_default_settings()
#     settings.temperature = 0.7
#     settings.top_k = 40
#     settings.top_p = 0.95
#     settings.max_tokens = 2048
#     settings.repeat_penalty = 1.1
#     settings.stream = True

#     messages = BasicChatHistory()

#     stream = agent.get_chat_response(
#         prompt + ' : ' + pdftext,
#         llm_sampling_settings=settings,
#         chat_history=messages,
#         returns_streaming_generator=True,
#         print_output=False
#     )

#     outputs = ""
#     for output in stream:
#         outputs += output
    
#     return outputs

#     # output = model.create_chat_completion(
#     #     messages=[
#     #         {"role": "assistant", "content": prompt},
#     #         {
#     #             "role": "user",
#     #             "content": pdftext
#     #         }
#     #     ],
#     #     max_tokens=maxtokens,
#     #     temperature=temperature
#     # )
#     # output = output['choices'][0]['message']['content']

#     # prompt = "Perform the following actions on given text: 1. Replace any person age with term [redacted] 2. DO NOT REPLACE ANY MEDICAL MEASUREMENTS 3. Replace only the CALENDAR DATES of format 'day/month/year' with term [redacted]"
#     # output = model.create_chat_completion(
#     #     messages=[
#     #         {"role": "assistant", "content": prompt},
#     #         {
#     #             "role": "user",
#     #             "content": output
#     #         }
#     #     ],
#     #     max_tokens=maxtokens,
#     #     temperature=temperature
#     # )
#     # output = output['choices'][0]['message']['content']

#     # print(prompt)
#     # print(output)
#     # print('-------------------------------------------------------')

#     # return outputs

# def pdf_to_text(files, maxtokens=2048, temperature=0, top_probability=0.95):
#     print('Control 0-----------------------------------')
#     files=[files]#remove later
#     for file in files:
#         file_name = os.path.basename(file)
#         file_name_splt = file_name.split('.')
#         # print('File name is ', file_name)
#         if (len(file_name_splt) > 1 and file_name_splt[1] == 'pdf'):
#             page2content = process_document(file, page_ids=[0])
#             pdftext = page2content[1]
#             print(pdftext)
#         # pdftext = file # remove later
#             if (pdftext): #shift this if block to right later
#                 anonymized_text = deidentify_doc(pdftext, maxtokens, temperature, top_probability)
#     return anonymized_text


# css = ".gradio-container {background: 'logo.png'}"
# temp_slider = gr.Slider(minimum=0, maximum=2, value=0.9, label="Temperature Value")
# prob_slider = gr.Slider(minimum=0, maximum=1, value=0.95, label="Max Probability Value")
# max_tokens = gr.Number(value=600, label="Max Tokens")
# input_folder = gr.File(file_count='multiple')
# input_folder_text = gr.Textbox(label='Enter output folder path')
# output_text = gr.Textbox()
# output_path_component = gr.File(label="Select Output Path")
# iface = gr.Interface(
#     fn=pdf_to_text,
#     inputs='file',
#     # inputs=["textbox", input_folder_text, "textbox", max_tokens, temp_slider, prob_slider],
#     outputs=output_text,
#     title='COBIx Endoscopy Report De-Identification',
#     description="This application assists to remove personal information from the uploaded clinical report",
#     theme=gr.themes.Soft(),
# )
# iface.launch()

# import spaces
# import json
# import subprocess
# from llama_cpp import Llama
# from llama_cpp_agent import LlamaCppAgent, MessagesFormatterType
# from llama_cpp_agent.providers import LlamaCppPythonProvider
# from llama_cpp_agent.chat_history import BasicChatHistory
# from llama_cpp_agent.chat_history.messages import Roles
# import gradio as gr
# from huggingface_hub import hf_hub_download

# hf_hub_download(
#     repo_id="bartowski/Meta-Llama-3-70B-Instruct-GGUF",
#     filename="Meta-Llama-3-70B-Instruct-Q3_K_M.gguf",
#     local_dir = "./models"
# )
# # hf_hub_download(
# #     repo_id="bartowski/Mistral-7B-Instruct-v0.3-GGUF",
# #     filename="Mistral-7B-Instruct-v0.3-f32.gguf",
# #     local_dir = "./models"
# # )

css = """
.message-row {
    justify-content: space-evenly !important;
}
.message-bubble-border {
    border-radius: 6px !important;
}
.message-buttons-bot, .message-buttons-user {
    right: 10px !important;
    left: auto !important;
    bottom: 2px !important;
}
.dark.message-bubble-border {
    border-color: #343140 !important;
}
.dark.user {
    background: #1e1c26 !important;
}
.dark.assistant.dark, .dark.pending.dark {
    background: #16141c !important;
}
"""

def get_messages_formatter_type(model_name):
    if "Llama" in model_name:
        return MessagesFormatterType.LLAMA_3
    elif "Mistral" in model_name:
        return MessagesFormatterType.MISTRAL
    else:
        raise ValueError(f"Unsupported model: {model_name}")

@spaces.GPU(duration=60)
def respond(
    message,
    history: list[tuple[str, str]],
    model,
    system_message,
    max_tokens,
    temperature,
    top_p,
    top_k,
    repeat_penalty,
):
    chat_template = get_messages_formatter_type(model)

    llm = Llama(
        model_path=f"models/{model}",
        flash_attn=True,
        n_gpu_layers=81,
        n_batch=1024,
        n_ctx=8192,
    )
    provider = LlamaCppPythonProvider(llm)

    agent = LlamaCppAgent(
        provider,
        system_prompt=f"{system_message}",
        predefined_messages_formatter_type=chat_template,
        debug_output=True
    )
    
    settings = provider.get_provider_default_settings()
    settings.temperature = temperature
    settings.top_k = top_k
    settings.top_p = top_p
    settings.max_tokens = max_tokens
    settings.repeat_penalty = repeat_penalty
    settings.stream = True

    messages = BasicChatHistory()

    for msn in history:
        user = {
            'role': Roles.user,
            'content': msn[0]
        }
        assistant = {
            'role': Roles.assistant,
            'content': msn[1]
        }
        messages.add_message(user)
        messages.add_message(assistant)

    stream = agent.get_chat_response(
        message,
        llm_sampling_settings=settings,
        chat_history=messages,
        returns_streaming_generator=True,
        print_output=False
    )
    
    outputs = ""
    for output in stream:
        outputs += output
        yield outputs

PLACEHOLDER = """
<div class="message-bubble-border" style="display:flex; max-width: 600px; border-radius: 6px; border-width: 1px; box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1); backdrop-filter: blur(10px);">
    <figure style="margin: 0;">
        <img src="https://huggingface.co/spaces/pabloce/llama-cpp-agent/resolve/main/llama.jpg" alt="Logo" style="width: 100%; height: 100%; border-radius: 8px;">
    </figure>
    <div style="padding: .5rem 1.5rem;">
        <h2 style="text-align: left; font-size: 1.5rem; font-weight: 700; margin-bottom: 0.5rem;">llama-cpp-agent</h2>
        <p style="text-align: left; font-size: 16px; line-height: 1.5; margin-bottom: 15px;">The llama-cpp-agent framework simplifies interactions with Large Language Models (LLMs), providing an interface for chatting, executing function calls, generating structured output, performing retrieval augmented generation, and processing text using agentic chains with tools.</p>
        <div style="display: flex; justify-content: space-between; align-items: center;">
            <div style="display: flex; flex-flow: column; justify-content: space-between;">
                <span style="display: inline-flex; align-items: center; border-radius: 0.375rem; background-color: rgba(229, 70, 77, 0.1); padding: 0.1rem 0.75rem; font-size: 0.75rem; font-weight: 500; color: #f88181; margin-bottom: 2.5px;">
                    Mistral 7B Instruct v0.3
                </span>
                <span style="display: inline-flex; align-items: center; border-radius: 0.375rem; background-color: rgba(79, 70, 229, 0.1); padding: 0.1rem 0.75rem; font-size: 0.75rem; font-weight: 500; color: #60a5fa; margin-top: 2.5px;">
                    Meta Llama 3 70B Instruct
                </span>
            </div>
            <div style="display: flex; justify-content: flex-end; align-items: center;">
                <a href="https://discord.gg/sRMvWKrh" target="_blank" rel="noreferrer" style="padding: .5rem;">
                    <svg width="24" height="24" fill="currentColor" xmlns="http://www.w3.org/2000/svg" viewBox="0 5 30.67 23.25">
                        <title>Discord</title>
                        <path d="M26.0015 6.9529C24.0021 6.03845 21.8787 5.37198 19.6623 5C19.3833 5.48048 19.0733 6.13144 18.8563 6.64292C16.4989 6.30193 14.1585 6.30193 11.8336 6.64292C11.6166 6.13144 11.2911 5.48048 11.0276 5C8.79575 5.37198 6.67235 6.03845 4.6869 6.9529C0.672601 12.8736 -0.41235 18.6548 0.130124 24.3585C2.79599 26.2959 5.36889 27.4739 7.89682 28.2489C8.51679 27.4119 9.07477 26.5129 9.55525 25.5675C8.64079 25.2265 7.77283 24.808 6.93587 24.312C7.15286 24.1571 7.36986 23.9866 7.57135 23.8161C12.6241 26.1255 18.0969 26.1255 23.0876 23.8161C23.3046 23.9866 23.5061 24.1571 23.7231 24.312C22.8861 24.808 22.0182 25.2265 21.1037 25.5675C21.5842 26.5129 22.1422 27.4119 22.7621 28.2489C25.2885 27.4739 27.8769 26.2959 30.5288 24.3585C31.1952 17.7559 29.4733 12.0212 26.0015 6.9529ZM10.2527 20.8402C8.73376 20.8402 7.49382 19.4608 7.49382 17.7714C7.49382 16.082 8.70276 14.7025 10.2527 14.7025C11.7871 14.7025 13.0425 16.082 13.0115 17.7714C13.0115 19.4608 11.7871 20.8402 10.2527 20.8402ZM20.4373 20.8402C18.9183 20.8402 17.6768 19.4608 17.6768 17.7714C17.6768 16.082 18.8873 14.7025 20.4373 14.7025C21.9717 14.7025 23.2271 16.082 23.1961 17.7714C23.1961 19.4608 21.9872 20.8402 20.4373 20.8402Z"></path>
                    </svg>
                </a>
                <a href="https://github.com/Maximilian-Winter/llama-cpp-agent" target="_blank" rel="noreferrer" style="padding: .5rem;">
                    <svg width="24" height="24" fill="currentColor" viewBox="3 3 18 18">
                        <title>GitHub</title>
                        <path d="M12 3C7.0275 3 3 7.12937 3 12.2276C3 16.3109 5.57625 19.7597 9.15374 20.9824C9.60374 21.0631 9.77249 20.7863 9.77249 20.5441C9.77249 20.3249 9.76125 19.5982 9.76125 18.8254C7.5 19.2522 6.915 18.2602 6.735 17.7412C6.63375 17.4759 6.19499 16.6569 5.8125 16.4378C5.4975 16.2647 5.0475 15.838 5.80124 15.8264C6.51 15.8149 7.01625 16.4954 7.18499 16.7723C7.99499 18.1679 9.28875 17.7758 9.80625 17.5335C9.885 16.9337 10.1212 16.53 10.38 16.2993C8.3775 16.0687 6.285 15.2728 6.285 11.7432C6.285 10.7397 6.63375 9.9092 7.20749 9.26326C7.1175 9.03257 6.8025 8.08674 7.2975 6.81794C7.2975 6.81794 8.05125 6.57571 9.77249 7.76377C10.4925 7.55615 11.2575 7.45234 12.0225 7.45234C12.7875 7.45234 13.5525 7.55615 14.2725 7.76377C15.9937 6.56418 16.7475 6.81794 16.7475 6.81794C17.2424 8.08674 16.9275 9.03257 16.8375 9.26326C17.4113 9.9092 17.76 10.7281 17.76 11.7432C17.76 15.2843 15.6563 16.0687 13.6537 16.2993C13.98 16.5877 14.2613 17.1414 14.2613 18.0065C14.2613 19.2407 14.25 20.2326 14.25 20.5441C14.25 20.7863 14.4188 21.0746 14.8688 20.9824C16.6554 20.364 18.2079 19.1866 19.3078 17.6162C20.4077 16.0457 20.9995 14.1611 21 12.2276C21 7.12937 16.9725 3 12 3Z"></path>
                    </svg>
                </a>
            </div>
        </div>
    </div>
</div>
"""

demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Dropdown([
                'Meta-Llama-3-70B-Instruct-Q3_K_M.gguf',
                'Mistral-7B-Instruct-v0.3-f32.gguf'
            ],
            value="Meta-Llama-3-70B-Instruct-Q3_K_M.gguf",
            label="Model"
        ),
        gr.Textbox(value="You are a helpful assistant.", label="System message"),
        gr.Slider(minimum=1, maximum=4096, value=2048, step=1, label="Max tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p",
        ),
        gr.Slider(
            minimum=0,
            maximum=100,
            value=40,
            step=1,
            label="Top-k",
        ),
        gr.Slider(
            minimum=0.0,
            maximum=2.0,
            value=1.1,
            step=0.1,
            label="Repetition penalty",
        ),
    ],
    theme=gr.themes.Soft(primary_hue="violet", secondary_hue="violet", neutral_hue="gray",font=[gr.themes.GoogleFont("Exo"), "ui-sans-serif", "system-ui", "sans-serif"]).set(
        body_background_fill_dark="#16141c",
        block_background_fill_dark="#16141c",
        block_border_width="1px",
        block_title_background_fill_dark="#1e1c26",
        input_background_fill_dark="#292733",
        button_secondary_background_fill_dark="#24212b",
        border_color_accent_dark="#343140",
        border_color_primary_dark="#343140",
        background_fill_secondary_dark="#16141c",
        color_accent_soft_dark="transparent",
        code_background_fill_dark="#292733",
    ),
    css=css,
    retry_btn="Retry",
    undo_btn="Undo",
    clear_btn="Clear",
    submit_btn="Send",
    description="Llama-cpp-agent: Chat multi llm selection",
    chatbot=gr.Chatbot(
        scale=1, 
        placeholder=PLACEHOLDER,
        likeable=False,
        show_copy_button=True
    )
)

# if __name__ == "__main__":
demo.launch()