mirror of https://github.com/leafspark/AutoGGUF
adjust window size and code formatting
This commit is contained in:
BuildTools
parent
41ebb7d609
commit
3e8d7b1415
491
src/AutoGGUF.py
491
src/AutoGGUF.py
|
|
@ -22,15 +22,17 @@
|
|||
from Logger import Logger
|
||||
from localizations import *
|
||||
|
||||
|
||||
class AutoGGUF(QMainWindow):
|
||||
def __init__(self):
|
||||
super().__init__()
|
||||
self.logger = Logger("AutoGGUF", "logs")
|
||||
ensure_directory(os.path.abspath("quantized_models"))
|
||||
|
||||
self.logger.info(INITIALIZING_AUTOGGUF)
|
||||
self.setWindowTitle(WINDOW_TITLE)
|
||||
self.setGeometry(100, 100, 1300, 1100)
|
||||
self.setGeometry(100, 100, 1600, 1200)
|
||||
|
||||
ensure_directory(os.path.abspath("quantized_models"))
|
||||
|
||||
main_layout = QHBoxLayout()
|
||||
left_layout = QVBoxLayout()
|
||||
|
|
@ -89,7 +91,7 @@ def __init__(self):
|
|||
right_layout.addWidget(download_group)
|
||||
|
||||
# Initialize releases and backends
|
||||
if os.environ.get('AUTOGGUF_CHECK_BACKEND', '').lower() == 'enabled':
|
||||
if os.environ.get("AUTOGGUF_CHECK_BACKEND", "").lower() == "enabled":
|
||||
self.refresh_releases()
|
||||
self.refresh_backends()
|
||||
|
||||
|
|
@ -140,20 +142,64 @@ def __init__(self):
|
|||
quant_options_layout = QFormLayout()
|
||||
|
||||
self.quant_type = QComboBox()
|
||||
self.quant_type.addItems([
|
||||
"Q4_0", "Q4_1", "Q5_0", "Q5_1", "IQ2_XXS", "IQ2_XS", "IQ2_S", "IQ2_M", "IQ1_S", "IQ1_M",
|
||||
"Q2_K", "Q2_K_S", "IQ3_XXS", "IQ3_S", "IQ3_M", "Q3_K", "IQ3_XS", "Q3_K_S", "Q3_K_M", "Q3_K_L",
|
||||
"IQ4_NL", "IQ4_XS", "Q4_K", "Q4_K_S", "Q4_K_M", "Q5_K", "Q5_K_S", "Q5_K_M", "Q6_K", "Q8_0",
|
||||
"Q4_0_4_4", "Q4_0_4_8", "Q4_0_8_8", "F16", "BF16", "F32", "COPY"
|
||||
])
|
||||
quant_options_layout.addRow(self.create_label(QUANTIZATION_TYPE, SELECT_QUANTIZATION_TYPE), self.quant_type)
|
||||
self.quant_type.addItems(
|
||||
[
|
||||
"Q4_0",
|
||||
"Q4_1",
|
||||
"Q5_0",
|
||||
"Q5_1",
|
||||
"IQ2_XXS",
|
||||
"IQ2_XS",
|
||||
"IQ2_S",
|
||||
"IQ2_M",
|
||||
"IQ1_S",
|
||||
"IQ1_M",
|
||||
"Q2_K",
|
||||
"Q2_K_S",
|
||||
"IQ3_XXS",
|
||||
"IQ3_S",
|
||||
"IQ3_M",
|
||||
"Q3_K",
|
||||
"IQ3_XS",
|
||||
"Q3_K_S",
|
||||
"Q3_K_M",
|
||||
"Q3_K_L",
|
||||
"IQ4_NL",
|
||||
"IQ4_XS",
|
||||
"Q4_K",
|
||||
"Q4_K_S",
|
||||
"Q4_K_M",
|
||||
"Q5_K",
|
||||
"Q5_K_S",
|
||||
"Q5_K_M",
|
||||
"Q6_K",
|
||||
"Q8_0",
|
||||
"Q4_0_4_4",
|
||||
"Q4_0_4_8",
|
||||
"Q4_0_8_8",
|
||||
"F16",
|
||||
"BF16",
|
||||
"F32",
|
||||
"COPY",
|
||||
]
|
||||
)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label(QUANTIZATION_TYPE, SELECT_QUANTIZATION_TYPE),
|
||||
self.quant_type,
|
||||
)
|
||||
|
||||
self.allow_requantize = QCheckBox(ALLOW_REQUANTIZE)
|
||||
self.leave_output_tensor = QCheckBox(LEAVE_OUTPUT_TENSOR)
|
||||
self.pure = QCheckBox(PURE)
|
||||
quant_options_layout.addRow(self.create_label("", ALLOWS_REQUANTIZING), self.allow_requantize)
|
||||
quant_options_layout.addRow(self.create_label("", LEAVE_OUTPUT_WEIGHT), self.leave_output_tensor)
|
||||
quant_options_layout.addRow(self.create_label("", DISABLE_K_QUANT_MIXTURES), self.pure)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label("", ALLOWS_REQUANTIZING), self.allow_requantize
|
||||
)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label("", LEAVE_OUTPUT_WEIGHT), self.leave_output_tensor
|
||||
)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label("", DISABLE_K_QUANT_MIXTURES), self.pure
|
||||
)
|
||||
|
||||
self.imatrix = QLineEdit()
|
||||
self.imatrix_button = QPushButton(BROWSE)
|
||||
|
|
@ -161,36 +207,61 @@ def __init__(self):
|
|||
imatrix_layout = QHBoxLayout()
|
||||
imatrix_layout.addWidget(self.imatrix)
|
||||
imatrix_layout.addWidget(self.imatrix_button)
|
||||
quant_options_layout.addRow(self.create_label(IMATRIX, USE_DATA_AS_IMPORTANCE_MATRIX), imatrix_layout)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label(IMATRIX, USE_DATA_AS_IMPORTANCE_MATRIX), imatrix_layout
|
||||
)
|
||||
|
||||
self.include_weights = QLineEdit()
|
||||
self.exclude_weights = QLineEdit()
|
||||
quant_options_layout.addRow(self.create_label(INCLUDE_WEIGHTS, USE_IMPORTANCE_MATRIX_FOR_TENSORS), self.include_weights)
|
||||
quant_options_layout.addRow(self.create_label(EXCLUDE_WEIGHTS, DONT_USE_IMPORTANCE_MATRIX_FOR_TENSORS), self.exclude_weights)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label(INCLUDE_WEIGHTS, USE_IMPORTANCE_MATRIX_FOR_TENSORS),
|
||||
self.include_weights,
|
||||
)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label(EXCLUDE_WEIGHTS, DONT_USE_IMPORTANCE_MATRIX_FOR_TENSORS),
|
||||
self.exclude_weights,
|
||||
)
|
||||
|
||||
self.use_output_tensor_type = QCheckBox(USE_OUTPUT_TENSOR_TYPE)
|
||||
self.output_tensor_type = QComboBox()
|
||||
self.output_tensor_type.addItems(["F32", "F16", "Q4_0", "Q4_1", "Q5_0", "Q5_1", "Q8_0"])
|
||||
self.output_tensor_type.addItems(
|
||||
["F32", "F16", "Q4_0", "Q4_1", "Q5_0", "Q5_1", "Q8_0"]
|
||||
)
|
||||
self.output_tensor_type.setEnabled(False)
|
||||
self.use_output_tensor_type.toggled.connect(lambda checked: self.output_tensor_type.setEnabled(checked))
|
||||
self.use_output_tensor_type.toggled.connect(
|
||||
lambda checked: self.output_tensor_type.setEnabled(checked)
|
||||
)
|
||||
output_tensor_layout = QHBoxLayout()
|
||||
output_tensor_layout.addWidget(self.use_output_tensor_type)
|
||||
output_tensor_layout.addWidget(self.output_tensor_type)
|
||||
quant_options_layout.addRow(self.create_label(OUTPUT_TENSOR_TYPE, USE_THIS_TYPE_FOR_OUTPUT_WEIGHT), output_tensor_layout)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label(OUTPUT_TENSOR_TYPE, USE_THIS_TYPE_FOR_OUTPUT_WEIGHT),
|
||||
output_tensor_layout,
|
||||
)
|
||||
|
||||
self.use_token_embedding_type = QCheckBox(USE_TOKEN_EMBEDDING_TYPE)
|
||||
self.token_embedding_type = QComboBox()
|
||||
self.token_embedding_type.addItems(["F32", "F16", "Q4_0", "Q4_1", "Q5_0", "Q5_1", "Q8_0"])
|
||||
self.token_embedding_type.addItems(
|
||||
["F32", "F16", "Q4_0", "Q4_1", "Q5_0", "Q5_1", "Q8_0"]
|
||||
)
|
||||
self.token_embedding_type.setEnabled(False)
|
||||
self.use_token_embedding_type.toggled.connect(lambda checked: self.token_embedding_type.setEnabled(checked))
|
||||
self.use_token_embedding_type.toggled.connect(
|
||||
lambda checked: self.token_embedding_type.setEnabled(checked)
|
||||
)
|
||||
token_embedding_layout = QHBoxLayout()
|
||||
token_embedding_layout.addWidget(self.use_token_embedding_type)
|
||||
token_embedding_layout.addWidget(self.token_embedding_type)
|
||||
quant_options_layout.addRow(self.create_label(TOKEN_EMBEDDING_TYPE, USE_THIS_TYPE_FOR_TOKEN_EMBEDDINGS), token_embedding_layout)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label(TOKEN_EMBEDDING_TYPE, USE_THIS_TYPE_FOR_TOKEN_EMBEDDINGS),
|
||||
token_embedding_layout,
|
||||
)
|
||||
|
||||
self.keep_split = QCheckBox(KEEP_SPLIT)
|
||||
self.override_kv = QLineEdit()
|
||||
quant_options_layout.addRow(self.create_label("", WILL_GENERATE_QUANTIZED_MODEL_IN_SAME_SHARDS), self.keep_split)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label("", WILL_GENERATE_QUANTIZED_MODEL_IN_SAME_SHARDS),
|
||||
self.keep_split,
|
||||
)
|
||||
|
||||
# KV Override section
|
||||
self.kv_override_widget = QWidget()
|
||||
|
|
@ -209,7 +280,10 @@ def __init__(self):
|
|||
kv_override_main_layout.addWidget(kv_override_scroll)
|
||||
kv_override_main_layout.addWidget(add_override_button)
|
||||
|
||||
quant_options_layout.addRow(self.create_label(KV_OVERRIDES, OVERRIDE_MODEL_METADATA), kv_override_main_layout)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label(KV_OVERRIDES, OVERRIDE_MODEL_METADATA),
|
||||
kv_override_main_layout,
|
||||
)
|
||||
|
||||
quant_options_widget.setLayout(quant_options_layout)
|
||||
quant_options_scroll.setWidget(quant_options_widget)
|
||||
|
|
@ -218,7 +292,10 @@ def __init__(self):
|
|||
|
||||
# Add this after the KV override section
|
||||
self.extra_arguments = QLineEdit()
|
||||
quant_options_layout.addRow(self.create_label(EXTRA_ARGUMENTS, "Additional command-line arguments"), self.extra_arguments)
|
||||
quant_options_layout.addRow(
|
||||
self.create_label(EXTRA_ARGUMENTS, "Additional command-line arguments"),
|
||||
self.extra_arguments,
|
||||
)
|
||||
|
||||
# Quantize button layout
|
||||
quantize_layout = QHBoxLayout()
|
||||
|
|
@ -250,7 +327,10 @@ def __init__(self):
|
|||
imatrix_datafile_layout = QHBoxLayout()
|
||||
imatrix_datafile_layout.addWidget(self.imatrix_datafile)
|
||||
imatrix_datafile_layout.addWidget(self.imatrix_datafile_button)
|
||||
imatrix_layout.addRow(self.create_label(DATA_FILE, INPUT_DATA_FILE_FOR_IMATRIX), imatrix_datafile_layout)
|
||||
imatrix_layout.addRow(
|
||||
self.create_label(DATA_FILE, INPUT_DATA_FILE_FOR_IMATRIX),
|
||||
imatrix_datafile_layout,
|
||||
)
|
||||
|
||||
self.imatrix_model = QLineEdit()
|
||||
self.imatrix_model_button = QPushButton(BROWSE)
|
||||
|
|
@ -258,7 +338,9 @@ def __init__(self):
|
|||
imatrix_model_layout = QHBoxLayout()
|
||||
imatrix_model_layout.addWidget(self.imatrix_model)
|
||||
imatrix_model_layout.addWidget(self.imatrix_model_button)
|
||||
imatrix_layout.addRow(self.create_label(MODEL, MODEL_TO_BE_QUANTIZED), imatrix_model_layout)
|
||||
imatrix_layout.addRow(
|
||||
self.create_label(MODEL, MODEL_TO_BE_QUANTIZED), imatrix_model_layout
|
||||
)
|
||||
|
||||
self.imatrix_output = QLineEdit()
|
||||
self.imatrix_output_button = QPushButton(BROWSE)
|
||||
|
|
@ -266,18 +348,27 @@ def __init__(self):
|
|||
imatrix_output_layout = QHBoxLayout()
|
||||
imatrix_output_layout.addWidget(self.imatrix_output)
|
||||
imatrix_output_layout.addWidget(self.imatrix_output_button)
|
||||
imatrix_layout.addRow(self.create_label(OUTPUT, OUTPUT_PATH_FOR_GENERATED_IMATRIX), imatrix_output_layout)
|
||||
imatrix_layout.addRow(
|
||||
self.create_label(OUTPUT, OUTPUT_PATH_FOR_GENERATED_IMATRIX),
|
||||
imatrix_output_layout,
|
||||
)
|
||||
|
||||
self.imatrix_frequency = QSpinBox()
|
||||
self.imatrix_frequency.setRange(1, 100) # Set the range from 1 to 100
|
||||
self.imatrix_frequency.setValue(1) # Set a default value
|
||||
imatrix_layout.addRow(self.create_label(OUTPUT_FREQUENCY, HOW_OFTEN_TO_SAVE_IMATRIX), self.imatrix_frequency)
|
||||
imatrix_layout.addRow(
|
||||
self.create_label(OUTPUT_FREQUENCY, HOW_OFTEN_TO_SAVE_IMATRIX),
|
||||
self.imatrix_frequency,
|
||||
)
|
||||
|
||||
# Context size input (now a spinbox)
|
||||
self.imatrix_ctx_size = QSpinBox()
|
||||
self.imatrix_ctx_size.setRange(1, 1048576) # Up to one million tokens
|
||||
self.imatrix_ctx_size.setValue(512) # Set a default value
|
||||
imatrix_layout.addRow(self.create_label(CONTEXT_SIZE, CONTEXT_SIZE_FOR_IMATRIX), self.imatrix_ctx_size)
|
||||
imatrix_layout.addRow(
|
||||
self.create_label(CONTEXT_SIZE, CONTEXT_SIZE_FOR_IMATRIX),
|
||||
self.imatrix_ctx_size,
|
||||
)
|
||||
|
||||
# Threads input with slider and spinbox
|
||||
threads_layout = QHBoxLayout()
|
||||
|
|
@ -292,7 +383,9 @@ def __init__(self):
|
|||
|
||||
threads_layout.addWidget(self.threads_slider)
|
||||
threads_layout.addWidget(self.threads_spinbox)
|
||||
imatrix_layout.addRow(self.create_label(THREADS, NUMBER_OF_THREADS_FOR_IMATRIX), threads_layout)
|
||||
imatrix_layout.addRow(
|
||||
self.create_label(THREADS, NUMBER_OF_THREADS_FOR_IMATRIX), threads_layout
|
||||
)
|
||||
|
||||
# GPU Offload for IMatrix (corrected version)
|
||||
gpu_offload_layout = QHBoxLayout()
|
||||
|
|
@ -311,7 +404,9 @@ def __init__(self):
|
|||
gpu_offload_layout.addWidget(self.gpu_offload_slider)
|
||||
gpu_offload_layout.addWidget(self.gpu_offload_spinbox)
|
||||
gpu_offload_layout.addWidget(self.gpu_offload_auto)
|
||||
imatrix_layout.addRow(self.create_label(GPU_OFFLOAD, SET_GPU_OFFLOAD_VALUE), gpu_offload_layout)
|
||||
imatrix_layout.addRow(
|
||||
self.create_label(GPU_OFFLOAD, SET_GPU_OFFLOAD_VALUE), gpu_offload_layout
|
||||
)
|
||||
|
||||
imatrix_generate_button = QPushButton(GENERATE_IMATRIX)
|
||||
imatrix_generate_button.clicked.connect(self.generate_imatrix)
|
||||
|
|
@ -336,7 +431,10 @@ def __init__(self):
|
|||
lora_input_layout = QHBoxLayout()
|
||||
lora_input_layout.addWidget(self.lora_input)
|
||||
lora_input_layout.addWidget(lora_input_button)
|
||||
lora_layout.addRow(self.create_label(LORA_INPUT_PATH, SELECT_LORA_INPUT_DIRECTORY), lora_input_layout)
|
||||
lora_layout.addRow(
|
||||
self.create_label(LORA_INPUT_PATH, SELECT_LORA_INPUT_DIRECTORY),
|
||||
lora_input_layout,
|
||||
)
|
||||
|
||||
self.lora_output = QLineEdit()
|
||||
lora_output_button = QPushButton(BROWSE)
|
||||
|
|
@ -344,13 +442,21 @@ def __init__(self):
|
|||
lora_output_layout = QHBoxLayout()
|
||||
lora_output_layout.addWidget(self.lora_output)
|
||||
lora_output_layout.addWidget(lora_output_button)
|
||||
lora_layout.addRow(self.create_label(LORA_OUTPUT_PATH, SELECT_LORA_OUTPUT_FILE), lora_output_layout)
|
||||
lora_layout.addRow(
|
||||
self.create_label(LORA_OUTPUT_PATH, SELECT_LORA_OUTPUT_FILE),
|
||||
lora_output_layout,
|
||||
)
|
||||
|
||||
# Output Type Dropdown
|
||||
self.lora_output_type_combo = QComboBox()
|
||||
self.lora_output_type_combo.addItems(["GGML", "GGUF"])
|
||||
self.lora_output_type_combo.currentIndexChanged.connect(self.update_base_model_visibility) # Connect to update visibility
|
||||
lora_layout.addRow(self.create_label(OUTPUT_TYPE, SELECT_OUTPUT_TYPE), self.lora_output_type_combo)
|
||||
self.lora_output_type_combo.currentIndexChanged.connect(
|
||||
self.update_base_model_visibility
|
||||
) # Connect to update visibility
|
||||
lora_layout.addRow(
|
||||
self.create_label(OUTPUT_TYPE, SELECT_OUTPUT_TYPE),
|
||||
self.lora_output_type_combo,
|
||||
)
|
||||
|
||||
# Base Model Path (initially hidden)
|
||||
self.base_model_path = QLineEdit()
|
||||
|
|
@ -362,7 +468,10 @@ def __init__(self):
|
|||
self.base_model_widget = QWidget()
|
||||
self.base_model_widget.setLayout(base_model_layout)
|
||||
self.base_model_widget.setVisible(False) # Initially hidden
|
||||
lora_layout.addRow(self.create_label(BASE_MODEL, SELECT_BASE_MODEL_FILE), self.base_model_widget)
|
||||
lora_layout.addRow(
|
||||
self.create_label(BASE_MODEL, SELECT_BASE_MODEL_FILE),
|
||||
self.base_model_widget,
|
||||
)
|
||||
|
||||
lora_convert_button = QPushButton(CONVERT_LORA)
|
||||
lora_convert_button.clicked.connect(self.convert_lora)
|
||||
|
|
@ -381,7 +490,9 @@ def __init__(self):
|
|||
export_lora_model_layout = QHBoxLayout()
|
||||
export_lora_model_layout.addWidget(self.export_lora_model)
|
||||
export_lora_model_layout.addWidget(export_lora_model_button)
|
||||
export_lora_layout.addRow(self.create_label(MODEL, SELECT_MODEL_FILE), export_lora_model_layout)
|
||||
export_lora_layout.addRow(
|
||||
self.create_label(MODEL, SELECT_MODEL_FILE), export_lora_model_layout
|
||||
)
|
||||
|
||||
self.export_lora_output = QLineEdit()
|
||||
export_lora_output_button = QPushButton(BROWSE)
|
||||
|
|
@ -389,7 +500,9 @@ def __init__(self):
|
|||
export_lora_output_layout = QHBoxLayout()
|
||||
export_lora_output_layout.addWidget(self.export_lora_output)
|
||||
export_lora_output_layout.addWidget(export_lora_output_button)
|
||||
export_lora_layout.addRow(self.create_label(OUTPUT, SELECT_OUTPUT_FILE), export_lora_output_layout)
|
||||
export_lora_layout.addRow(
|
||||
self.create_label(OUTPUT, SELECT_OUTPUT_FILE), export_lora_output_layout
|
||||
)
|
||||
|
||||
# GGML LoRA Adapters
|
||||
self.export_lora_adapters = QListWidget()
|
||||
|
|
@ -400,20 +513,28 @@ def __init__(self):
|
|||
buttons_layout = QHBoxLayout()
|
||||
buttons_layout.addWidget(add_adapter_button)
|
||||
adapters_layout.addLayout(buttons_layout)
|
||||
export_lora_layout.addRow(self.create_label(GGML_LORA_ADAPTERS, SELECT_LORA_ADAPTER_FILES), adapters_layout)
|
||||
export_lora_layout.addRow(
|
||||
self.create_label(GGML_LORA_ADAPTERS, SELECT_LORA_ADAPTER_FILES),
|
||||
adapters_layout,
|
||||
)
|
||||
|
||||
# Threads
|
||||
self.export_lora_threads = QSpinBox()
|
||||
self.export_lora_threads.setRange(1, 64)
|
||||
self.export_lora_threads.setValue(8) # Default value
|
||||
export_lora_layout.addRow(self.create_label(THREADS, NUMBER_OF_THREADS_FOR_LORA_EXPORT), self.export_lora_threads)
|
||||
export_lora_layout.addRow(
|
||||
self.create_label(THREADS, NUMBER_OF_THREADS_FOR_LORA_EXPORT),
|
||||
self.export_lora_threads,
|
||||
)
|
||||
|
||||
export_lora_button = QPushButton(EXPORT_LORA)
|
||||
export_lora_button.clicked.connect(self.export_lora)
|
||||
export_lora_layout.addRow(export_lora_button)
|
||||
|
||||
export_lora_group.setLayout(export_lora_layout)
|
||||
right_layout.addWidget(export_lora_group) # Add the Export LoRA group to the right layout
|
||||
right_layout.addWidget(
|
||||
export_lora_group
|
||||
) # Add the Export LoRA group to the right layout
|
||||
|
||||
# Modify the task list to support right-click menu
|
||||
self.task_list.setContextMenuPolicy(Qt.ContextMenuPolicy.CustomContextMenu)
|
||||
|
|
@ -445,14 +566,18 @@ def refresh_backends(self):
|
|||
if valid_backends:
|
||||
for name, path in valid_backends:
|
||||
self.backend_combo.addItem(name, userData=path)
|
||||
self.backend_combo.setEnabled(True) # Enable the combo box if there are valid backends
|
||||
self.backend_combo.setEnabled(
|
||||
True
|
||||
) # Enable the combo box if there are valid backends
|
||||
else:
|
||||
self.backend_combo.addItem(NO_BACKENDS_AVAILABLE)
|
||||
self.backend_combo.setEnabled(False)
|
||||
self.logger.info(FOUND_VALID_BACKENDS.format(self.backend_combo.count()))
|
||||
|
||||
def update_base_model_visibility(self, index):
|
||||
self.base_model_widget.setVisible(self.lora_output_type_combo.itemText(index) == "GGUF")
|
||||
self.base_model_widget.setVisible(
|
||||
self.lora_output_type_combo.itemText(index) == "GGUF"
|
||||
)
|
||||
|
||||
def save_preset(self):
|
||||
self.logger.info(SAVING_PRESET)
|
||||
|
|
@ -469,15 +594,19 @@ def save_preset(self):
|
|||
"use_token_embedding_type": self.use_token_embedding_type.isChecked(),
|
||||
"token_embedding_type": self.token_embedding_type.currentText(),
|
||||
"keep_split": self.keep_split.isChecked(),
|
||||
"kv_overrides": [entry.get_override_string() for entry in self.kv_override_entries],
|
||||
"extra_arguments": self.extra_arguments.text()
|
||||
"kv_overrides": [
|
||||
entry.get_override_string() for entry in self.kv_override_entries
|
||||
],
|
||||
"extra_arguments": self.extra_arguments.text(),
|
||||
}
|
||||
|
||||
file_name, _ = QFileDialog.getSaveFileName(self, SAVE_PRESET, "", JSON_FILES)
|
||||
if file_name:
|
||||
with open(file_name, 'w') as f:
|
||||
with open(file_name, "w") as f:
|
||||
json.dump(preset, f, indent=4)
|
||||
QMessageBox.information(self, PRESET_SAVED, PRESET_SAVED_TO.format(file_name))
|
||||
QMessageBox.information(
|
||||
self, PRESET_SAVED, PRESET_SAVED_TO.format(file_name)
|
||||
)
|
||||
self.logger.info(PRESET_SAVED_TO.format(file_name))
|
||||
|
||||
def load_preset(self):
|
||||
|
|
@ -485,20 +614,30 @@ def load_preset(self):
|
|||
file_name, _ = QFileDialog.getOpenFileName(self, LOAD_PRESET, "", JSON_FILES)
|
||||
if file_name:
|
||||
try:
|
||||
with open(file_name, 'r') as f:
|
||||
with open(file_name, "r") as f:
|
||||
preset = json.load(f)
|
||||
|
||||
self.quant_type.setCurrentText(preset.get("quant_type", ""))
|
||||
self.allow_requantize.setChecked(preset.get("allow_requantize", False))
|
||||
self.leave_output_tensor.setChecked(preset.get("leave_output_tensor", False))
|
||||
self.leave_output_tensor.setChecked(
|
||||
preset.get("leave_output_tensor", False)
|
||||
)
|
||||
self.pure.setChecked(preset.get("pure", False))
|
||||
self.imatrix.setText(preset.get("imatrix", ""))
|
||||
self.include_weights.setText(preset.get("include_weights", ""))
|
||||
self.exclude_weights.setText(preset.get("exclude_weights", ""))
|
||||
self.use_output_tensor_type.setChecked(preset.get("use_output_tensor_type", False))
|
||||
self.output_tensor_type.setCurrentText(preset.get("output_tensor_type", ""))
|
||||
self.use_token_embedding_type.setChecked(preset.get("use_token_embedding_type", False))
|
||||
self.token_embedding_type.setCurrentText(preset.get("token_embedding_type", ""))
|
||||
self.use_output_tensor_type.setChecked(
|
||||
preset.get("use_output_tensor_type", False)
|
||||
)
|
||||
self.output_tensor_type.setCurrentText(
|
||||
preset.get("output_tensor_type", "")
|
||||
)
|
||||
self.use_token_embedding_type.setChecked(
|
||||
preset.get("use_token_embedding_type", False)
|
||||
)
|
||||
self.token_embedding_type.setCurrentText(
|
||||
preset.get("token_embedding_type", "")
|
||||
)
|
||||
self.keep_split.setChecked(preset.get("keep_split", False))
|
||||
self.extra_arguments.setText(preset.get("extra_arguments", ""))
|
||||
|
||||
|
|
@ -508,7 +647,9 @@ def load_preset(self):
|
|||
for override in preset.get("kv_overrides", []):
|
||||
self.add_kv_override(override)
|
||||
|
||||
QMessageBox.information(self, PRESET_LOADED, PRESET_LOADED_FROM.format(file_name))
|
||||
QMessageBox.information(
|
||||
self, PRESET_LOADED, PRESET_LOADED_FROM.format(file_name)
|
||||
)
|
||||
except Exception as e:
|
||||
QMessageBox.critical(self, ERROR, FAILED_TO_LOAD_PRESET.format(str(e)))
|
||||
self.logger.info(PRESET_LOADED_FROM.format(file_name))
|
||||
|
|
@ -520,30 +661,40 @@ def save_task_preset(self, task_item):
|
|||
preset = {
|
||||
"command": thread.command,
|
||||
"backend_path": thread.cwd,
|
||||
"log_file": thread.log_file
|
||||
"log_file": thread.log_file,
|
||||
}
|
||||
file_name, _ = QFileDialog.getSaveFileName(self, SAVE_TASK_PRESET, "", JSON_FILES)
|
||||
file_name, _ = QFileDialog.getSaveFileName(
|
||||
self, SAVE_TASK_PRESET, "", JSON_FILES
|
||||
)
|
||||
if file_name:
|
||||
with open(file_name, 'w') as f:
|
||||
with open(file_name, "w") as f:
|
||||
json.dump(preset, f, indent=4)
|
||||
QMessageBox.information(self, TASK_PRESET_SAVED, TASK_PRESET_SAVED_TO.format(file_name))
|
||||
QMessageBox.information(
|
||||
self, TASK_PRESET_SAVED, TASK_PRESET_SAVED_TO.format(file_name)
|
||||
)
|
||||
break
|
||||
|
||||
def browse_export_lora_model(self):
|
||||
self.logger.info(BROWSING_FOR_EXPORT_LORA_MODEL_FILE)
|
||||
model_file, _ = QFileDialog.getOpenFileName(self, SELECT_MODEL_FILE, "", GGUF_FILES)
|
||||
model_file, _ = QFileDialog.getOpenFileName(
|
||||
self, SELECT_MODEL_FILE, "", GGUF_FILES
|
||||
)
|
||||
if model_file:
|
||||
self.export_lora_model.setText(os.path.abspath(model_file))
|
||||
|
||||
def browse_export_lora_output(self):
|
||||
self.logger.info(BROWSING_FOR_EXPORT_LORA_OUTPUT_FILE)
|
||||
output_file, _ = QFileDialog.getSaveFileName(self, SELECT_OUTPUT_FILE, "", GGUF_FILES)
|
||||
output_file, _ = QFileDialog.getSaveFileName(
|
||||
self, SELECT_OUTPUT_FILE, "", GGUF_FILES
|
||||
)
|
||||
if output_file:
|
||||
self.export_lora_output.setText(os.path.abspath(output_file))
|
||||
|
||||
def add_lora_adapter(self):
|
||||
self.logger.info(ADDING_LORA_ADAPTER)
|
||||
adapter_path, _ = QFileDialog.getOpenFileName(self, SELECT_LORA_ADAPTER_FILE, "", LORA_FILES)
|
||||
adapter_path, _ = QFileDialog.getOpenFileName(
|
||||
self, SELECT_LORA_ADAPTER_FILE, "", LORA_FILES
|
||||
)
|
||||
if adapter_path:
|
||||
# Create a widget to hold the path and scale input
|
||||
adapter_widget = QWidget()
|
||||
|
|
@ -557,7 +708,9 @@ def add_lora_adapter(self):
|
|||
adapter_layout.addWidget(scale_input)
|
||||
|
||||
delete_button = QPushButton(DELETE_ADAPTER)
|
||||
delete_button.clicked.connect(lambda: self.delete_lora_adapter_item(adapter_widget))
|
||||
delete_button.clicked.connect(
|
||||
lambda: self.delete_lora_adapter_item(adapter_widget)
|
||||
)
|
||||
adapter_layout.addWidget(delete_button)
|
||||
|
||||
# Add the widget to the list
|
||||
|
|
@ -568,7 +721,9 @@ def add_lora_adapter(self):
|
|||
|
||||
def browse_base_model(self):
|
||||
self.logger.info(BROWSING_FOR_BASE_MODEL_FOLDER) # Updated log message
|
||||
base_model_folder = QFileDialog.getExistingDirectory(self, SELECT_BASE_MODEL_FOLDER)
|
||||
base_model_folder = QFileDialog.getExistingDirectory(
|
||||
self, SELECT_BASE_MODEL_FOLDER
|
||||
)
|
||||
if base_model_folder:
|
||||
self.base_model_path.setText(os.path.abspath(base_model_folder))
|
||||
|
||||
|
|
@ -608,16 +763,22 @@ def export_lora(self):
|
|||
if not backend_path:
|
||||
raise ValueError(NO_BACKEND_SELECTED)
|
||||
|
||||
command = [os.path.join(backend_path, "llama-export-lora"),
|
||||
"--model", model_path,
|
||||
"--output", output_path]
|
||||
command = [
|
||||
os.path.join(backend_path, "llama-export-lora"),
|
||||
"--model",
|
||||
model_path,
|
||||
"--output",
|
||||
output_path,
|
||||
]
|
||||
|
||||
for adapter_path, adapter_scale in lora_adapters:
|
||||
if adapter_path:
|
||||
if adapter_scale:
|
||||
try:
|
||||
scale_value = float(adapter_scale)
|
||||
command.extend(["--lora-scaled", adapter_path, str(scale_value)])
|
||||
command.extend(
|
||||
["--lora-scaled", adapter_path, str(scale_value)]
|
||||
)
|
||||
except ValueError:
|
||||
raise ValueError(INVALID_LORA_SCALE_VALUE)
|
||||
else:
|
||||
|
|
@ -655,11 +816,17 @@ def restart_task(self, task_item):
|
|||
self.logger.info(RESTARTING_TASK.format(task_item.task_name))
|
||||
for thread in self.quant_threads:
|
||||
if thread.log_file == task_item.log_file:
|
||||
new_thread = QuantizationThread(thread.command, thread.cwd, thread.log_file)
|
||||
new_thread = QuantizationThread(
|
||||
thread.command, thread.cwd, thread.log_file
|
||||
)
|
||||
self.quant_threads.append(new_thread)
|
||||
new_thread.status_signal.connect(task_item.update_status)
|
||||
new_thread.finished_signal.connect(lambda: self.task_finished(new_thread))
|
||||
new_thread.error_signal.connect(lambda err: self.handle_error(err, task_item))
|
||||
new_thread.finished_signal.connect(
|
||||
lambda: self.task_finished(new_thread)
|
||||
)
|
||||
new_thread.error_signal.connect(
|
||||
lambda err: self.handle_error(err, task_item)
|
||||
)
|
||||
new_thread.model_info_signal.connect(self.update_model_info)
|
||||
new_thread.start()
|
||||
task_item.update_status(IN_PROGRESS)
|
||||
|
|
@ -667,14 +834,18 @@ def restart_task(self, task_item):
|
|||
|
||||
def browse_lora_input(self):
|
||||
self.logger.info(BROWSING_FOR_LORA_INPUT_DIRECTORY)
|
||||
lora_input_path = QFileDialog.getExistingDirectory(self, SELECT_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)
|
||||
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))
|
||||
|
||||
|
|
@ -691,7 +862,13 @@ def convert_lora(self):
|
|||
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]
|
||||
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)
|
||||
|
|
@ -708,7 +885,9 @@ def convert_lora(self):
|
|||
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_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)
|
||||
list_item = QListWidgetItem(self.task_list)
|
||||
list_item.setSizeHint(task_item.sizeHint())
|
||||
|
|
@ -716,7 +895,11 @@ def convert_lora(self):
|
|||
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.finished_signal.connect(
|
||||
lambda: self.lora_conversion_finished(
|
||||
thread, lora_input_path, lora_output_path
|
||||
)
|
||||
)
|
||||
thread.error_signal.connect(lambda err: self.handle_error(err, task_item))
|
||||
thread.start()
|
||||
self.logger.info(LORA_CONVERSION_TASK_STARTED)
|
||||
|
|
@ -746,15 +929,28 @@ def download_finished(self, extract_dir):
|
|||
self.download_button.setEnabled(True)
|
||||
self.download_progress.setValue(100)
|
||||
|
||||
if self.cuda_extract_checkbox.isChecked() and self.cuda_extract_checkbox.isVisible():
|
||||
if (
|
||||
self.cuda_extract_checkbox.isChecked()
|
||||
and self.cuda_extract_checkbox.isVisible()
|
||||
):
|
||||
cuda_backend = self.backend_combo_cuda.currentData()
|
||||
if cuda_backend and cuda_backend != NO_SUITABLE_CUDA_BACKENDS:
|
||||
self.extract_cuda_files(extract_dir, cuda_backend)
|
||||
QMessageBox.information(self, DOWNLOAD_COMPLETE, LLAMACPP_DOWNLOADED_AND_EXTRACTED.format(extract_dir, cuda_backend))
|
||||
QMessageBox.information(
|
||||
self,
|
||||
DOWNLOAD_COMPLETE,
|
||||
LLAMACPP_DOWNLOADED_AND_EXTRACTED.format(extract_dir, cuda_backend),
|
||||
)
|
||||
else:
|
||||
QMessageBox.warning(self, CUDA_EXTRACTION_FAILED, NO_SUITABLE_CUDA_BACKEND_FOUND)
|
||||
QMessageBox.warning(
|
||||
self, CUDA_EXTRACTION_FAILED, NO_SUITABLE_CUDA_BACKEND_FOUND
|
||||
)
|
||||
else:
|
||||
QMessageBox.information(self, DOWNLOAD_COMPLETE, LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED.format(extract_dir))
|
||||
QMessageBox.information(
|
||||
self,
|
||||
DOWNLOAD_COMPLETE,
|
||||
LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED.format(extract_dir),
|
||||
)
|
||||
|
||||
self.refresh_backends() # Refresh the backends after successful download
|
||||
self.update_cuda_option() # Update CUDA options in case a CUDA-capable backend was downloaded
|
||||
|
|
@ -768,11 +964,13 @@ def download_finished(self, extract_dir):
|
|||
def refresh_releases(self):
|
||||
self.logger.info(REFRESHING_LLAMACPP_RELEASES)
|
||||
try:
|
||||
response = requests.get("https://api.github.com/repos/ggerganov/llama.cpp/releases")
|
||||
response = requests.get(
|
||||
"https://api.github.com/repos/ggerganov/llama.cpp/releases"
|
||||
)
|
||||
releases = response.json()
|
||||
self.release_combo.clear()
|
||||
for release in releases:
|
||||
self.release_combo.addItem(release['tag_name'], userData=release)
|
||||
self.release_combo.addItem(release["tag_name"], userData=release)
|
||||
self.release_combo.currentIndexChanged.connect(self.update_assets)
|
||||
self.update_assets()
|
||||
except Exception as e:
|
||||
|
|
@ -783,14 +981,14 @@ def update_assets(self):
|
|||
self.asset_combo.clear()
|
||||
release = self.release_combo.currentData()
|
||||
if release:
|
||||
for asset in release['assets']:
|
||||
self.asset_combo.addItem(asset['name'], userData=asset)
|
||||
for asset in release["assets"]:
|
||||
self.asset_combo.addItem(asset["name"], userData=asset)
|
||||
self.update_cuda_option()
|
||||
|
||||
def update_cuda_option(self):
|
||||
self.logger.debug(UPDATING_CUDA_OPTIONS)
|
||||
asset = self.asset_combo.currentData()
|
||||
is_cuda = asset and "cudart" in asset['name'].lower()
|
||||
is_cuda = asset and "cudart" in asset["name"].lower()
|
||||
self.cuda_extract_checkbox.setVisible(is_cuda)
|
||||
self.cuda_backend_label.setVisible(is_cuda)
|
||||
self.backend_combo_cuda.setVisible(is_cuda)
|
||||
|
|
@ -808,9 +1006,9 @@ def download_llama_cpp(self):
|
|||
if not os.path.exists(llama_bin):
|
||||
os.makedirs(llama_bin)
|
||||
|
||||
save_path = os.path.join(llama_bin, asset['name'])
|
||||
save_path = os.path.join(llama_bin, asset["name"])
|
||||
|
||||
self.download_thread = DownloadThread(asset['browser_download_url'], save_path)
|
||||
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)
|
||||
|
|
@ -843,15 +1041,28 @@ def download_finished(self, extract_dir):
|
|||
self.download_button.setEnabled(True)
|
||||
self.download_progress.setValue(100)
|
||||
|
||||
if self.cuda_extract_checkbox.isChecked() and self.cuda_extract_checkbox.isVisible():
|
||||
if (
|
||||
self.cuda_extract_checkbox.isChecked()
|
||||
and self.cuda_extract_checkbox.isVisible()
|
||||
):
|
||||
cuda_backend = self.backend_combo_cuda.currentData()
|
||||
if cuda_backend:
|
||||
self.extract_cuda_files(extract_dir, cuda_backend)
|
||||
QMessageBox.information(self, DOWNLOAD_COMPLETE, LLAMACPP_DOWNLOADED_AND_EXTRACTED.format(extract_dir, cuda_backend))
|
||||
QMessageBox.information(
|
||||
self,
|
||||
DOWNLOAD_COMPLETE,
|
||||
LLAMACPP_DOWNLOADED_AND_EXTRACTED.format(extract_dir, cuda_backend),
|
||||
)
|
||||
else:
|
||||
QMessageBox.warning(self, CUDA_EXTRACTION_FAILED, NO_CUDA_BACKEND_SELECTED)
|
||||
QMessageBox.warning(
|
||||
self, CUDA_EXTRACTION_FAILED, NO_CUDA_BACKEND_SELECTED
|
||||
)
|
||||
else:
|
||||
QMessageBox.information(self, DOWNLOAD_COMPLETE, LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED.format(extract_dir))
|
||||
QMessageBox.information(
|
||||
self,
|
||||
DOWNLOAD_COMPLETE,
|
||||
LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED.format(extract_dir),
|
||||
)
|
||||
|
||||
self.refresh_backends()
|
||||
|
||||
|
|
@ -859,12 +1070,11 @@ def extract_cuda_files(self, extract_dir, destination):
|
|||
self.logger.info(EXTRACTING_CUDA_FILES.format(extract_dir, destination))
|
||||
for root, dirs, files in os.walk(extract_dir):
|
||||
for file in files:
|
||||
if file.lower().endswith('.dll'):
|
||||
if file.lower().endswith(".dll"):
|
||||
source_path = os.path.join(root, file)
|
||||
dest_path = os.path.join(destination, file)
|
||||
shutil.copy2(source_path, dest_path)
|
||||
|
||||
|
||||
def download_error(self, error_message):
|
||||
self.logger.error(DOWNLOAD_ERROR.format(error_message))
|
||||
self.download_button.setEnabled(True)
|
||||
|
|
@ -874,7 +1084,7 @@ def download_error(self, error_message):
|
|||
# Clean up any partially downloaded files
|
||||
asset = self.asset_combo.currentData()
|
||||
if asset:
|
||||
partial_file = os.path.join(os.path.abspath("llama_bin"), asset['name'])
|
||||
partial_file = os.path.join(os.path.abspath("llama_bin"), asset["name"])
|
||||
if os.path.exists(partial_file):
|
||||
os.remove(partial_file)
|
||||
|
||||
|
|
@ -900,7 +1110,9 @@ def show_task_context_menu(self, position):
|
|||
context_menu.addAction(restart_action)
|
||||
|
||||
save_preset_action = QAction(SAVE_PRESET, self)
|
||||
save_preset_action.triggered.connect(lambda: self.save_task_preset(task_item))
|
||||
save_preset_action.triggered.connect(
|
||||
lambda: self.save_task_preset(task_item)
|
||||
)
|
||||
context_menu.addAction(save_preset_action)
|
||||
|
||||
delete_action = QAction(DELETE, self)
|
||||
|
|
@ -951,10 +1163,13 @@ def retry_task(self, item):
|
|||
|
||||
def delete_task(self, item):
|
||||
self.logger.info(DELETING_TASK.format(item.text()))
|
||||
reply = QMessageBox.question(self, CONFIRM_DELETION_TITLE,
|
||||
reply = QMessageBox.question(
|
||||
self,
|
||||
CONFIRM_DELETION_TITLE,
|
||||
CONFIRM_DELETION,
|
||||
QMessageBox.StandardButton.Yes | QMessageBox.StandardButton.No,
|
||||
QMessageBox.StandardButton.No)
|
||||
QMessageBox.StandardButton.No,
|
||||
)
|
||||
if reply == QMessageBox.StandardButton.Yes:
|
||||
row = self.task_list.row(item)
|
||||
self.task_list.takeItem(row)
|
||||
|
|
@ -981,7 +1196,6 @@ def load_models(self):
|
|||
self.model_list.addItem(file)
|
||||
self.logger.info(LOADED_MODELS.format(self.model_list.count()))
|
||||
|
||||
|
||||
def browse_models(self):
|
||||
self.logger.info(BROWSING_FOR_MODELS_DIRECTORY)
|
||||
models_path = QFileDialog.getExistingDirectory(self, SELECT_MODELS_DIRECTORY)
|
||||
|
|
@ -1006,7 +1220,9 @@ def browse_logs(self):
|
|||
|
||||
def browse_imatrix(self):
|
||||
self.logger.info(BROWSING_FOR_IMATRIX_FILE)
|
||||
imatrix_file, _ = QFileDialog.getOpenFileName(self, SELECT_IMATRIX_FILE, "", DAT_FILES)
|
||||
imatrix_file, _ = QFileDialog.getOpenFileName(
|
||||
self, SELECT_IMATRIX_FILE, "", DAT_FILES
|
||||
)
|
||||
if imatrix_file:
|
||||
self.imatrix.setText(os.path.abspath(imatrix_file))
|
||||
|
||||
|
|
@ -1014,7 +1230,11 @@ def update_system_info(self):
|
|||
ram = psutil.virtual_memory()
|
||||
cpu = psutil.cpu_percent()
|
||||
self.ram_bar.setValue(int(ram.percent))
|
||||
self.ram_bar.setFormat(RAM_USAGE_FORMAT.format(ram.percent, ram.used // 1024 // 1024, ram.total // 1024 // 1024))
|
||||
self.ram_bar.setFormat(
|
||||
RAM_USAGE_FORMAT.format(
|
||||
ram.percent, ram.used // 1024 // 1024, ram.total // 1024 // 1024
|
||||
)
|
||||
)
|
||||
self.cpu_label.setText(CPU_USAGE_FORMAT.format(cpu))
|
||||
|
||||
def validate_quantization_inputs(self):
|
||||
|
|
@ -1038,8 +1258,8 @@ def add_kv_override(self, override_string=None):
|
|||
entry = KVOverrideEntry()
|
||||
entry.deleted.connect(self.remove_kv_override)
|
||||
if override_string:
|
||||
key, value = override_string.split('=')
|
||||
type_, val = value.split(':')
|
||||
key, value = override_string.split("=")
|
||||
type_, val = value.split(":")
|
||||
entry.key_input.setText(key)
|
||||
entry.type_combo.setCurrentText(type_)
|
||||
entry.value_input.setText(val)
|
||||
|
|
@ -1070,10 +1290,17 @@ def quantize_model(self):
|
|||
quant_type = self.quant_type.currentText()
|
||||
|
||||
# Start building the output name
|
||||
output_name_parts = [os.path.splitext(model_name)[0], "converted", quant_type]
|
||||
output_name_parts = [
|
||||
os.path.splitext(model_name)[0],
|
||||
"converted",
|
||||
quant_type,
|
||||
]
|
||||
|
||||
# Check for output tensor options
|
||||
if self.use_output_tensor_type.isChecked() or self.leave_output_tensor.isChecked():
|
||||
if (
|
||||
self.use_output_tensor_type.isChecked()
|
||||
or self.leave_output_tensor.isChecked()
|
||||
):
|
||||
output_tensor_part = "o"
|
||||
if self.use_output_tensor_type.isChecked():
|
||||
output_tensor_part += "." + self.output_tensor_type.currentText()
|
||||
|
|
@ -1118,9 +1345,13 @@ def quantize_model(self):
|
|||
if self.exclude_weights.text():
|
||||
command.extend(["--exclude-weights", self.exclude_weights.text()])
|
||||
if self.use_output_tensor_type.isChecked():
|
||||
command.extend(["--output-tensor-type", self.output_tensor_type.currentText()])
|
||||
command.extend(
|
||||
["--output-tensor-type", self.output_tensor_type.currentText()]
|
||||
)
|
||||
if self.use_token_embedding_type.isChecked():
|
||||
command.extend(["--token-embedding-type", self.token_embedding_type.currentText()])
|
||||
command.extend(
|
||||
["--token-embedding-type", self.token_embedding_type.currentText()]
|
||||
)
|
||||
if self.keep_split.isChecked():
|
||||
command.append("--keep-split")
|
||||
if self.override_kv.text():
|
||||
|
|
@ -1139,12 +1370,16 @@ def quantize_model(self):
|
|||
ensure_directory(logs_path)
|
||||
|
||||
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
|
||||
log_file = os.path.join(logs_path, f"{model_name}_{timestamp}_{quant_type}.log")
|
||||
log_file = os.path.join(
|
||||
logs_path, f"{model_name}_{timestamp}_{quant_type}.log"
|
||||
)
|
||||
|
||||
thread = QuantizationThread(command, backend_path, log_file)
|
||||
self.quant_threads.append(thread)
|
||||
|
||||
task_item = TaskListItem(QUANTIZING_MODEL_TO.format(model_name, quant_type), log_file)
|
||||
task_item = TaskListItem(
|
||||
QUANTIZING_MODEL_TO.format(model_name, quant_type), log_file
|
||||
)
|
||||
list_item = QListWidgetItem(self.task_list)
|
||||
list_item.setSizeHint(task_item.sizeHint())
|
||||
self.task_list.addItem(list_item)
|
||||
|
|
@ -1188,7 +1423,7 @@ def show_task_details(self, item):
|
|||
|
||||
# Load existing content
|
||||
if os.path.exists(task_item.log_file):
|
||||
with open_file_safe(task_item.log_file, 'r') as f:
|
||||
with open_file_safe(task_item.log_file, "r") as f:
|
||||
log_text.setPlainText(f.read())
|
||||
|
||||
# Connect to the thread if it's still running
|
||||
|
|
@ -1207,13 +1442,17 @@ def browse_imatrix_datafile(self):
|
|||
|
||||
def browse_imatrix_model(self):
|
||||
self.logger.info(BROWSING_FOR_IMATRIX_MODEL_FILE)
|
||||
model_file, _ = QFileDialog.getOpenFileName(self, SELECT_MODEL_FILE, "", GGUF_FILES)
|
||||
model_file, _ = QFileDialog.getOpenFileName(
|
||||
self, SELECT_MODEL_FILE, "", GGUF_FILES
|
||||
)
|
||||
if model_file:
|
||||
self.imatrix_model.setText(os.path.abspath(model_file))
|
||||
|
||||
def browse_imatrix_output(self):
|
||||
self.logger.info(BROWSING_FOR_IMATRIX_OUTPUT_FILE)
|
||||
output_file, _ = QFileDialog.getSaveFileName(self, SELECT_OUTPUT_FILE, "", DAT_FILES)
|
||||
output_file, _ = QFileDialog.getSaveFileName(
|
||||
self, SELECT_OUTPUT_FILE, "", DAT_FILES
|
||||
)
|
||||
if output_file:
|
||||
self.imatrix_output.setText(os.path.abspath(output_file))
|
||||
|
||||
|
|
@ -1241,12 +1480,18 @@ def generate_imatrix(self):
|
|||
|
||||
command = [
|
||||
os.path.join(backend_path, "llama-imatrix"),
|
||||
"-f", self.imatrix_datafile.text(),
|
||||
"-m", self.imatrix_model.text(),
|
||||
"-o", self.imatrix_output.text(),
|
||||
"--output-frequency", str(self.imatrix_frequency.value()),
|
||||
"--ctx-size", str(self.imatrix_ctx_size.value()),
|
||||
"--threads", str(self.threads_spinbox.value())
|
||||
"-f",
|
||||
self.imatrix_datafile.text(),
|
||||
"-m",
|
||||
self.imatrix_model.text(),
|
||||
"-o",
|
||||
self.imatrix_output.text(),
|
||||
"--output-frequency",
|
||||
str(self.imatrix_frequency.value()),
|
||||
"--ctx-size",
|
||||
str(self.imatrix_ctx_size.value()),
|
||||
"--threads",
|
||||
str(self.threads_spinbox.value()),
|
||||
]
|
||||
|
||||
if self.gpu_offload_auto.isChecked():
|
||||
|
|
@ -1260,7 +1505,9 @@ def generate_imatrix(self):
|
|||
thread = QuantizationThread(command, backend_path, log_file)
|
||||
self.quant_threads.append(thread)
|
||||
|
||||
task_name = GENERATING_IMATRIX_FOR.format(os.path.basename(self.imatrix_model.text()))
|
||||
task_name = GENERATING_IMATRIX_FOR.format(
|
||||
os.path.basename(self.imatrix_model.text())
|
||||
)
|
||||
task_item = TaskListItem(task_name, log_file)
|
||||
list_item = QListWidgetItem(self.task_list)
|
||||
list_item.setSizeHint(task_item.sizeHint())
|
||||
|
|
@ -1287,10 +1534,13 @@ def handle_error(self, error_message, task_item):
|
|||
def closeEvent(self, event: QCloseEvent):
|
||||
self.logger.info(APPLICATION_CLOSING)
|
||||
if self.quant_threads:
|
||||
reply = QMessageBox.question(self, WARNING,
|
||||
reply = QMessageBox.question(
|
||||
self,
|
||||
WARNING,
|
||||
TASK_RUNNING_WARNING,
|
||||
QMessageBox.StandardButton.Yes | QMessageBox.StandardButton.No,
|
||||
QMessageBox.StandardButton.No)
|
||||
QMessageBox.StandardButton.No,
|
||||
)
|
||||
|
||||
if reply == QMessageBox.StandardButton.Yes:
|
||||
for thread in self.quant_threads:
|
||||
|
|
@ -1302,6 +1552,7 @@ def closeEvent(self, event: QCloseEvent):
|
|||
event.accept()
|
||||
self.logger.info(APPLICATION_CLOSED)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
app = QApplication(sys.argv)
|
||||
window = AutoGGUF()
|
||||
|
|
|
|||
Loading…
Reference in New Issue