File size: 8,764 Bytes
5416372
a2e6c05
 
d81ed7c
a2e6c05
d81ed7c
a2e6c05
 
 
 
a65de5c
fd701cd
 
a2e6c05
 
 
 
 
 
 
5fe2c9a
 
3b39700
a2e6c05
5fe2c9a
63ba25e
a2e6c05
e976361
63ba25e
a2e6c05
d81ed7c
 
90e8d67
 
 
 
 
 
 
32f8e9c
90e8d67
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1a382ff
d81ed7c
 
 
 
 
 
e976361
 
a2e6c05
d81ed7c
a2e6c05
 
 
 
 
 
0a910e6
 
 
a2e6c05
 
e976361
 
a2e6c05
e976361
4993069
a2e6c05
 
d81ed7c
a2e6c05
 
776563f
a2e6c05
 
 
 
 
e976361
 
 
a2e6c05
e976361
a2e6c05
d81ed7c
a2e6c05
d81ed7c
a2e6c05
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d81ed7c
 
 
 
e976361
d81ed7c
 
 
 
 
 
e976361
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d81ed7c
3cae82a
d81ed7c
b23a519
 
 
88a7fc3
18e5a55
72fd759
18e5a55
3b39700
014d21e
 
a2e6c05
 
 
 
 
 
08a4d08
 
d81ed7c
 
 
a2e6c05
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
import spaces
import json
import subprocess
import gradio as gr
from huggingface_hub import hf_hub_download

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="bartowski/dolphin-2.9.1-yi-1.5-34b-GGUF", filename="dolphin-2.9.1-yi-1.5-34b-Q6_K.gguf",  local_dir = "./models")
hf_hub_download(repo_id="crusoeai/dolphin-2.9.1-llama-3-70b-GGUF", filename="dolphin-2.9.1-llama-3-70b.Q3_K_M.gguf",  local_dir = "./models")
# hf_hub_download(repo_id="bartowski/dolphin-2.9.1-yi-1.5-9b-GGUF", filename="dolphin-2.9.1-yi-1.5-9b-f32.gguf",  local_dir = "./models")
# hf_hub_download(repo_id="crusoeai/dolphin-2.9.1-llama-3-8b-GGUF", filename="dolphin-2.9.1-llama-3-8b.Q6_K.gguf",  local_dir = "./models")

css = """
.message-row {
    justify-content: space-evenly !important;
}
.message-bubble-border {
    border-radius: 6px !important;
}
.dark.message-bubble-border {
    border-color: #21293b !important;
}
.dark.user {
    background: #0a1120 !important;
}
.dark.assistant {
    background: transparent !important;
}
"""

PLACEHOLDER = """
<div class="message-bubble-border" style="display:flex; max-width: 600px; border-radius: 8px; 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/cognitivecomputations/chat/resolve/main/image.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;">Cognitive Computations</h2>
        <p style="text-align: left; font-size: 16px; line-height: 1.5; margin-bottom: 15px;">We are dedicated to embracing innovation, exploring cutting-edge ideas, and fearlessly pushing the boundaries of what is possible in the realm of artificial intelligence.</p>
        <div style="display: flex; justify-content: flex-end;">
            <a href="https://discord.gg/cognitivecomputations" 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/cognitivecomputations" 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>
"""

@spaces.GPU(duration=120)
def respond(
    message,
    history: list[tuple[str, str]],
    max_tokens,
    temperature,
    top_p,
    top_k,
    repeat_penalty,
    model,
):
    from llama_cpp import Llama
    from llama_cpp_agent import LlamaCppAgent
    from llama_cpp_agent import 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
    print(message)
    print(history)
    
    llm = Llama(
        model_path=f"models/{model}",
        flash_attn=True,
        n_threads=40,
        n_gpu_layers=81,
        n_batch=1024,
        n_ctx=8192,
    )
    provider = LlamaCppPythonProvider(llm)

    agent = LlamaCppAgent(
        provider,
        system_prompt="You are Dolphin an AI assistant that helps humanity.",
        predefined_messages_formatter_type=MessagesFormatterType.CHATML,
        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

demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Slider(minimum=1, maximum=8192, value=8192, 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",
        ),
        gr.Dropdown(['dolphin-2.9.1-yi-1.5-34b-Q6_K.gguf', 'dolphin-2.9.1-llama-3-70b.Q3_K_M.gguf'], value="dolphin-2.9.1-llama-3-70b.Q3_K_M.gguf", label="Model"),
    ],
    theme=gr.themes.Soft(primary_hue="indigo", secondary_hue="blue", neutral_hue="gray",font=[gr.themes.GoogleFont("Exo"), "ui-sans-serif", "system-ui", "sans-serif"]).set(
        body_background_fill_dark="#0f172a",
        block_background_fill_dark="#0f172a",
        block_border_width="1px",
        block_title_background_fill_dark="#070d1b",
        input_background_fill_dark="#0c1425",
        button_secondary_background_fill_dark="#070d1b",
        border_color_primary_dark="#21293b",
        background_fill_secondary_dark="#0f172a",
        color_accent_soft_dark="transparent"
    ),
    css=css,
    retry_btn="Retry",
    undo_btn="Undo",
    clear_btn="Clear",
    submit_btn="Send",
    description="Cognitive Computation: Chat Dolphin 🐬 2.9.1-llama-3-70b & 2.9.1-yi-1.5-34b",
    chatbot=gr.Chatbot(scale=1, placeholder=PLACEHOLDER)
)

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