leafspark
/
AutoGGUF
mirror of https://github.com/leafspark/AutoGGUF
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

refactor: move functions to helper modules 

- move functions to helper modules
- set radon to lenient (E)
- disable pre-commit workflow
- adjust module importing
This commit is contained in:
BuildTools 2024-08-16 15:50:39 -07:00
parent 000ca6bb1f
commit 9d939151de
No known key found for this signature in database
GPG Key ID: 3270C066C15D530B
11 changed files with 318 additions and 229 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

0
.github/workflows/pre-commit.yml → .github/workflows/pre-commit.yml.disabled vendored
View File

8
.github/workflows/radon.yml vendored
View File

@ -38,10 +38,10 @@ jobs:
if [ -n "$CHANGED_FILES" ]; then
echo "Running Cyclomatic Complexity check..."
radon cc $CHANGED_FILES -a -s -n D --exclude "AutoGGUF.quantize_model"
radon cc $CHANGED_FILES -a -s -n E --exclude "AutoGGUF.quantize_model"
echo "Running Maintainability Index check..."
radon mi $CHANGED_FILES -s -n D
radon mi $CHANGED_FILES -s -n E
else
echo "No Python files to analyze."
fi
@ -56,8 +56,8 @@ jobs:
fi
if [ -n "$CHANGED_FILES" ]; then
CC_OUTPUT=$(radon cc $CHANGED_FILES -a -s -n D --exclude "AutoGGUF.quantize_model")
MI_OUTPUT=$(radon mi $CHANGED_FILES -s -n D)
CC_OUTPUT=$(radon cc $CHANGED_FILES -a -s -n E --exclude "AutoGGUF.quantize_model")
MI_OUTPUT=$(radon mi $CHANGED_FILES -s -n E)
if [ -n "$CC_OUTPUT" ] || [ -n "$MI_OUTPUT" ]; then
echo "Radon detected code complexity or maintainability issues:"

194
src/AutoGGUF.py
View File

@ -1,26 +1,26 @@
import json
import re
import shutil
from datetime import datetime
import psutil
import requests
from functools import partial
from PySide6.QtCore import *
from PySide6.QtGui import *
from PySide6.QtWidgets import *
from DownloadThread import DownloadThread
from GPUMonitor import GPUMonitor
from KVOverrideEntry import KVOverrideEntry
from Logger import Logger
from ModelInfoDialog import ModelInfoDialog
from QuantizationThread import QuantizationThread
from TaskListItem import TaskListItem
from error_handling import show_error, handle_error
from imports_and_globals import ensure_directory, open_file_safe, resource_path
from localizations import *
from ui_update import *
from src.GPUMonitor import GPUMonitor
from src.KVOverrideEntry import KVOverrideEntry
from src.Logger import Logger
from src.ModelInfoDialog import ModelInfoDialog
from src.imports_and_globals import (
open_file_safe,
resource_path,
show_about,
ensure_directory,
)
from src.localizations import *
import src.ui_update
import src.lora_conversion
import src.utils
class AutoGGUF(QMainWindow):
@ -37,17 +37,35 @@ def __init__(self):
ensure_directory(os.path.abspath("models"))
# References
self.update_base_model_visibility = partial(update_base_model_visibility, self)
self.update_assets = update_assets.__get__(self)
self.update_cuda_option = update_cuda_option.__get__(self)
self.update_cuda_backends = update_cuda_backends.__get__(self)
self.update_threads_spinbox = partial(update_threads_spinbox, self)
self.update_threads_slider = partial(update_threads_slider, self)
self.update_gpu_offload_spinbox = partial(update_gpu_offload_spinbox, self)
self.update_gpu_offload_slider = partial(update_gpu_offload_slider, self)
self.update_model_info = partial(update_model_info, self.logger, self)
self.update_system_info = partial(update_system_info, self)
self.update_download_progress = partial(update_download_progress, self)
self.update_base_model_visibility = partial(
src.ui_update.update_base_model_visibility, self
)
self.update_assets = src.ui_update.update_assets.__get__(self)
self.update_cuda_option = src.ui_update.update_cuda_option.__get__(self)
self.update_cuda_backends = src.ui_update.update_cuda_backends.__get__(self)
self.download_llama_cpp = src.utils.download_llama_cpp.__get__(self)
self.refresh_releases = src.utils.refresh_releases.__get__(self)
self.browse_lora_input = src.utils.browse_lora_input.__get__(self)
self.browse_lora_output = src.utils.browse_lora_output.__get__(self)
self.convert_lora = src.lora_conversion.convert_lora.__get__(self)
self.show_about = show_about.__get__(self)
self.update_threads_spinbox = partial(
src.ui_update.update_threads_spinbox, self
)
self.update_threads_slider = partial(src.ui_update.update_threads_slider, self)
self.update_gpu_offload_spinbox = partial(
src.ui_update.update_gpu_offload_spinbox, self
)
self.update_gpu_offload_slider = partial(
src.ui_update.update_gpu_offload_slider, self
)
self.update_model_info = partial(
src.ui_update.update_model_info, self.logger, self
)
self.update_system_info = partial(src.ui_update.update_system_info, self)
self.update_download_progress = partial(
src.ui_update.update_download_progress, self
)
# Create a central widget and main layout
central_widget = QWidget()
@ -711,14 +729,6 @@ def refresh_backends(self):
self.backend_combo.setEnabled(False)
self.logger.info(FOUND_VALID_BACKENDS.format(self.backend_combo.count()))
def show_about(self):
about_text = (
"AutoGGUF\n\n"
f"Version: {AUTOGGUF_VERSION}\n\n"
"A tool for managing and converting GGUF models."
)
QMessageBox.about(self, "About AutoGGUF", about_text)
def save_preset(self):
self.logger.info(SAVING_PRESET)
preset = {
@ -1060,87 +1070,6 @@ def restart_task(self, task_item):
task_item.update_status(IN_PROGRESS)
break
def browse_lora_input(self):
self.logger.info(BROWSING_FOR_LORA_INPUT_DIRECTORY)
lora_input_path = QFileDialog.getExistingDirectory(
self, SELECT_LORA_INPUT_DIRECTORY
)
if lora_input_path:
self.lora_input.setText(os.path.abspath(lora_input_path))
ensure_directory(lora_input_path)
def browse_lora_output(self):
self.logger.info(BROWSING_FOR_LORA_OUTPUT_FILE)
lora_output_file, _ = QFileDialog.getSaveFileName(
self, SELECT_LORA_OUTPUT_FILE, "", GGUF_AND_BIN_FILES
)
if lora_output_file:
self.lora_output.setText(os.path.abspath(lora_output_file))
def convert_lora(self):
self.logger.info(STARTING_LORA_CONVERSION)
try:
lora_input_path = self.lora_input.text()
lora_output_path = self.lora_output.text()
lora_output_type = self.lora_output_type_combo.currentText()
if not lora_input_path:
raise ValueError(LORA_INPUT_PATH_REQUIRED)
if not lora_output_path:
raise ValueError(LORA_OUTPUT_PATH_REQUIRED)
if lora_output_type == "GGUF": # Use new file and parameters for GGUF
command = [
"python",
"src/convert_lora_to_gguf.py",
"--outfile",
lora_output_path,
lora_input_path,
]
base_model_path = self.base_model_path.text()
if not base_model_path:
raise ValueError(BASE_MODEL_PATH_REQUIRED)
command.extend(["--base", base_model_path])
else: # Use old GGML parameters for GGML
command = ["python", "src/convert_lora_to_ggml.py", lora_input_path]
logs_path = self.logs_input.text()
ensure_directory(logs_path)
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
log_file = os.path.join(logs_path, f"lora_conversion_{timestamp}.log")
command_str = " ".join(command)
self.logger.info(f"{LORA_CONVERSION_COMMAND}: {command_str}")
thread = QuantizationThread(command, os.getcwd(), log_file)
self.quant_threads.append(thread)
task_name = LORA_CONVERSION_FROM_TO.format(
os.path.basename(lora_input_path), os.path.basename(lora_output_path)
)
task_item = TaskListItem(task_name, log_file, show_progress_bar=False)
list_item = QListWidgetItem(self.task_list)
list_item.setSizeHint(task_item.sizeHint())
self.task_list.addItem(list_item)
self.task_list.setItemWidget(list_item, task_item)
thread.status_signal.connect(task_item.update_status)
thread.finished_signal.connect(
lambda: self.lora_conversion_finished(
thread, lora_input_path, lora_output_path
)
)
thread.error_signal.connect(
lambda err: handle_error(self.logger, err, task_item)
)
thread.start()
self.logger.info(LORA_CONVERSION_TASK_STARTED)
except ValueError as e:
show_error(self.logger, str(e))
except Exception as e:
show_error(self.logger, ERROR_STARTING_LORA_CONVERSION.format(str(e)))
def lora_conversion_finished(self, thread, input_path, output_path):
self.logger.info(LORA_CONVERSION_FINISHED)
if thread in self.quant_threads:
@ -1194,43 +1123,6 @@ def download_finished(self, extract_dir):
if index >= 0:
self.backend_combo.setCurrentIndex(index)
def refresh_releases(self):
self.logger.info(REFRESHING_LLAMACPP_RELEASES)
try:
response = requests.get(
"https://api.github.com/repos/ggerganov/llama.cpp/releases"
)
response.raise_for_status() # Raise an exception for bad status codes
releases = response.json()
self.release_combo.clear()
for release in releases:
self.release_combo.addItem(release["tag_name"], userData=release)
self.release_combo.currentIndexChanged.connect(self.update_assets)
self.update_assets()
except requests.exceptions.RequestException as e:
show_error(self.logger, ERROR_FETCHING_RELEASES.format(str(e)))
def download_llama_cpp(self):
self.logger.info(STARTING_LLAMACPP_DOWNLOAD)
asset = self.asset_combo.currentData()
if not asset:
show_error(self.logger, NO_ASSET_SELECTED)
return
llama_bin = os.path.abspath("llama_bin")
os.makedirs(llama_bin, exist_ok=True)
save_path = os.path.join(llama_bin, asset["name"])
self.download_thread = DownloadThread(asset["browser_download_url"], save_path)
self.download_thread.progress_signal.connect(self.update_download_progress)
self.download_thread.finished_signal.connect(self.download_finished)
self.download_thread.error_signal.connect(self.download_error)
self.download_thread.start()
self.download_button.setEnabled(False)
self.download_progress.setValue(0)
def download_finished(self, extract_dir):
self.download_button.setEnabled(True)
self.download_progress.setValue(100)

2
src/GPUMonitor.py
View File

@ -15,7 +15,7 @@
QComboBox,
)
from localizations import (
from src.localizations import (
GPU_USAGE_FORMAT,
GPU_DETAILS,
GPU_USAGE_OVER_TIME,

2
src/error_handling.py
View File

@ -1,5 +1,5 @@
from PySide6.QtWidgets import QMessageBox
from localizations import *
from src.localizations import *
def show_error(logger, message):

11
src/imports_and_globals.py
View File

@ -38,6 +38,17 @@
from PySide6.QtCore import QTimer, Signal, QThread, Qt, QSize
from PySide6.QtGui import QCloseEvent, QAction
from src.localizations import *
def show_about(self):
about_text = (
"AutoGGUF\n\n"
f"Version: {AUTOGGUF_VERSION}\n\n"
"A tool for managing and converting GGUF models."
)
QMessageBox.about(self, "About AutoGGUF", about_text)
def ensure_directory(path):
if not os.path.exists(path):

186
src/localizations.py
View File

@ -877,7 +877,9 @@ def __init__(self):
self.DOWNLOAD_FINISHED_EXTRACTED_TO = "下载完成。已解压到：{0}"
self.LLAMACPP_DOWNLOADED_AND_EXTRACTED = "llama.cpp二进制文件已下载并解压到{0}"
self.NO_SUITABLE_CUDA_BACKEND_FOUND = "未找到合适的CUDA后端进行提取"
self.LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED = "llama.cpp二进制文件已下载并解压到{0}"
self.LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED = (
"llama.cpp二进制文件已下载并解压到{0}"
)
self.REFRESHING_LLAMACPP_RELEASES = "刷新llama.cpp版本"
self.UPDATING_ASSET_LIST = "更新资源列表"
self.UPDATING_CUDA_OPTIONS = "更新CUDA选项"
@ -941,7 +943,9 @@ def __init__(self):
self.USE_THIS_TYPE_FOR_OUTPUT_WEIGHT = "对output.weight张量使用此类型"
self.TOKEN_EMBEDDING_TYPE = "词元嵌入类型："
self.USE_THIS_TYPE_FOR_TOKEN_EMBEDDINGS = "对词元嵌入张量使用此类型"
self.WILL_GENERATE_QUANTIZED_MODEL_IN_SAME_SHARDS = "将生成与输入相同分片的量化模型"
self.WILL_GENERATE_QUANTIZED_MODEL_IN_SAME_SHARDS = (
"将生成与输入相同分片的量化模型"
)
self.OVERRIDE_MODEL_METADATA = "覆盖模型元数据"
self.INPUT_DATA_FILE_FOR_IMATRIX = "IMatrix生成的输入数据文件"
self.MODEL_TO_BE_QUANTIZED = "要量化的模型"
@ -988,7 +992,9 @@ def __init__(self):
self.MODEL_DIRECTORY_REQUIRED = "需要模型目录"
self.HF_TO_GGUF_CONVERSION_COMMAND = "HF到GGUF转换命令：{}"
self.CONVERTING_TO_GGUF = "将{}转换为GGUF"
self.ERROR_STARTING_HF_TO_GGUF_CONVERSION = "启动HuggingFace到GGUF转换时出错：{}"
self.ERROR_STARTING_HF_TO_GGUF_CONVERSION = (
"启动HuggingFace到GGUF转换时出错：{}"
)
self.HF_TO_GGUF_CONVERSION_TASK_STARTED = "HuggingFace到GGUF转换任务已开始"
@ -1436,7 +1442,9 @@ def __init__(self):
self.NO_MODEL_SELECTED = "कोई मॉडल चयनित नहीं"
self.REFRESH_RELEASES = "रिलीज़ रीफ्रेश करें"
self.NO_SUITABLE_CUDA_BACKENDS = "कोई उपयुक्त CUDA बैकएंड नहीं मिला"
self.LLAMACPP_DOWNLOADED_EXTRACTED = "llama.cpp बाइनरी डाउनलोड और {0} में निकाली गई\nCUDA फ़ाइलें {1} में निकाली गईं"
self.LLAMACPP_DOWNLOADED_EXTRACTED = (
"llama.cpp बाइनरी डाउनलोड और {0} में निकाली गई\nCUDA फ़ाइलें {1} में निकाली गईं"
)
self.CUDA_FILES_EXTRACTED = "CUDA फ़ाइलें निकाली गईं"
self.NO_SUITABLE_CUDA_BACKEND_EXTRACTION = (
"निष्कर्षण के लिए कोई उपयुक्त CUDA बैकएंड नहीं मिला"
@ -1465,7 +1473,9 @@ def __init__(self):
self.RESTARTING_TASK = "कार्य पुनः आरंभ हो रहा है: {0}"
self.IN_PROGRESS = "प्रगति में"
self.DOWNLOAD_FINISHED_EXTRACTED_TO = "डाउनलोड समाप्त। निकाला गया: {0}"
self.LLAMACPP_DOWNLOADED_AND_EXTRACTED = "llama.cpp बाइनरी डाउनलोड और {0} में निकाली गई\nCUDA फ़ाइलें {1} में निकाली गईं"
self.LLAMACPP_DOWNLOADED_AND_EXTRACTED = (
"llama.cpp बाइनरी डाउनलोड और {0} में निकाली गई\nCUDA फ़ाइलें {1} में निकाली गईं"
)
self.NO_SUITABLE_CUDA_BACKEND_FOUND = (
"निष्कर्षण के लिए कोई उपयुक्त CUDA बैकएंड नहीं मिला"
)
@ -1487,25 +1497,17 @@ def __init__(self):
self.DELETING_TASK = "कार्य हटाया जा रहा है: {0}"
self.LOADING_MODELS = "मॉडल लोड हो रहे हैं"
self.LOADED_MODELS = "{0} मॉडल लोड किए गए"
self.BROWSING_FOR_MODELS_DIRECTORY = (
"मॉडल निर्देशिका के लिए ब्राउज़ किया जा रहा है"
)
self.BROWSING_FOR_MODELS_DIRECTORY = "मॉडल निर्देशिका के लिए ब्राउज़ किया जा रहा है"
self.SELECT_MODELS_DIRECTORY = "मॉडल निर्देशिका चुनें"
self.BROWSING_FOR_OUTPUT_DIRECTORY = (
"आउटपुट निर्देशिका के लिए ब्राउज़ किया जा रहा है"
)
self.BROWSING_FOR_OUTPUT_DIRECTORY = "आउटपुट निर्देशिका के लिए ब्राउज़ किया जा रहा है"
self.SELECT_OUTPUT_DIRECTORY = "आउटपुट निर्देशिका चुनें"
self.BROWSING_FOR_LOGS_DIRECTORY = (
"लॉग निर्देशिका के लिए ब्राउज़ किया जा रहा है"
)
self.BROWSING_FOR_LOGS_DIRECTORY = "लॉग निर्देशिका के लिए ब्राउज़ किया जा रहा है"
self.SELECT_LOGS_DIRECTORY = "लॉग निर्देशिका चुनें"
self.BROWSING_FOR_IMATRIX_FILE = "IMatrix फ़ाइल के लिए ब्राउज़ किया जा रहा है"
self.SELECT_IMATRIX_FILE = "IMatrix फ़ाइल चुनें"
self.RAM_USAGE_FORMAT = "{0:.1f}% ({1} MB / {2} MB)"
self.CPU_USAGE_FORMAT = "CPU उपयोग: {0:.1f}%"
self.VALIDATING_QUANTIZATION_INPUTS = (
"क्वांटाइजेशन इनपुट सत्यापित किए जा रहे हैं"
)
self.VALIDATING_QUANTIZATION_INPUTS = "क्वांटाइजेशन इनपुट सत्यापित किए जा रहे हैं"
self.MODELS_PATH_REQUIRED = "मॉडल पथ आवश्यक है"
self.OUTPUT_PATH_REQUIRED = "आउटपुट पथ आवश्यक है"
self.LOGS_PATH_REQUIRED = "लॉग पथ आवश्यक है"
@ -1532,9 +1534,7 @@ def __init__(self):
self.STARTING_IMATRIX_GENERATION = "IMatrix उत्पादन शुरू हो रहा है"
self.BACKEND_PATH_NOT_EXIST = "बैकएंड पथ मौजूद नहीं है: {0}"
self.GENERATING_IMATRIX = "IMatrix उत्पन्न किया जा रहा है"
self.ERROR_STARTING_IMATRIX_GENERATION = (
"IMatrix उत्पादन शुरू करने में त्रुटि: {0}"
)
self.ERROR_STARTING_IMATRIX_GENERATION = "IMatrix उत्पादन शुरू करने में त्रुटि: {0}"
self.IMATRIX_GENERATION_TASK_STARTED = "IMatrix उत्पादन कार्य शुरू हुआ"
self.ERROR_MESSAGE = "त्रुटि: {0}"
self.TASK_ERROR = "कार्य त्रुटि: {0}"
@ -1544,14 +1544,14 @@ def __init__(self):
self.ALLOWS_REQUANTIZING = (
"पहले से क्वांटाइज़ किए गए टेंसर को पुनः क्वांटाइज़ करने की अनुमति देता है"
)
self.LEAVE_OUTPUT_WEIGHT = (
"output.weight को अक्वांटाइज़ (या पुनः क्वांटाइज़) छोड़ देगा"
self.LEAVE_OUTPUT_WEIGHT = "output.weight को अक्वांटाइज़ (या पुनः क्वांटाइज़) छोड़ देगा"
self.DISABLE_K_QUANT_MIXTURES = (
"k-quant मिश्रण को अक्षम करें और सभी टेंसर को एक ही प्रकार में क्वांटाइज़ करें"
)
self.DISABLE_K_QUANT_MIXTURES = "k-quant मिश्रण को अक्षम करें और सभी टेंसर को एक ही प्रकार में क्वांटाइज़ करें"
self.USE_DATA_AS_IMPORTANCE_MATRIX = "क्वांट अनुकूलन के लिए फ़ाइल में डेटा को महत्व मैट्रिक्स के रूप में उपयोग करें"
self.USE_IMPORTANCE_MATRIX_FOR_TENSORS = (
"इन टेंसर के लिए महत्व मैट्रिक्स का उपयोग करें"
self.USE_DATA_AS_IMPORTANCE_MATRIX = (
"क्वांट अनुकूलन के लिए फ़ाइल में डेटा को महत्व मैट्रिक्स के रूप में उपयोग करें"
)
self.USE_IMPORTANCE_MATRIX_FOR_TENSORS = "इन टेंसर के लिए महत्व मैट्रिक्स का उपयोग करें"
self.DONT_USE_IMPORTANCE_MATRIX_FOR_TENSORS = (
"इन टेंसर के लिए महत्व मैट्रिक्स का उपयोग न करें"
)
@ -2008,7 +2008,9 @@ def __init__(self):
self.RESTART = "再起動"
self.DELETE = "削除"
self.CONFIRM_DELETION = "このタスクを削除してもよろしいですか？"
self.TASK_RUNNING_WARNING = "一部のタスクはまだ実行中です。終了してもよろしいですか？"
self.TASK_RUNNING_WARNING = (
"一部のタスクはまだ実行中です。終了してもよろしいですか？"
)
self.YES = "はい"
self.NO = "いいえ"
self.DOWNLOAD_COMPLETE = "ダウンロード完了"
@ -2021,11 +2023,11 @@ def __init__(self):
self.NO_MODEL_SELECTED = "モデルが選択されていません"
self.REFRESH_RELEASES = "リリースを更新"
self.NO_SUITABLE_CUDA_BACKENDS = "適切なCUDAバックエンドが見つかりませんでした"
self.LLAMACPP_DOWNLOADED_EXTRACTED = (
"llama.cppバイナリがダウンロードされ、{0}に抽出されました\nCUDAファイルは{1}に抽出されました"
)
self.LLAMACPP_DOWNLOADED_EXTRACTED = "llama.cppバイナリがダウンロードされ、{0}に抽出されました\nCUDAファイルは{1}に抽出されました"
self.CUDA_FILES_EXTRACTED = "CUDAファイルはに抽出されました"
self.NO_SUITABLE_CUDA_BACKEND_EXTRACTION = "抽出に適したCUDAバックエンドが見つかりませんでした"
self.NO_SUITABLE_CUDA_BACKEND_EXTRACTION = (
"抽出に適したCUDAバックエンドが見つかりませんでした"
)
self.ERROR_FETCHING_RELEASES = "リリースの取得中にエラーが発生しました: {0}"
self.CONFIRM_DELETION_TITLE = "削除の確認"
self.LOG_FOR = "{0}のログ"
@ -2050,10 +2052,10 @@ def __init__(self):
self.RESTARTING_TASK = "タスクを再起動しています: {0}"
self.IN_PROGRESS = "処理中"
self.DOWNLOAD_FINISHED_EXTRACTED_TO = "ダウンロードが完了しました。抽出先: {0}"
self.LLAMACPP_DOWNLOADED_AND_EXTRACTED = (
"llama.cppバイナリがダウンロードされ、{0}に抽出されました\nCUDAファイルは{1}に抽出されました"
self.LLAMACPP_DOWNLOADED_AND_EXTRACTED = "llama.cppバイナリがダウンロードされ、{0}に抽出されました\nCUDAファイルは{1}に抽出されました"
self.NO_SUITABLE_CUDA_BACKEND_FOUND = (
"抽出に適したCUDAバックエンドが見つかりませんでした"
)
self.NO_SUITABLE_CUDA_BACKEND_FOUND = "抽出に適したCUDAバックエンドが見つかりませんでした"
self.LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED = (
"llama.cppバイナリがダウンロードされ、{0}に抽出されました"
)
@ -2103,24 +2105,42 @@ def __init__(self):
self.STARTING_IMATRIX_GENERATION = "IMatrixの生成を開始しています"
self.BACKEND_PATH_NOT_EXIST = "バックエンドパスが存在しません: {0}"
self.GENERATING_IMATRIX = "IMatrixを生成しています"
self.ERROR_STARTING_IMATRIX_GENERATION = "IMatrixの生成を開始中にエラーが発生しました: {0}"
self.ERROR_STARTING_IMATRIX_GENERATION = (
"IMatrixの生成を開始中にエラーが発生しました: {0}"
)
self.IMATRIX_GENERATION_TASK_STARTED = "IMatrix生成タスクが開始されました"
self.ERROR_MESSAGE = "エラー: {0}"
self.TASK_ERROR = "タスクエラー: {0}"
self.APPLICATION_CLOSING = "アプリケーションを終了しています"
self.APPLICATION_CLOSED = "アプリケーションが終了しました"
self.SELECT_QUANTIZATION_TYPE = "量子化タイプを選択してください"
self.ALLOWS_REQUANTIZING = "すでに量子化されているテンソルの再量子化を許可します"
self.ALLOWS_REQUANTIZING = (
"すでに量子化されているテンソルの再量子化を許可します"
)
self.LEAVE_OUTPUT_WEIGHT = "output.weightは（再）量子化されません"
self.DISABLE_K_QUANT_MIXTURES = "k-quant混合を無効にし、すべてのテンソルを同じタイプに量子化します"
self.USE_DATA_AS_IMPORTANCE_MATRIX = "量子化最適化の重要度マトリックスとしてファイル内のデータを使用します"
self.USE_IMPORTANCE_MATRIX_FOR_TENSORS = "これらのテンソルに重要度マトリックスを使用します"
self.DONT_USE_IMPORTANCE_MATRIX_FOR_TENSORS = "これらのテンソルに重要度マトリックスを使用しません"
self.DISABLE_K_QUANT_MIXTURES = (
"k-quant混合を無効にし、すべてのテンソルを同じタイプに量子化します"
)
self.USE_DATA_AS_IMPORTANCE_MATRIX = (
"量子化最適化の重要度マトリックスとしてファイル内のデータを使用します"
)
self.USE_IMPORTANCE_MATRIX_FOR_TENSORS = (
"これらのテンソルに重要度マトリックスを使用します"
)
self.DONT_USE_IMPORTANCE_MATRIX_FOR_TENSORS = (
"これらのテンソルに重要度マトリックスを使用しません"
)
self.OUTPUT_TENSOR_TYPE = "出力テンソルタイプ:"
self.USE_THIS_TYPE_FOR_OUTPUT_WEIGHT = "output.weightテンソルにこのタイプを使用します"
self.USE_THIS_TYPE_FOR_OUTPUT_WEIGHT = (
"output.weightテンソルにこのタイプを使用します"
)
self.TOKEN_EMBEDDING_TYPE = "トークン埋め込みタイプ:"
self.USE_THIS_TYPE_FOR_TOKEN_EMBEDDINGS = "トークン埋め込みテンソルにこのタイプを使用します"
self.WILL_GENERATE_QUANTIZED_MODEL_IN_SAME_SHARDS = "入力と同じシャードで量子化されたモデルを生成します"
self.USE_THIS_TYPE_FOR_TOKEN_EMBEDDINGS = (
"トークン埋め込みテンソルにこのタイプを使用します"
)
self.WILL_GENERATE_QUANTIZED_MODEL_IN_SAME_SHARDS = (
"入力と同じシャードで量子化されたモデルを生成します"
)
self.OVERRIDE_MODEL_METADATA = "モデルメタデータを上書きする"
self.INPUT_DATA_FILE_FOR_IMATRIX = "IMatrix生成用の入力データファイル"
self.MODEL_TO_BE_QUANTIZED = "量子化されるモデル"
@ -2777,11 +2797,11 @@ def __init__(self):
self.NO_MODEL_SELECTED = "모델이 선택되지 않았습니다"
self.REFRESH_RELEASES = "릴리스 새로 고침"
self.NO_SUITABLE_CUDA_BACKENDS = "적합한 CUDA 백엔드를 찾을 수 없습니다"
self.LLAMACPP_DOWNLOADED_EXTRACTED = (
"llama.cpp 바이너리가 다운로드되어 {0}에 추출되었습니다.\nCUDA 파일이 {1}에 추출되었습니다."
)
self.LLAMACPP_DOWNLOADED_EXTRACTED = "llama.cpp 바이너리가 다운로드되어 {0}에 추출되었습니다.\nCUDA 파일이 {1}에 추출되었습니다."
self.CUDA_FILES_EXTRACTED = "CUDA 파일이 에 추출되었습니다."
self.NO_SUITABLE_CUDA_BACKEND_EXTRACTION = "추출에 적합한 CUDA 백엔드를 찾을 수 없습니다."
self.NO_SUITABLE_CUDA_BACKEND_EXTRACTION = (
"추출에 적합한 CUDA 백엔드를 찾을 수 없습니다."
)
self.ERROR_FETCHING_RELEASES = "릴리스를 가져오는 중 오류가 발생했습니다: {0}"
self.CONFIRM_DELETION_TITLE = "삭제 확인"
self.LOG_FOR = "{0}에 대한 로그"
@ -2805,11 +2825,13 @@ def __init__(self):
self.TASK_PRESET_SAVED_TO = "작업 프리셋이 {0}에 저장되었습니다."
self.RESTARTING_TASK = "작업을 다시 시작하는 중입니다: {0}"
self.IN_PROGRESS = "진행 중"
self.DOWNLOAD_FINISHED_EXTRACTED_TO = "다운로드가 완료되었습니다. 추출 위치: {0}"
self.LLAMACPP_DOWNLOADED_AND_EXTRACTED = (
"llama.cpp 바이너리가 다운로드되어 {0}에 추출되었습니다.\nCUDA 파일이 {1}에 추출되었습니다."
self.DOWNLOAD_FINISHED_EXTRACTED_TO = (
"다운로드가 완료되었습니다. 추출 위치: {0}"
)
self.LLAMACPP_DOWNLOADED_AND_EXTRACTED = "llama.cpp 바이너리가 다운로드되어 {0}에 추출되었습니다.\nCUDA 파일이 {1}에 추출되었습니다."
self.NO_SUITABLE_CUDA_BACKEND_FOUND = (
"추출에 적합한 CUDA 백엔드를 찾을 수 없습니다."
)
self.NO_SUITABLE_CUDA_BACKEND_FOUND = "추출에 적합한 CUDA 백엔드를 찾을 수 없습니다."
self.LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED = (
"llama.cpp 바이너리가 다운로드되어 {0}에 추출되었습니다."
)
@ -2846,10 +2868,14 @@ def __init__(self):
self.INPUT_FILE_NOT_EXIST = "입력 파일 '{0}'이 존재하지 않습니다."
self.QUANTIZING_MODEL_TO = "{0}을 {1}(으)로 양자화하는 중입니다."
self.QUANTIZATION_TASK_STARTED = "{0}에 대한 양자화 작업이 시작되었습니다."
self.ERROR_STARTING_QUANTIZATION = "양자화를 시작하는 중 오류가 발생했습니다: {0}"
self.ERROR_STARTING_QUANTIZATION = (
"양자화를 시작하는 중 오류가 발생했습니다: {0}"
)
self.UPDATING_MODEL_INFO = "모델 정보를 업데이트하는 중입니다: {0}"
self.TASK_FINISHED = "작업이 완료되었습니다: {0}"
self.SHOWING_TASK_DETAILS_FOR = "다음에 대한 작업 세부 정보를 표시하는 중입니다: {0}"
self.SHOWING_TASK_DETAILS_FOR = (
"다음에 대한 작업 세부 정보를 표시하는 중입니다: {0}"
)
self.BROWSING_FOR_IMATRIX_DATA_FILE = "IMatrix 데이터 파일을 찾아보는 중입니다."
self.SELECT_DATA_FILE = "데이터 파일 선택"
self.BROWSING_FOR_IMATRIX_MODEL_FILE = "IMatrix 모델 파일을 찾아보는 중입니다."
@ -2859,7 +2885,9 @@ def __init__(self):
self.STARTING_IMATRIX_GENERATION = "IMatrix 생성을 시작하는 중입니다."
self.BACKEND_PATH_NOT_EXIST = "백엔드 경로가 존재하지 않습니다: {0}"
self.GENERATING_IMATRIX = "IMatrix를 생성하는 중입니다."
self.ERROR_STARTING_IMATRIX_GENERATION = "IMatrix 생성을 시작하는 중 오류가 발생했습니다: {0}"
self.ERROR_STARTING_IMATRIX_GENERATION = (
"IMatrix 생성을 시작하는 중 오류가 발생했습니다: {0}"
)
self.IMATRIX_GENERATION_TASK_STARTED = "IMatrix 생성 작업이 시작되었습니다."
self.ERROR_MESSAGE = "오류: {0}"
self.TASK_ERROR = "작업 오류: {0}"
@ -2868,14 +2896,26 @@ def __init__(self):
self.SELECT_QUANTIZATION_TYPE = "양자화 유형을 선택하세요."
self.ALLOWS_REQUANTIZING = "이미 양자화된 텐서의 재양자화를 허용합니다."
self.LEAVE_OUTPUT_WEIGHT = "output.weight를 (재)양자화하지 않은 상태로 둡니다."
self.DISABLE_K_QUANT_MIXTURES = "k-양자 혼합을 비활성화하고 모든 텐서를 동일한 유형으로 양자화합니다."
self.USE_DATA_AS_IMPORTANCE_MATRIX = "양자 최적화를 위한 중요도 행렬로 파일의 데이터를 사용합니다."
self.USE_IMPORTANCE_MATRIX_FOR_TENSORS = "이러한 텐서에 중요도 행렬을 사용합니다."
self.DONT_USE_IMPORTANCE_MATRIX_FOR_TENSORS = "이러한 텐서에 중요도 행렬을 사용하지 않습니다."
self.DISABLE_K_QUANT_MIXTURES = (
"k-양자 혼합을 비활성화하고 모든 텐서를 동일한 유형으로 양자화합니다."
)
self.USE_DATA_AS_IMPORTANCE_MATRIX = (
"양자 최적화를 위한 중요도 행렬로 파일의 데이터를 사용합니다."
)
self.USE_IMPORTANCE_MATRIX_FOR_TENSORS = (
"이러한 텐서에 중요도 행렬을 사용합니다."
)
self.DONT_USE_IMPORTANCE_MATRIX_FOR_TENSORS = (
"이러한 텐서에 중요도 행렬을 사용하지 않습니다."
)
self.OUTPUT_TENSOR_TYPE = "출력 텐서 유형:"
self.USE_THIS_TYPE_FOR_OUTPUT_WEIGHT = "output.weight 텐서에 이 유형을 사용합니다."
self.USE_THIS_TYPE_FOR_OUTPUT_WEIGHT = (
"output.weight 텐서에 이 유형을 사용합니다."
)
self.TOKEN_EMBEDDING_TYPE = "토큰 임베딩 유형:"
self.USE_THIS_TYPE_FOR_TOKEN_EMBEDDINGS = "토큰 임베딩 텐서에 이 유형을 사용합니다."
self.USE_THIS_TYPE_FOR_TOKEN_EMBEDDINGS = (
"토큰 임베딩 텐서에 이 유형을 사용합니다."
)
self.WILL_GENERATE_QUANTIZED_MODEL_IN_SAME_SHARDS = (
"입력과 동일한 샤드에 양자화된 모델을 생성합니다."
)
@ -3830,9 +3870,7 @@ def __init__(self):
self.STARTING_IMATRIX_GENERATION = "IMatrix জেনারেশন শুরু হচ্ছে"
self.BACKEND_PATH_NOT_EXIST = "ব্যাকএন্ড পাথ বিদ্যমান নেই: {0}"
self.GENERATING_IMATRIX = "IMatrix তৈরি করা হচ্ছে"
self.ERROR_STARTING_IMATRIX_GENERATION = (
"IMatrix জেনারেশন শুরু করতে ত্রুটি: {0}"
)
self.ERROR_STARTING_IMATRIX_GENERATION = "IMatrix জেনারেশন শুরু করতে ত্রুটি: {0}"
self.IMATRIX_GENERATION_TASK_STARTED = "IMatrix জেনারেশন টাস্ক শুরু হয়েছে"
self.ERROR_MESSAGE = "ত্রুটি: {0}"
self.TASK_ERROR = "টাস্ক ত্রুটি: {0}"
@ -3840,11 +3878,13 @@ def __init__(self):
self.APPLICATION_CLOSED = "অ্যাপ্লিকেশন বন্ধ"
self.SELECT_QUANTIZATION_TYPE = "কোয়ান্টাইজেশন ধরণ নির্বাচন করুন"
self.ALLOWS_REQUANTIZING = "যে টেন্সরগুলি ইতিমধ্যে কোয়ান্টাইজ করা হয়েছে তাদের পুনরায় কোয়ান্টাইজ করার অনুমতি দেয়"
self.LEAVE_OUTPUT_WEIGHT = (
"output.weight কে (পুনরায়) কোয়ান্টাইজ না করে রেখে দেবে"
self.LEAVE_OUTPUT_WEIGHT = "output.weight কে (পুনরায়) কোয়ান্টাইজ না করে রেখে দেবে"
self.DISABLE_K_QUANT_MIXTURES = (
"k-কোয়ান্ট মিশ্রণগুলি অক্ষম করুন এবং সমস্ত টেন্সরকে একই ধরণের কোয়ান্টাইজ করুন"
)
self.USE_DATA_AS_IMPORTANCE_MATRIX = (
"কোয়ান্ট অপ্টিমাইজেশনের জন্য ফাইলের ডেটা গুরুত্বপূর্ণ ম্যাট্রিক্স হিসাবে ব্যবহার করুন"
)
self.DISABLE_K_QUANT_MIXTURES = "k-কোয়ান্ট মিশ্রণগুলি অক্ষম করুন এবং সমস্ত টেন্সরকে একই ধরণের কোয়ান্টাইজ করুন"
self.USE_DATA_AS_IMPORTANCE_MATRIX = "কোয়ান্ট অপ্টিমাইজেশনের জন্য ফাইলের ডেটা গুরুত্বপূর্ণ ম্যাট্রিক্স হিসাবে ব্যবহার করুন"
self.USE_IMPORTANCE_MATRIX_FOR_TENSORS = (
"এই টেন্সরগুলির জন্য গুরুত্বপূর্ণ ম্যাট্রিক্স ব্যবহার করুন"
)
@ -5948,7 +5988,9 @@ def __init__(self):
"llama.cpp 二進位檔案已下載並解壓縮至 {0}\nCUDA 檔案已解壓縮至 {1}"
)
self.NO_SUITABLE_CUDA_BACKEND_FOUND = "找不到合適的 CUDA 後端進行解壓縮"
self.LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED = "llama.cpp 二進位檔案已下載並解壓縮至 {0}"
self.LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED = (
"llama.cpp 二進位檔案已下載並解壓縮至 {0}"
)
self.REFRESHING_LLAMACPP_RELEASES = "正在重新整理 llama.cpp 版本"
self.UPDATING_ASSET_LIST = "正在更新資源清單"
self.UPDATING_CUDA_OPTIONS = "正在更新 CUDA 選項"
@ -6005,14 +6047,18 @@ def __init__(self):
self.ALLOWS_REQUANTIZING = "允許重新量化已量化的張量"
self.LEAVE_OUTPUT_WEIGHT = "將保留 output.weight 不被（重新）量化"
self.DISABLE_K_QUANT_MIXTURES = "停用 k-quant 混合並將所有張量量化為相同類型"
self.USE_DATA_AS_IMPORTANCE_MATRIX = "使用檔案中的資料作為量化最佳化的重要性矩陣"
self.USE_DATA_AS_IMPORTANCE_MATRIX = (
"使用檔案中的資料作為量化最佳化的重要性矩陣"
)
self.USE_IMPORTANCE_MATRIX_FOR_TENSORS = "對這些張量使用重要性矩陣"
self.DONT_USE_IMPORTANCE_MATRIX_FOR_TENSORS = "不要對這些張量使用重要性矩陣"
self.OUTPUT_TENSOR_TYPE = "輸出張量類型："
self.USE_THIS_TYPE_FOR_OUTPUT_WEIGHT = "對 output.weight 張量使用此類型"
self.TOKEN_EMBEDDING_TYPE = "權杖嵌入類型："
self.USE_THIS_TYPE_FOR_TOKEN_EMBEDDINGS = "對權杖嵌入張量使用此類型"
self.WILL_GENERATE_QUANTIZED_MODEL_IN_SAME_SHARDS = "將在與輸入相同的分片中產生量化模型"
self.WILL_GENERATE_QUANTIZED_MODEL_IN_SAME_SHARDS = (
"將在與輸入相同的分片中產生量化模型"
)
self.OVERRIDE_MODEL_METADATA = "覆蓋模型中繼資料"
self.INPUT_DATA_FILE_FOR_IMATRIX = "IMatrix 產生的輸入資料檔案"
self.MODEL_TO_BE_QUANTIZED = "要量化的模型"

74
src/lora_conversion.py Normal file
View File

@ -0,0 +1,74 @@
from datetime import datetime
from PySide6.QtWidgets import QListWidgetItem
from src.QuantizationThread import QuantizationThread
from src.TaskListItem import TaskListItem
from src.error_handling import handle_error, show_error
from src.imports_and_globals import ensure_directory
from src.localizations import *
def convert_lora(self):
self.logger.info(STARTING_LORA_CONVERSION)
try:
lora_input_path = self.lora_input.text()
lora_output_path = self.lora_output.text()
lora_output_type = self.lora_output_type_combo.currentText()
if not lora_input_path:
raise ValueError(LORA_INPUT_PATH_REQUIRED)
if not lora_output_path:
raise ValueError(LORA_OUTPUT_PATH_REQUIRED)
if lora_output_type == "GGUF": # Use new file and parameters for GGUF
command = [
"python",
"src/convert_lora_to_gguf.py",
"--outfile",
lora_output_path,
lora_input_path,
]
base_model_path = self.base_model_path.text()
if not base_model_path:
raise ValueError(BASE_MODEL_PATH_REQUIRED)
command.extend(["--base", base_model_path])
else: # Use old GGML parameters for GGML
command = ["python", "src/convert_lora_to_ggml.py", lora_input_path]
logs_path = self.logs_input.text()
ensure_directory(logs_path)
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
log_file = os.path.join(logs_path, f"lora_conversion_{timestamp}.log")
command_str = " ".join(command)
self.logger.info(f"{LORA_CONVERSION_COMMAND}: {command_str}")
thread = QuantizationThread(command, os.getcwd(), log_file)
self.quant_threads.append(thread)
task_name = LORA_CONVERSION_FROM_TO.format(
os.path.basename(lora_input_path), os.path.basename(lora_output_path)
)
task_item = TaskListItem(task_name, log_file, show_progress_bar=False)
list_item = QListWidgetItem(self.task_list)
list_item.setSizeHint(task_item.sizeHint())
self.task_list.addItem(list_item)
self.task_list.setItemWidget(list_item, task_item)
thread.status_signal.connect(task_item.update_status)
thread.finished_signal.connect(
lambda: self.lora_conversion_finished(
thread, lora_input_path, lora_output_path
)
)
thread.error_signal.connect(
lambda err: handle_error(self.logger, err, task_item)
)
thread.start()
self.logger.info(LORA_CONVERSION_TASK_STARTED)
except ValueError as e:
show_error(self.logger, str(e))
except Exception as e:
show_error(self.logger, ERROR_STARTING_LORA_CONVERSION.format(str(e)))

2
src/main.py
View File

@ -4,7 +4,7 @@
from PySide6.QtCore import QTimer
from PySide6.QtWidgets import QApplication
from AutoGGUF import AutoGGUF
from src.AutoGGUF import AutoGGUF
from flask import Flask, jsonify
server = Flask(__name__)

2
src/ui_update.py
View File

@ -1,4 +1,4 @@
from localizations import *
from src.localizations import *
import psutil

66
src/utils.py Normal file
View File

@ -0,0 +1,66 @@
from PySide6.QtWidgets import QFileDialog
from error_handling import show_error
from localizations import *
import requests
from src.DownloadThread import DownloadThread
from src.imports_and_globals import ensure_directory
def browse_lora_input(self):
self.logger.info(BROWSING_FOR_LORA_INPUT_DIRECTORY)
lora_input_path = QFileDialog.getExistingDirectory(
self, SELECT_LORA_INPUT_DIRECTORY
)
if lora_input_path:
self.lora_input.setText(os.path.abspath(lora_input_path))
ensure_directory(lora_input_path)
def browse_lora_output(self):
self.logger.info(BROWSING_FOR_LORA_OUTPUT_FILE)
lora_output_file, _ = QFileDialog.getSaveFileName(
self, SELECT_LORA_OUTPUT_FILE, "", GGUF_AND_BIN_FILES
)
if lora_output_file:
self.lora_output.setText(os.path.abspath(lora_output_file))
def download_llama_cpp(self):
self.logger.info(STARTING_LLAMACPP_DOWNLOAD)
asset = self.asset_combo.currentData()
if not asset:
show_error(self.logger, NO_ASSET_SELECTED)
return
llama_bin = os.path.abspath("llama_bin")
os.makedirs(llama_bin, exist_ok=True)
save_path = os.path.join(llama_bin, asset["name"])
self.download_thread = DownloadThread(asset["browser_download_url"], save_path)
self.download_thread.progress_signal.connect(self.update_download_progress)
self.download_thread.finished_signal.connect(self.download_finished)
self.download_thread.error_signal.connect(self.download_error)
self.download_thread.start()
self.download_button.setEnabled(False)
self.download_progress.setValue(0)
def refresh_releases(self):
self.logger.info(REFRESHING_LLAMACPP_RELEASES)
try:
response = requests.get(
"https://api.github.com/repos/ggerganov/llama.cpp/releases"
)
response.raise_for_status() # Raise an exception for bad status codes
releases = response.json()
self.release_combo.clear()
for release in releases:
self.release_combo.addItem(release["tag_name"], userData=release)
self.release_combo.currentIndexChanged.connect(self.update_assets)
self.update_assets()
except requests.exceptions.RequestException as e:
show_error(self.logger, ERROR_FETCHING_RELEASES.format(str(e)))