improve update for textgenui
This commit is contained in:
parent
08e3df166a
commit
eb60acc2a0
@ -2,6 +2,7 @@ from core.stack import Stack
|
||||
|
||||
|
||||
class TextGenerationWebui(Stack):
|
||||
|
||||
def __init__(self):
|
||||
super().__init__(
|
||||
'Text Generation',
|
||||
@ -10,6 +11,8 @@ class TextGenerationWebui(Stack):
|
||||
'https://github.com/oobabooga/text-generation-webui/'
|
||||
)
|
||||
|
||||
self.exllama = "0.2.1"
|
||||
|
||||
def _install(self):
|
||||
# Install LlamaCpp from prebuilt
|
||||
self.pip_install("llama-cpp-python", env=["CMAKE_ARGS=\"-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS\""]) # cpu
|
||||
@ -41,13 +44,19 @@ class TextGenerationWebui(Stack):
|
||||
# Fix llama trying to use cuda version
|
||||
self.remove_line_in_file("llama_cpp_cuda", "webui/modules/llama_cpp_python_hijack.py")
|
||||
|
||||
# Install useful packages
|
||||
# Install ExLlamaV2 and auto-gptq
|
||||
self.pip_install(
|
||||
"https://github.com/turboderp/exllamav2/releases/download/v0.1.9/exllamav2-0.1.9+rocm6.1.torch2.4.0-cp310-cp310-linux_x86_64.whl")
|
||||
f"https://github.com/turboderp/exllamav2/releases/download/v{self.exllama}/exllamav2-{self.exllama}+rocm6.1.torch2.4.0-cp310-cp310-linux_x86_64.whl")
|
||||
self.install_from_prebuilt("bitsandbytes")
|
||||
self.pip_install("auto-gptq", args=["--no-build-isolation", "--extra-index-url",
|
||||
"https://huggingface.github.io/autogptq-index/whl/rocm573/"])
|
||||
|
||||
def _update(self):
|
||||
self.pip_install(
|
||||
f"https://github.com/turboderp/exllamav2/releases/download/v{self.exllama}/exllamav2-{self.exllama}+rocm6.1.torch2.4.0-cp310-cp310-linux_x86_64.whl")
|
||||
self.pip_install("auto-gptq", args=["--no-build-isolation", "--extra-index-url",
|
||||
"https://huggingface.github.io/autogptq-index/whl/rocm573/"])
|
||||
|
||||
def _start(self):
|
||||
args = ["--listen", "--listen-port", str(self.port)]
|
||||
self.python(f"server.py", args=args, current_dir="webui",
|
||||
|
Loading…
Reference in New Issue
Block a user