File size: 5,966 Bytes
177e22a
c028b5a
 
4c8f20d
177e22a
880094b
177e22a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
880094b
 
 
 
 
 
 
 
 
 
 
177e22a
 
 
 
 
 
 
 
 
260ecb5
177e22a
260ecb5
177e22a
260ecb5
177e22a
260ecb5
 
 
 
 
 
 
 
 
b38ea7a
 
880094b
 
 
b38ea7a
 
880094b
b38ea7a
 
 
 
 
 
880094b
 
 
b38ea7a
 
880094b
b38ea7a
 
 
 
177e22a
260ecb5
177e22a
 
 
 
 
 
 
 
 
 
c028b5a
 
 
177e22a
 
 
 
 
c028b5a
 
 
177e22a
 
 
 
 
 
 
 
4cc80a6
177e22a
 
 
 
 
 
 
4cc80a6
 
 
177e22a
 
c028b5a
177e22a
c028b5a
 
 
 
 
177e22a
c028b5a
 
177e22a
 
 
c028b5a
177e22a
18845ec
177e22a
 
 
 
 
 
 
c028b5a
6c720bd
 
177e22a
6c720bd
81e7dde
c028b5a
 
 
 
 
 
 
 
 
 
 
 
177e22a
c028b5a
 
 
177e22a
c028b5a
 
 
177e22a
c028b5a
 
 
 
 
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
from typing import Any
import gradio as gr
import pandas as pd
import json
import requests
from html.parser import HTMLParser

quants = {
    "Q2_K": 3.35,
    "Q3_K_S": 3.5,
    "Q3_K_M": 3.91,
    "Q3_K_L": 4.27,
    "Q4_0": 4.55,
    "Q4_K_S": 4.58,
    "Q4_K_M": 4.85,
    "Q5_0": 5.54,
    "Q5_K_S": 5.54,
    "Q5_K_M": 5.69,
    "Q6_K": 6.59,
    "Q8_0": 8.5,
}

class SvelteHydratorExtractor(HTMLParser):
    def __init__(self):
        self.data = None
        super().__init__()

    def handle_starttag(self, tag, attrs):
        print("Start tag:", tag)
        for attr in attrs:
            if attr[0] == "data-props":
                self.data = attr[1].replace("&quot:", '"')


def calc_model_size(parameters: int, quant: float) -> int:
    return parameters * quant // 8


def get_model_config(hf_model: str) -> dict[str, Any]:
    config = requests.get(
        f"https://huggingface.co/{hf_model}/raw/main/config.json"
    ).json()
    model_size = 0
    try:
        model_size = requests.get(
            f"https://huggingface.co/{hf_model}/raw/main/model.safetensors.index.json"
        ).json()["metadta"]["total_size"]
    except:
        try:
            model_size = requests.get(
                f"https://huggingface.co/{hf_model}/raw/main/pytorch_model.bin.index.json"
            ).json()["metadta"]["total_size"]
        except:
            model_page = requests.get(
                f"https://huggingface.co/{hf_model}"
            ).text
            param_props_idx = model_page.find('data-target="ModelSafetensorsParams"')
            if param_props_idx != -1:
                param_props_start = model_page.rfind("<div", 0, param_props_idx)
                param_props_end = model_page.find(">", param_props_idx)
                extractor = SvelteHydratorExtractor()
                extractor.feed(model_page[param_props_start:param_props_end + 1])
                model_size = (
                    json.loads(
                        extractor.data
                    )["safetensors"]["total"]
                    * 2
                )
            else:
                param_props_idx = model_page.find('data-target="ModelHeader"')
                param_props_start = model_page.rfind("<div", 0, param_props_idx)
                param_props_end = model_page.find(">", param_props_idx)
                extractor = SvelteHydratorExtractor()
                extractor.feed(model_page[param_props_start:param_props_end + 1])
                model_size = (
                    json.loads(
                        extractor.data
                    )["model"]["safetensors"]["total"]
                    * 2
                )

    # assume fp16 weights
    config["parameters"] = model_size / 2
    return config


def calc_input_buffer_size(model_config, context: int) -> float:
    return 4096 + 2048 * model_config["hidden_size"] + context * 4 + context * 2048


def calc_compute_buffer_size(model_config, context: int) -> float:
    return (
        (context / 1024 * 2 + 0.75) * model_config["num_attention_heads"] * 1024 * 1024
    )


def calc_context_size(model_config, context: int) -> float:
    n_gqa = model_config["num_attention_heads"] / model_config["num_key_value_heads"]
    n_embd_gqa = model_config["hidden_size"] / n_gqa
    n_elements = n_embd_gqa * (model_config["num_hidden_layers"] * context)
    return 2 * n_elements * 2


def calc(model_base, context, quant_size):
    model_config = get_model_config(model_base)
    quant_bpw = 0
    try:
        quant_bpw = float(quant_size)
    except:
        quant_bpw = quants[quant_size]

    model_size = round(
        calc_model_size(model_config["parameters"], quant_bpw) / 1000 / 1000 / 1000, 2
    )
    context_size = round(
        (
            calc_input_buffer_size(model_config, context)
            + calc_context_size(model_config, context)
            + calc_compute_buffer_size(model_config, context)
        )
        / 1000
        / 1000
        / 1000,
        2,
    )

    return model_size, context_size, round(model_size + context_size, 2)


title = "GGUF VRAM Calculator"

with gr.Blocks(title=title, theme=gr.themes.Monochrome()) as app:
    default_model = "mistralai/Mistral-7B-v0.1"
    default_quant = "Q4_K_S"
    default_context = 8192
    default_size = calc(default_model, default_context, default_quant)
    default_model_size = default_size[0]
    default_context_size = default_size[1]

    gr.Markdown(
        f"# {app.title}\n## This space has been superseeded by the [NyxKrage/LLM-Model-VRAM-Calculator](https://huggingface.co/spaces/NyxKrage/LLM-Model-VRAM-Calculator), which has model search built in, and doesn't rely on gradio\nThis is meant only as a guide and is will not be 100% accurate, this also does not account for anything that might be running in the background on your system or CUDA system memory fallback on Windows"
    )
    model = gr.Textbox(
        value=default_model,
        label="Enter Unquantized HF Model Name (e.g. mistralai/Mistral-7B-v0.1)",
    )
    context = gr.Number(
        minimum=1, value=default_context, label="Desired Context Size (Tokens)"
    )
    quant = gr.Dropdown(
        choices=list(quants.keys()),
        value=default_quant,
        allow_custom_value=True,
        label="Enter GGUF Quant (e.g. Q4_K_S) or the specific BPW for other quantization schemes such as exl2 (e.g. 4.5)",
    )
    btn = gr.Button(value="Submit", variant="primary")
    btn.click(
        calc,
        inputs=[
            model,
            context,
            quant,
        ],
        outputs=[
            gr.Number(
                label="Model Size (GB)",
                value=default_size[0],
            ),
            gr.Number(
                label="Context Size (GB)",
                value=default_size[1],
            ),
            gr.Number(
                label="Total Size (GB)",
                value=default_size[2],
            ),
        ],
    )

    app.launch()