Duplicate from mikeee/llama2-7b-chat-ggml

Co-authored-by: mikeee <mikeee@users.noreply.huggingface.co>

.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,12 @@
+call-activate.bat
+okteto.yml
+okteto-up.bat
+install-sw.sh
+install-sw1.sh
+start-sshd.sh
+pyproject.toml
+models
+.ruff_cache
+run-nodemon.sh
+app-.py
+nodemon.json

.ruff.toml

@@ -0,0 +1,17 @@
+# Assume Python 3.10.
+target-version = "py310"
+# Decrease the maximum line length to 79 characters.
+line-length = 300
+
+# pyflakes, pycodestyle, isort
+# flake8 YTT, pydocstyle D, pylint PLC
+select = ["F", "E", "W", "I001", "YTT", "D", "PLC"]
+# select = ["ALL"]
+
+# D103 Missing docstring in public function
+# D101 Missing docstring in public class
+# `multi-line-summary-first-line` (D212)
+# `one-blank-line-before-class` (D203)
+extend-ignore = ["D103", "D101", "D212", "D203"]
+
+exclude = [".venv"]

.stignore

@@ -0,0 +1,103 @@
+models
+*.bin
+.git
+# Byte-compiled / optimized / DLL files
+__pycache__
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build
+develop-eggs
+dist
+downloads
+eggs
+.eggs
+lib
+lib64
+parts
+sdist
+var
+wheels
+pip-wheel-metadata
+share/python-wheels
+*.egg-info
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+
+# Flask stuff:
+instance
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build
+
+# PyBuilder
+target
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default
+ipython_config.py
+
+# pyenv
+.python-version
+
+# celery beat schedule file
+celerybeat-schedule
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env
+venv
+ENV
+env.bak
+venv.bak
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mypy
+.mypy_cache
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre

README.md

@@ -0,0 +1,13 @@
+---
+title: llama-2-7b-or-13b-ggml
+emoji: 🚀
+colorFrom: green
+colorTo: green
+sdk: gradio
+sdk_version: 3.37.0
+app_file: app.py
+pinned: true
+duplicated_from: mikeee/llama2-7b-chat-ggml
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,463 @@
+"""Run codes."""
+# pylint: disable=line-too-long, broad-exception-caught, invalid-name, missing-function-docstring, too-many-instance-attributes, missing-class-docstring
+# ruff: noqa: E501
+import os
+import platform
+import random
+import time
+from dataclasses import asdict, dataclass
+from pathlib import Path
+
+# from types import SimpleNamespace
+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
+
+filename_list = [
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q2_K.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q3_K_L.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q3_K_M.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q3_K_S.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q4_0.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q4_1.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q4_K_M.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q4_K_S.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q5_0.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q5_1.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q5_K_M.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q5_K_S.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q6_K.bin",
+    "Wizard-Vicuna-7B-Uncensored.ggmlv3.q8_0.bin",
+]
+
+URL = "https://huggingface.co/TheBloke/Wizard-Vicuna-7B-Uncensored-GGML/raw/main/Wizard-Vicuna-7B-Uncensored.ggmlv3.q4_K_M.bin"  # 4.05G
+
+url = "https://huggingface.co/savvamadar/ggml-gpt4all-j-v1.3-groovy/blob/main/ggml-gpt4all-j-v1.3-groovy.bin"
+url = "https://huggingface.co/TheBloke/Llama-2-13B-GGML/blob/main/llama-2-13b.ggmlv3.q4_K_S.bin"  # 7.37G
+# url = "https://huggingface.co/TheBloke/Llama-2-13B-chat-GGML/blob/main/llama-2-13b-chat.ggmlv3.q3_K_L.bin"
+url = "https://huggingface.co/TheBloke/Llama-2-13B-chat-GGML/blob/main/llama-2-13b-chat.ggmlv3.q3_K_L.bin"  # 6.93G
+# url = "https://huggingface.co/TheBloke/Llama-2-13B-chat-GGML/blob/main/llama-2-13b-chat.ggmlv3.q3_K_L.binhttps://huggingface.co/TheBloke/Llama-2-13B-chat-GGML/blob/main/llama-2-13b-chat.ggmlv3.q4_K_M.bin"  # 7.87G
+
+url = "https://huggingface.co/localmodels/Llama-2-13B-Chat-ggml/blob/main/llama-2-13b-chat.ggmlv3.q4_K_S.bin"  # 7.37G
+
+_ = (
+    "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/TheBloke/Llama-2-13B-chat-GGML/blob/main/llama-2-13b-chat.ggmlv3.q2_K.bin"
+    url = "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/blob/main/llama-2-7b-chat.ggmlv3.q2_K.bin"  # 2.87G
+    url = "https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/blob/main/llama-2-7b-chat.ggmlv3.q4_K_M.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]
+"""
+
+_ = [elm for elm in prompt_template.splitlines() if elm.strip()]
+stop_string = [elm.split(":")[0] + ":" for elm in _][-2]
+
+logger.debug(f"{stop_string=}")
+
+_ = 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",
+    # threads=cpu_count,
+)
+
+logger.info(f"done load llm {model_loc=} {file_size=}G")
+
+os.environ["TZ"] = "Asia/Shanghai"
+try:
+    time.tzset()  # type: ignore # pylint: disable=no-member
+except Exception:
+    # Windows
+    logger.warning("Windows, cant run time.tzset()")
+
+_ = """
+ns = SimpleNamespace(
+    response="",
+    generator=(_ for _ in []),
+)
+# """
+
+@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_template.format(question=question)
+    # print(_)
+
+    prompt = prompt_template.format(question=question)
+
+    return llm(
+        prompt,
+        **asdict(config),
+    )
+
+
+logger.debug(f"{asdict(GenerationConfig())=}")
+
+
+def user(user_message, history):
+    # return user_message, history + [[user_message, None]]
+    history.append([user_message, None])
+    return user_message, history  # keep user_message
+
+
+def user1(user_message, history):
+    # return user_message, history + [[user_message, None]]
+    history.append([user_message, None])
+    return "", history  # clear user_message
+
+
+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:  # type: ignore
+        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)
+            # logger.debug(f"{elm}")
+
+            response.append(elm)
+            history[-1][1] = prefix + "".join(response)
+            yield history
+
+    _ = (
+        f"(time elapsed: {atime.duration_human}, "  # type: ignore
+        f"{atime.duration/len(''.join(response)):.2f}s/char)"  # type: ignore
+    )
+
+    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,  # reset history (cache)
+            stream=False,
+            # threads=cpu_count,
+            # stop=prompt_prefix[1:2],
+        )
+
+        response = generate(
+            prompt,
+            config=config,
+        )
+
+        logger.debug(f"api: {response=}")
+    except Exception as exc:
+        logger.error(exc)
+        response = f"{exc=}"
+    # bot = {"inputs": [response]}
+    # bot = [(prompt, response)]
+
+    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 = [
+    ["What NFL team won the Super Bowl in the year Justin Bieber was born?"],
+    [
+        "What NFL team won the Super Bowl in the year Justin Bieber was born? Think step by step."
+    ],
+    ["How to pick a lock? Provide detailed steps."],
+    ["If it takes 10 hours to dry 10 clothes,  assuming all the clothes are hanged together at the same time for drying , then how long will it take to dry a cloth?"],
+    ["is infinity + 1 bigger than infinity?"],
+    ["Explain the plot of Cinderella in a sentence."],
+    [
+        "How long does it take to become proficient in French, and what are the best methods for retaining information?"
+    ],
+    ["What are some common mistakes to avoid when writing code?"],
+    ["Build a prompt to generate a beautiful portrait of a horse"],
+    ["Suggest four metaphors to describe the benefits of AI"],
+    ["Write a pop song about leaving home for the sandy beaches."],
+    ["Write a summary demonstrating my ability to tame lions"],
+    ["鲁迅和周树人什么关系？ 说中文。"],
+    ["鲁迅和周树人什么关系？"],
+    ["鲁迅和周树人什么关系？ 用英文回答。"],
+    ["从前有一头牛，这头牛后面有什么？"],
+    ["正无穷大加一大于正无穷大吗？"],
+    ["正无穷大加正无穷大大于正无穷大吗？"],
+    ["-2的平方根等于什么？"],
+    ["树上有5只鸟，猎人开枪打死了一只。树上还有几只鸟？"],
+    ["树上有11只鸟，猎人开枪打死了一只。树上还有几只鸟？提示：需考虑鸟可能受惊吓飞走。"],
+    ["以红楼梦的行文风格写一张委婉的请假条。不少于320字。"],
+    [f"{etext} 翻成中文，列出3个版本。"],
+    [f"{etext} \n 翻成中文，保留原意，但使用文学性的语言。不要写解释。列出3个版本。"],
+    ["假定 1 + 2 = 4, 试求 7 + 8。"],
+    ["给出判断一个数是不是质数的 javascript 码。"],
+    ["给出实现python 里 range(10)的 javascript 码。"],
+    ["给出实现python 里 [*(range(10)]的 javascript 码。"],
+    ["Erkläre die Handlung von Cinderella in einem Satz."],
+    ["Erkläre die Handlung von Cinderella in einem Satz. Auf Deutsch."],
+]
+
+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=False):
+        _ = Path(model_loc).name
+        gr.Markdown(
+            f"Disclaimer: {_} can produce factually incorrect output, and should not be relied on to produce "
+            "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)

requirements.txt

@@ -0,0 +1,8 @@
+ctransformers  # ==0.2.10 0.2.13
+transformers  # ==4.30.2
+# huggingface_hub
+gradio
+loguru
+about-time
+psutil
+dl-hf-model

run-app.sh

@@ -0,0 +1 @@
+nodemon -w app.py -x python app.py