update_doc.py 6.4 KB
Newer Older
jixx's avatar
init  
jixx committed
1
2
3
4
5
6
7
import subprocess
import argparse
import ast
import json
import os

TEMPLATE = """
jixx's avatar
jixx committed
8
# Supported Models
jixx's avatar
init  
jixx committed
9

jixx's avatar
jixx committed
10
Text Generation Inference enables serving optimized models. The following sections list which models (VLMs & LLMs) are supported.
jixx's avatar
init  
jixx committed
11
12
13

SUPPORTED_MODELS

jixx's avatar
jixx committed
14

jixx's avatar
init  
jixx committed
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
If the above list lacks the model you would like to serve, depending on the model's pipeline type, you can try to initialize and serve the model anyways to see how well it performs, but performance isn't guaranteed for non-optimized models:

```python
# for causal LMs/text-generation models
AutoModelForCausalLM.from_pretrained(<model>, device_map="auto")`
# or, for text-to-text generation models
AutoModelForSeq2SeqLM.from_pretrained(<model>, device_map="auto")
```

If you wish to serve a supported model that already exists on a local folder, just point to the local folder.

```bash
text-generation-launcher --model-id <PATH-TO-LOCAL-BLOOM>
```
"""


def check_cli(check: bool):
    output = subprocess.check_output(["text-generation-launcher", "--help"]).decode(
        "utf-8"
    )

    wrap_code_blocks_flag = "<!-- WRAP CODE BLOCKS -->"
    final_doc = f"# Text-generation-launcher arguments\n\n{wrap_code_blocks_flag}\n\n"

    lines = output.split("\n")

    header = ""
    block = []
    for line in lines:
        if line.startswith("  -") or line.startswith("      -"):
            rendered_block = "\n".join(block)
            if header:
                final_doc += f"## {header}\n```shell\n{rendered_block}\n```\n"
            else:
                final_doc += f"```shell\n{rendered_block}\n```\n"
            block = []
            tokens = line.split("<")
            if len(tokens) > 1:
                header = tokens[-1][:-1]
            else:
                header = line.split("--")[-1]
            header = header.upper().replace("-", "_")

        block.append(line)

    rendered_block = "\n".join(block)
    final_doc += f"## {header}\n```shell\n{rendered_block}\n```\n"
    block = []

jixx's avatar
jixx committed
65
    filename = "docs/source/reference/launcher.md"
jixx's avatar
init  
jixx committed
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
    if check:
        with open(filename, "r") as f:
            doc = f.read()
            if doc != final_doc:
                tmp = "launcher.md"
                with open(tmp, "w") as g:
                    g.write(final_doc)
                diff = subprocess.run(
                    ["diff", tmp, filename], capture_output=True
                ).stdout.decode("utf-8")
                print(diff)
                raise Exception(
                    "Cli arguments Doc is not up-to-date, run `python update_doc.py` in order to update it"
                )
    else:
        with open(filename, "w") as f:
            f.write(final_doc)


def check_supported_models(check: bool):
    filename = "server/text_generation_server/models/__init__.py"
    with open(filename, "r") as f:
        tree = ast.parse(f.read())

    enum_def = [
        x for x in tree.body if isinstance(x, ast.ClassDef) and x.name == "ModelType"
    ][0]
    _locals = {}
    _globals = {}
    exec(f"import enum\n{ast.unparse(enum_def)}", _globals, _locals)
    ModelType = _locals["ModelType"]
    list_string = ""
    for data in ModelType:
        list_string += f"- [{data.value['name']}]({data.value['url']})"
        if data.value.get("multimodal", None):
            list_string += " (Multimodal)"
        list_string += "\n"

    final_doc = TEMPLATE.replace("SUPPORTED_MODELS", list_string)

    filename = "docs/source/supported_models.md"
    if check:
        with open(filename, "r") as f:
            doc = f.read()
            if doc != final_doc:
                tmp = "supported.md"
                with open(tmp, "w") as g:
                    g.write(final_doc)
                diff = subprocess.run(
                    ["diff", tmp, filename], capture_output=True
                ).stdout.decode("utf-8")
                print(diff)
                raise Exception(
                    "Supported models is not up-to-date, run `python update_doc.py` in order to update it"
                )
    else:
        with open(filename, "w") as f:
            f.write(final_doc)


def get_openapi_schema():
    try:
        output = subprocess.check_output(["text-generation-router", "print-schema"])
        return json.loads(output)
    except subprocess.CalledProcessError as e:
        print(f"Error running text-generation-router print-schema: {e}")
        raise SystemExit(1)
    except json.JSONDecodeError:
        print("Error: Invalid JSON received from text-generation-router print-schema")
        raise SystemExit(1)


def check_openapi(check: bool):
    new_openapi_data = get_openapi_schema()
    filename = "docs/openapi.json"
    tmp_filename = "openapi_tmp.json"

    with open(tmp_filename, "w") as f:
        json.dump(new_openapi_data, f, indent=2)

    if check:
        diff = subprocess.run(
            [
                "diff",
                # allow for trailing whitespace since it's not significant
                # and the precommit hook will remove it
                "--ignore-trailing-space",
                tmp_filename,
                filename,
            ],
            capture_output=True,
jixx's avatar
jixx committed
157
        ).stdout.decode("utf-8")
jixx's avatar
init  
jixx committed
158
159
160
161
162
163
164
165
166
167
168
        os.remove(tmp_filename)

        if diff:
            print(diff)
            raise Exception(
                "OpenAPI documentation is not up-to-date, run `python update_doc.py` in order to update it"
            )

    else:
        os.rename(tmp_filename, filename)
        print("OpenAPI documentation updated.")
jixx's avatar
jixx committed
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
    p = subprocess.run(
        [
            "redocly",
            # allow for trailing whitespace since it's not significant
            # and the precommit hook will remove it
            "lint",
            "--skip-rule",
            "security-defined",
            filename,
        ],
        capture_output=True,
    )
    errors = p.stderr.decode("utf-8")
    # The openapi specs fails on `exclusive_minimum` which is expected to be a boolean where
    # utoipa outputs a value instead: https://github.com/juhaku/utoipa/issues/969
    print(errors)
    if p.returncode != 0:
        print(errors)
        raise Exception(
            f"OpenAPI documentation is invalid, `redocly lint {filename}` showed some error:\n {errors}"
        )
    return True
jixx's avatar
init  
jixx committed
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("--check", action="store_true")

    args = parser.parse_args()

    check_cli(args.check)
    check_supported_models(args.check)
    check_openapi(args.check)


if __name__ == "__main__":
    main()