File size: 16,162 Bytes
0eeee8c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import argparse
import glob
import hashlib
import os
import platform
import re
import site
import subprocess
import sys

script_dir = os.getcwd()
conda_env_path = os.path.join(script_dir, "installer_files", "env")

# Remove the '# ' from the following lines as needed for your AMD GPU on Linux
# os.environ["ROCM_PATH"] = '/opt/rocm'
# os.environ["HSA_OVERRIDE_GFX_VERSION"] = '10.3.0'
# os.environ["HCC_AMDGPU_TARGET"] = 'gfx1030'

# Command-line flags
cmd_flags_path = os.path.join(script_dir, "CMD_FLAGS.txt")
if os.path.exists(cmd_flags_path):
    with open(cmd_flags_path, 'r') as f:
        CMD_FLAGS = ' '.join(line.strip() for line in f if line.strip() and not line.strip().startswith('#'))
else:
    CMD_FLAGS = ''

flags = f"{' '.join([flag for flag in sys.argv[1:] if flag != '--update'])} {CMD_FLAGS}"


def is_linux():
    return sys.platform.startswith("linux")


def is_windows():
    return sys.platform.startswith("win")


def is_macos():
    return sys.platform.startswith("darwin")


def is_x86_64():
    return platform.machine() == "x86_64"


def cpu_has_avx2():
    try:
        import cpuinfo

        info = cpuinfo.get_cpu_info()
        if 'avx2' in info['flags']:
            return True
        else:
            return False
    except:
        return True


def cpu_has_amx():
    try:
        import cpuinfo

        info = cpuinfo.get_cpu_info()
        if 'amx' in info['flags']:
            return True
        else:
            return False
    except:
        return True


def torch_version():
    site_packages_path = None
    for sitedir in site.getsitepackages():
        if "site-packages" in sitedir and conda_env_path in sitedir:
            site_packages_path = sitedir
            break

    if site_packages_path:
        torch_version_file = open(os.path.join(site_packages_path, 'torch', 'version.py')).read().splitlines()
        torver = [line for line in torch_version_file if '__version__' in line][0].split('__version__ = ')[1].strip("'")
    else:
        from torch import __version__ as torver
    return torver


def is_installed():
    site_packages_path = None
    for sitedir in site.getsitepackages():
        if "site-packages" in sitedir and conda_env_path in sitedir:
            site_packages_path = sitedir
            break

    if site_packages_path:
        return os.path.isfile(os.path.join(site_packages_path, 'torch', '__init__.py'))
    else:
        return os.path.isdir(conda_env_path)


def check_env():
    # If we have access to conda, we are probably in an environment
    conda_exist = run_cmd("conda", environment=True, capture_output=True).returncode == 0
    if not conda_exist:
        print("Conda is not installed. Exiting...")
        sys.exit(1)

    # Ensure this is a new environment and not the base environment
    if os.environ["CONDA_DEFAULT_ENV"] == "base":
        print("Create an environment for this project and activate it. Exiting...")
        sys.exit(1)


def clear_cache():
    run_cmd("conda clean -a -y", environment=True)
    run_cmd("python -m pip cache purge", environment=True)


def print_big_message(message):
    message = message.strip()
    lines = message.split('\n')
    print("\n\n*******************************************************************")
    for line in lines:
        if line.strip() != '':
            print("*", line)

    print("*******************************************************************\n\n")


def calculate_file_hash(file_path):
    p = os.path.join(script_dir, file_path)
    if os.path.isfile(p):
        with open(p, 'rb') as f:
            return hashlib.sha256(f.read()).hexdigest()
    else:
        return ''


def run_cmd(cmd, assert_success=False, environment=False, capture_output=False, env=None):
    # Use the conda environment
    if environment:
        if is_windows():
            conda_bat_path = os.path.join(script_dir, "installer_files", "conda", "condabin", "conda.bat")
            cmd = "\"" + conda_bat_path + "\" activate \"" + conda_env_path + "\" >nul && " + cmd
        else:
            conda_sh_path = os.path.join(script_dir, "installer_files", "conda", "etc", "profile.d", "conda.sh")
            cmd = ". \"" + conda_sh_path + "\" && conda activate \"" + conda_env_path + "\" && " + cmd

    # Run shell commands
    result = subprocess.run(cmd, shell=True, capture_output=capture_output, env=env)

    # Assert the command ran successfully
    if assert_success and result.returncode != 0:
        print("Command '" + cmd + "' failed with exit status code '" + str(result.returncode) + "'.\n\nExiting now.\nTry running the start/update script again.")
        sys.exit(1)

    return result


def install_webui():
    # Select your GPU, or choose to run in CPU mode
    if "GPU_CHOICE" in os.environ:
        choice = os.environ["GPU_CHOICE"].upper()
        print_big_message(f"Selected GPU choice \"{choice}\" based on the GPU_CHOICE environment variable.")
    else:
        print()
        print("What is your GPU?")
        print()
        print("A) NVIDIA")
        print("B) AMD (Linux/MacOS only. Requires ROCm SDK 5.6 on Linux)")
        print("C) Apple M Series")
        print("D) Intel Arc (IPEX)")
        print("N) None (I want to run models in CPU mode)")
        print()

        choice = input("Input> ").upper()
        while choice not in 'ABCDN':
            print("Invalid choice. Please try again.")
            choice = input("Input> ").upper()

    if choice == "N":
        print_big_message("Once the installation ends, make sure to open CMD_FLAGS.txt with\na text editor and add the --cpu flag.")

    # Find the proper Pytorch installation command
    install_git = "conda install -y -k ninja git"
    install_pytorch = "python -m pip install torch torchvision torchaudio"

    use_cuda118 = "N"
    if any((is_windows(), is_linux())) and choice == "A":
        if "USE_CUDA118" in os.environ:
            use_cuda118 = "Y" if os.environ.get("USE_CUDA118", "").lower() in ("yes", "y", "true", "1", "t", "on") else "N"
        else:
            # Ask for CUDA version if using NVIDIA
            print("\nWould you like to use CUDA 11.8 instead of 12.1? This is only necessary for older GPUs like Kepler.\nIf unsure, say \"N\".\n")
            use_cuda118 = input("Input (Y/N)> ").upper().strip('"\'').strip()
            while use_cuda118 not in 'YN':
                print("Invalid choice. Please try again.")
                use_cuda118 = input("Input> ").upper().strip('"\'').strip()
        if use_cuda118 == 'Y':
            print("CUDA: 11.8")
        else:
            print("CUDA: 12.1")

        install_pytorch = f"python -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/{'cu121' if use_cuda118 == 'N' else 'cu118'}"
    elif not is_macos() and choice == "B":
        if is_linux():
            install_pytorch = "python -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.6"
        else:
            print("AMD GPUs are only supported on Linux. Exiting...")
            sys.exit(1)
    elif is_linux() and (choice == "C" or choice == "N"):
        install_pytorch = "python -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu"
    elif choice == "D":
        install_pytorch = "python -m pip install torch==2.0.1a0 torchvision==0.15.2a0 intel_extension_for_pytorch==2.0.110+xpu -f https://developer.intel.com/ipex-whl-stable-xpu"

    # Install Git and then Pytorch
    run_cmd(f"{install_git} && {install_pytorch} && python -m pip install py-cpuinfo==9.0.0", assert_success=True, environment=True)

    # Install CUDA libraries (this wasn't necessary for Pytorch before...)
    if choice == "A":
        run_cmd(f"conda install -y -c \"nvidia/label/{'cuda-12.1.1' if use_cuda118 == 'N' else 'cuda-11.8.0'}\" cuda-runtime", assert_success=True, environment=True)

    # Install the webui requirements
    update_requirements(initial_installation=True)


def update_requirements(initial_installation=False):
    # Create .git directory if missing
    if not os.path.isdir(os.path.join(script_dir, ".git")):
        git_creation_cmd = 'git init -b main && git remote add origin https://github.com/oobabooga/text-generation-webui && git fetch && git symbolic-ref refs/remotes/origin/HEAD refs/remotes/origin/main && git reset --hard origin/main && git branch --set-upstream-to=origin/main'
        run_cmd(git_creation_cmd, environment=True, assert_success=True)

    files_to_check = [
        'start_linux.sh', 'start_macos.sh', 'start_windows.bat', 'start_wsl.bat',
        'update_linux.sh', 'update_macos.sh', 'update_windows.bat', 'update_wsl.bat',
        'one_click.py'
    ]

    before_pull_hashes = {file_name: calculate_file_hash(file_name) for file_name in files_to_check}
    run_cmd("git pull --autostash", assert_success=True, environment=True)
    after_pull_hashes = {file_name: calculate_file_hash(file_name) for file_name in files_to_check}

    # Check for differences in installation file hashes
    for file_name in files_to_check:
        if before_pull_hashes[file_name] != after_pull_hashes[file_name]:
            print(f"File '{file_name}' was updated during 'git pull'. Please run the script again.")
            exit(1)

    # Extensions requirements are installed only during the initial install by default.
    # That can be changed with the INSTALL_EXTENSIONS environment variable.
    install = initial_installation
    if "INSTALL_EXTENSIONS" in os.environ:
        install = os.environ["INSTALL_EXTENSIONS"].lower() in ("yes", "y", "true", "1", "t", "on")

    if install:
        print_big_message("Installing extensions requirements.")
        extensions = next(os.walk("extensions"))[1]
        for extension in extensions:
            if extension in ['superbooga', 'superboogav2']:  # No wheels available for requirements
                continue

            extension_req_path = os.path.join("extensions", extension, "requirements.txt")
            if os.path.exists(extension_req_path):
                run_cmd("python -m pip install -r " + extension_req_path + " --upgrade", assert_success=True, environment=True)
    elif initial_installation:
        print_big_message("Will not install extensions due to INSTALL_EXTENSIONS environment variable.")

    # Detect the Python and PyTorch versions
    torver = torch_version()
    print(f"TORCH: {torver}")
    is_cuda = '+cu' in torver
    is_cuda118 = '+cu118' in torver  # 2.1.0+cu118
    is_cuda117 = '+cu117' in torver  # 2.0.1+cu117
    is_rocm = '+rocm' in torver  # 2.0.1+rocm5.4.2
    is_intel = '+cxx11' in torver  # 2.0.1a0+cxx11.abi
    is_cpu = '+cpu' in torver  # 2.0.1+cpu

    if is_rocm:
        if cpu_has_avx2():
            requirements_file = "requirements_amd.txt"
        else:
            requirements_file = "requirements_amd_noavx2.txt"
    elif is_cpu:
        if cpu_has_avx2():
            requirements_file = "requirements_cpu_only.txt"
        else:
            requirements_file = "requirements_cpu_only_noavx2.txt"
    elif is_macos():
        if is_x86_64():
            requirements_file = "requirements_apple_intel.txt"
        else:
            requirements_file = "requirements_apple_silicon.txt"
    else:
        if cpu_has_avx2():
            requirements_file = "requirements.txt"
        else:
            requirements_file = "requirements_noavx2.txt"

    # Prepare the requirements file
    print_big_message(f"Installing webui requirements from file: {requirements_file}")
    textgen_requirements = open(requirements_file).read().splitlines()
    if is_cuda117:
        textgen_requirements = [req.replace('+cu121', '+cu117').replace('+cu122', '+cu117').replace('torch2.1', 'torch2.0') for req in textgen_requirements]
    elif is_cuda118:
        textgen_requirements = [req.replace('+cu121', '+cu118').replace('+cu122', '+cu118') for req in textgen_requirements]
    if is_windows() and (is_cuda117 or is_cuda118):  # No flash-attention on Windows for CUDA 11
        textgen_requirements = [req for req in textgen_requirements if 'bdashore3/flash-attention' not in req]

    with open('temp_requirements.txt', 'w') as file:
        file.write('\n'.join(textgen_requirements))

    # Workaround for git+ packages not updating properly.
    git_requirements = [req for req in textgen_requirements if req.startswith("git+")]
    for req in git_requirements:
        url = req.replace("git+", "")
        package_name = url.split("/")[-1].split("@")[0].rstrip(".git")
        run_cmd("python -m pip uninstall -y " + package_name, environment=True)
        print(f"Uninstalled {package_name}")

    # Install/update the project requirements
    run_cmd("python -m pip install -r temp_requirements.txt --upgrade", assert_success=True, environment=True)
    os.remove('temp_requirements.txt')

    # Check for '+cu' or '+rocm' in version string to determine if torch uses CUDA or ROCm. Check for pytorch-cuda as well for backwards compatibility
    if not any((is_cuda, is_rocm)) and run_cmd("conda list -f pytorch-cuda | grep pytorch-cuda", environment=True, capture_output=True).returncode == 1:
        clear_cache()
        return

    if not os.path.exists("repositories/"):
        os.mkdir("repositories")

    os.chdir("repositories")

    # Install or update ExLlama as needed
    if not os.path.exists("exllama/"):
        run_cmd("git clone https://github.com/turboderp/exllama.git", environment=True)
    else:
        os.chdir("exllama")
        run_cmd("git pull", environment=True)
        os.chdir("..")

    if is_linux():
        # Fix JIT compile issue with ExLlama in Linux/WSL
        if not os.path.exists(f"{conda_env_path}/lib64"):
            run_cmd(f'ln -s "{conda_env_path}/lib" "{conda_env_path}/lib64"', environment=True)

        # On some Linux distributions, g++ may not exist or be the wrong version to compile GPTQ-for-LLaMa
        gxx_output = run_cmd("g++ -dumpfullversion -dumpversion", environment=True, capture_output=True)
        if gxx_output.returncode != 0 or int(gxx_output.stdout.strip().split(b".")[0]) > 11:
            # Install the correct version of g++
            run_cmd("conda install -y -k conda-forge::gxx_linux-64=11.2.0", environment=True)

    clear_cache()


def download_model():
    run_cmd("python download-model.py", environment=True)


def launch_webui():
    run_cmd(f"python server.py {flags}", environment=True)


if __name__ == "__main__":
    # Verifies we are in a conda environment
    check_env()

    parser = argparse.ArgumentParser(add_help=False)
    parser.add_argument('--update', action='store_true', help='Update the web UI.')
    args, _ = parser.parse_known_args()

    if args.update:
        update_requirements()
    else:
        # If webui has already been installed, skip and run
        if not is_installed():
            install_webui()
            os.chdir(script_dir)

        if os.environ.get("LAUNCH_AFTER_INSTALL", "").lower() in ("no", "n", "false", "0", "f", "off"):
            print_big_message("Install finished successfully and will now exit due to LAUNCH_AFTER_INSTALL.")
            sys.exit()

        # Check if a model has been downloaded yet
        if '--model-dir' in flags:
            # Splits on ' ' or '=' while maintaining spaces within quotes
            flags_list = re.split(' +(?=(?:[^\"]*\"[^\"]*\")*[^\"]*$)|=', flags)
            model_dir = [flags_list[(flags_list.index(flag)+1)] for flag in flags_list if flag == '--model-dir'][0].strip('"\'')
        else:
            model_dir = 'models'

        if len([item for item in glob.glob(f'{model_dir}/*') if not item.endswith(('.txt', '.yaml'))]) == 0:
            print_big_message("WARNING: You haven't downloaded any model yet.\nOnce the web UI launches, head over to the \"Model\" tab and download one.")

        # Workaround for llama-cpp-python loading paths in CUDA env vars even if they do not exist
        conda_path_bin = os.path.join(conda_env_path, "bin")
        if not os.path.exists(conda_path_bin):
            os.mkdir(conda_path_bin)

        # Launch the webui
        launch_webui()