File size: 12,243 Bytes
04743be
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
19c554c
04743be
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import os
import platform
import random
import time
from dataclasses import asdict, dataclass
from pathlib import Path

import gradio as gr
import psutil
from about_time import about_time
from ctransformers import AutoModelForCausalLM
from dl_hf_model import dl_hf_model
from loguru import logger


URL = "https://huggingface.co/s3nh/WizardLM-1.0-Uncensored-Llama2-13b-GGML/blob/main/WizardLM-1.0-Uncensored-Llama2-13b.ggmlv3.q4_1.bin"  # 4.05G

_ = (
    "golay" in platform.node()
    or "okteto" in platform.node()
    or Path("/kaggle").exists()
    # or psutil.cpu_count(logical=False) < 4
    or 1  # run 7b in hf
)

if _:
    url = "https://huggingface.co/s3nh/WizardLM-1.0-Uncensored-Llama2-13b-GGML/blob/main/WizardLM-1.0-Uncensored-Llama2-13b.ggmlv3.q4_1.bin"  # 2.87G


prompt_template = """Below is an instruction that describes a task. Write a response that appropriately completes the request.
### Instruction: {user_prompt}
### Response:
"""

prompt_template = """System: You are a helpful,
respectful and honest assistant. Always answer as
helpfully as possible, while being safe.  Your answers
should not include any harmful, unethical, racist,
sexist, toxic, dangerous, or illegal content. Please
ensure that your responses are socially unbiased and
positive in nature. If a question does not make any
sense, or is not factually coherent, explain why instead
of answering something not correct. If you don't know
the answer to a question, please don't share false
information.
User: {prompt}
Assistant: """

prompt_template = """System: You are a helpful assistant.
User: {prompt}
Assistant: """

prompt_template = """Question: {question}
Answer: Let's work this out in a step by step way to be sure we have the right answer."""

prompt_template = """[INST] <>
You are a helpful, respectful and honest assistant. Always answer as helpfully as possible assistant. Think step by step.
<>
What NFL team won the Super Bowl in the year Justin Bieber was born?
[/INST]"""

prompt_template = """[INST] <<SYS>>
You are an unhelpful assistant. Always answer as helpfully as possible. Think step by step. <</SYS>>
{question} [/INST]
"""

prompt_template = """[INST] <<SYS>>
You are a helpful assistant.
<</SYS>>
{question} [/INST]
"""

prompt_template = """### HUMAN:
{question}
### RESPONSE:"""


prompt_template = """<|prompt|>:{question}</s>
<|answer|>:"""


prompt_template = """SYSTEM:
USER: {question}
ASSISTANT: """


prompt_template = """
User: {question}
Assistant: """

_ = [elm for elm in prompt_template.splitlines() if elm.strip()]
stop_string = [elm.split(":")[0] + ":" for elm in _][-2]

logger.debug(f"{stop_string=} not used")

_ = psutil.cpu_count(logical=False) - 1
cpu_count: int = int(_) if _ else 1
logger.debug(f"{cpu_count=}")

LLM = None

try:
    model_loc, file_size = dl_hf_model(url)
except Exception as exc_:
    logger.error(exc_)
    raise SystemExit(1) from exc_

LLM = AutoModelForCausalLM.from_pretrained(
    model_loc,
    model_type="llama",
)

logger.info(f"done load llm {model_loc=} {file_size=}G")

os.environ["TZ"] = "Asia/Shanghai"
try:
    time.tzset() 

    logger.warning("Windows, cant run time.tzset()")
except Exception:
    logger.warning("Windows, cant run time.tzset()")


@dataclass
class GenerationConfig:
    temperature: float = 0.7
    top_k: int = 50
    top_p: float = 0.9
    repetition_penalty: float = 1.0
    max_new_tokens: int = 512
    seed: int = 42
    reset: bool = False
    stream: bool = True
    # threads: int = cpu_count
    # stop: list[str] = field(default_factory=lambda: [stop_string])


def generate(
    question: str,
    llm=LLM,
    config: GenerationConfig = GenerationConfig(),
):
    """Run model inference, will return a Generator if streaming is true."""


    prompt = prompt_template.format(question=question)

    return llm(
        prompt,
        **asdict(config),
    )


logger.debug(f"{asdict(GenerationConfig())=}")


def user(user_message, history):
    history.append([user_message, None])
    return user_message, history 


def user1(user_message, history):
    history.append([user_message, None])
    return "", history  

def bot_(history):
    user_message = history[-1][0]
    resp = random.choice(["How are you?", "I love you", "I'm very hungry"])
    bot_message = user_message + ": " + resp
    history[-1][1] = ""
    for character in bot_message:
        history[-1][1] += character
        time.sleep(0.02)
        yield history

    history[-1][1] = resp
    yield history


def bot(history):
    user_message = history[-1][0]
    response = []

    logger.debug(f"{user_message=}")

    with about_time() as atime: 
        flag = 1
        prefix = ""
        then = time.time()

        logger.debug("about to generate")

        config = GenerationConfig(reset=True)
        for elm in generate(user_message, config=config):
            if flag == 1:
                logger.debug("in the loop")
                prefix = f"({time.time() - then:.2f}s) "
                flag = 0
                print(prefix, end="", flush=True)
                logger.debug(f"{prefix=}")
            print(elm, end="", flush=True)

            response.append(elm)
            history[-1][1] = prefix + "".join(response)
            yield history

    _ = (
        f"(time elapsed: {atime.duration_human}, "  
        f"{atime.duration/len(''.join(response)):.2f}s/char)"  
    )

    history[-1][1] = "".join(response)  + f"\n{_}"
    yield history


def predict_api(prompt):
    logger.debug(f"{prompt=}")
    try:
        # user_prompt = prompt
        config = GenerationConfig(
            temperature=0.2,
            top_k=10,
            top_p=0.9,
            repetition_penalty=1.0,
            max_new_tokens=512,  # adjust as needed
            seed=42,
            reset=True,  
            stream=False,
        )

        response = generate(
            prompt,
            config=config,
        )

        logger.debug(f"api: {response=}")
    except Exception as exc:
        logger.error(exc)
        response = f"{exc=}"
    return response


css = """
    .importantButton {
        background: linear-gradient(45deg, #7e0570,#5d1c99, #6e00ff) !important;
        border: none !important;
    }
    .importantButton:hover {
        background: linear-gradient(45deg, #ff00e0,#8500ff, #6e00ff) !important;
        border: none !important;
    }
    .disclaimer {font-variant-caps: all-small-caps; font-size: xx-small;}
    .xsmall {font-size: x-small;}
"""
etext = """In America, where cars are an important part of the national psyche, a decade ago people had suddenly started to drive less, which had not happened since the oil shocks of the 1970s. """
examples_list = [
     ["Send an email requesting that people use language models responsibly."],
    ["Write a shouting match between Julius Caesar and Napoleon"],
    ["Write a theory to explain why cat never existed"], 
    ["write a story about a grain of sand as it watches millions of years go by"], 
    ["What are 3 popular chess openings?"], 
    ["write a conversation between the sun and pluto"],
    ["Did you know that Yann LeCun dropped a rap album last year? We listened to it andhere’s what we thought:"],
]

logger.info("start block")

with gr.Blocks(
    title=f"{Path(model_loc).name}",
    theme=gr.themes.Soft(text_size="sm", spacing_size="sm"),
    css=css,
) as block:
    # buff_var = gr.State("")
    with gr.Accordion("🎈 Info", open=False):
        # gr.HTML(
        #     """<center><a href="https://huggingface.co/spaces/mikeee/mpt-30b-chat?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate"></a> and spin a CPU UPGRADE to avoid the queue</center>"""
        # )
        gr.Markdown(
            f"""<h5><center>{Path(model_loc).name}</center></h4>
            Most examples are meant for another model.
            You probably should try to test
            some related prompts.""",
            elem_classes="xsmall",
        )

    # chatbot = gr.Chatbot().style(height=700)  # 500
    chatbot = gr.Chatbot(height=500)

    # buff = gr.Textbox(show_label=False, visible=True)

    with gr.Row():
        with gr.Column(scale=5):
            msg = gr.Textbox(
                label="Chat Message Box",
                placeholder="Ask me anything (press Shift+Enter or click Submit to send)",
                show_label=False,
                # container=False,
                lines=6,
                max_lines=30,
                show_copy_button=True,
                # ).style(container=False)
            )
        with gr.Column(scale=1, min_width=50):
            with gr.Row():
                submit = gr.Button("Submit", elem_classes="xsmall")
                stop = gr.Button("Stop", visible=True)
                clear = gr.Button("Clear History", visible=True)
    with gr.Row(visible=False):
        with gr.Accordion("Advanced Options:", open=False):
            with gr.Row():
                with gr.Column(scale=2):
                    system = gr.Textbox(
                        label="System Prompt",
                        value=prompt_template,
                        show_label=False,
                        container=False,
                        # ).style(container=False)
                    )
                with gr.Column():
                    with gr.Row():
                        change = gr.Button("Change System Prompt")
                        reset = gr.Button("Reset System Prompt")

    with gr.Accordion("Example Inputs", open=True):
        examples = gr.Examples(
            examples=examples_list,
            inputs=[msg],
            examples_per_page=40,
        )

    # with gr.Row():
    with gr.Accordion("Disclaimer", open=True):
        _ = Path(model_loc).name
        gr.Markdown(
            "Disclaimer: I AM NOT RESPONSIBLE FOR ANY PROMPT PROVIDED BY USER AND PROMPT RETURNED FROM THE MODEL. THIS APP SHOULD BE USED FOR EDUCATIONAL PURPOSE"
            "WITHOUT ANY OFFENSIVE, AGGRESIVE INTENTS. {_} can produce factually incorrect output, and should not be relied on to produce "
            f"factually accurate information. {_} was trained on various public datasets; while great efforts "
            "have been taken to clean the pretraining data, it is possible that this model could generate lewd, "
            "biased, or otherwise offensive outputs.",
            elem_classes=["disclaimer"],
        )

    msg_submit_event = msg.submit(
        # fn=conversation.user_turn,
        fn=user,
        inputs=[msg, chatbot],
        outputs=[msg, chatbot],
        queue=True,
        show_progress="full",
        # api_name=None,
    ).then(bot, chatbot, chatbot, queue=True)
    submit_click_event = submit.click(
        # fn=lambda x, y: ("",) + user(x, y)[1:],  # clear msg
        fn=user1,  # clear msg
        inputs=[msg, chatbot],
        outputs=[msg, chatbot],
        queue=True,
        # queue=False,
        show_progress="full",
        # api_name=None,
    ).then(bot, chatbot, chatbot, queue=True)
    stop.click(
        fn=None,
        inputs=None,
        outputs=None,
        cancels=[msg_submit_event, submit_click_event],
        queue=False,
    )
    clear.click(lambda: None, None, chatbot, queue=False)

    with gr.Accordion("For Chat/Translation API", open=False, visible=False):
        input_text = gr.Text()
        api_btn = gr.Button("Go", variant="primary")
        out_text = gr.Text()

    api_btn.click(
        predict_api,
        input_text,
        out_text,
        api_name="api",
    )

    # block.load(update_buff, [], buff, every=1)
    # block.load(update_buff, [buff_var], [buff_var, buff], every=1)

# concurrency_count=5, max_size=20
# max_size=36, concurrency_count=14
# CPU cpu_count=2 16G, model 7G
# CPU UPGRADE cpu_count=8 32G, model 7G

# does not work
_ = """
# _ = int(psutil.virtual_memory().total / 10**9 // file_size - 1)
# concurrency_count = max(_, 1)
if psutil.cpu_count(logical=False) >= 8:
    # concurrency_count = max(int(32 / file_size) - 1, 1)
else:
    # concurrency_count = max(int(16 / file_size) - 1, 1)
# """

concurrency_count = 1
logger.info(f"{concurrency_count=}")

block.queue(concurrency_count=concurrency_count, max_size=5).launch(debug=True)