mirror of https://github.com/leafspark/AutoGGUF
Compare commits
No commits in common. "main" and "v1.9.1" have entirely different histories.
|
|
@ -10,4 +10,3 @@ AUTOGGUF_SERVER=enabled
|
|||
AUTOGGUF_SERVER_PORT=7001
|
||||
AUTOGGUF_SERVER_API_KEY=
|
||||
AUTOGGUF_LANGUAGE=en-US
|
||||
AUTOGGUF_BACKEND_REPO=ggerganov/llama.cpp
|
||||
|
|
|
|||
|
|
@ -18,9 +18,10 @@ jobs:
|
|||
matrix:
|
||||
os: [windows-latest, ubuntu-latest, macos-latest]
|
||||
arch: [x64]
|
||||
include:
|
||||
- os: windows-latest
|
||||
arch: x86
|
||||
runs-on: ${{ matrix.os }}
|
||||
outputs:
|
||||
artifact-names: ${{ steps.set-outputs.outputs.artifact-names }}
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
|
|
@ -66,7 +67,6 @@ jobs:
|
|||
Copy-Item -Path "src\convert_lora_to_gguf.py" -Destination "$distPath\src"
|
||||
Copy-Item -Path "src\convert_lora_to_ggml.py" -Destination "$distPath\src"
|
||||
Copy-Item -Path "src\quantize_to_fp8_dynamic.py" -Destination "$distPath\src"
|
||||
Copy-Item -Path ".env.example" -Destination "$distPath\"
|
||||
|
||||
- name: Copy additional files (Linux/macOS)
|
||||
if: matrix.os != 'windows-latest'
|
||||
|
|
@ -78,58 +78,46 @@ jobs:
|
|||
cp src/convert_lora_to_gguf.py $distPath/src/
|
||||
cp src/convert_lora_to_ggml.py $distPath/src/
|
||||
cp src/quantize_to_fp8_dynamic.py $distPath/src/
|
||||
cp .env.example $distPath/
|
||||
|
||||
- name: Set outputs for artifact name
|
||||
id: set-outputs
|
||||
run: echo "artifact-name=AutoGGUF-${{ matrix.os }}-${{ matrix.arch }}-${{ github.event.inputs.build_type }}-${{ github.sha }}" >> $GITHUB_OUTPUT
|
||||
- name: Generate SHA256 (Windows)
|
||||
if: matrix.os == 'windows-latest'
|
||||
run: |
|
||||
$distPath = if ("${{ github.event.inputs.build_type }}" -eq "RELEASE") { "build\release\dist" } else { "build\dev\dist" }
|
||||
$archSuffix = if ("${{ matrix.arch }}" -eq "x86") { "-x86" } else { "-x64" }
|
||||
$exeName = "AutoGGUF$archSuffix.exe"
|
||||
$versionHash = "${{ github.sha }}".Substring(0, 7)
|
||||
$hashFile = "AutoGGUF-${{ matrix.os }}-${{ matrix.arch }}-$versionHash.sha256"
|
||||
$hash = (Get-FileHash "$distPath\$exeName" -Algorithm SHA256).Hash.ToLower()
|
||||
"$hash *$exeName" | Out-File -FilePath "$distPath\$hashFile" -Encoding utf8
|
||||
|
||||
- name: Generate SHA256 (Linux)
|
||||
if: matrix.os == 'ubuntu-latest'
|
||||
run: |
|
||||
distPath=$(if [ "${{ github.event.inputs.build_type }}" = "RELEASE" ]; then echo "build/release/dist"; else echo "build/dev/dist"; fi)
|
||||
exeName="AutoGGUF-x64"
|
||||
versionHash=$(echo ${{ github.sha }} | cut -c1-7)
|
||||
hashFile="AutoGGUF-${{ matrix.os }}-x64-$versionHash.sha256"
|
||||
cd $distPath && sha256sum $exeName > $hashFile
|
||||
|
||||
- name: Generate SHA256 (macOS)
|
||||
if: matrix.os == 'macos-latest'
|
||||
run: |
|
||||
distPath=$(if [ "${{ github.event.inputs.build_type }}" = "RELEASE" ]; then echo "build/release/dist"; else echo "build/dev/dist"; fi)
|
||||
exeName="AutoGGUF-x64"
|
||||
versionHash=$(echo ${{ github.sha }} | cut -c1-7)
|
||||
hashFile="AutoGGUF-${{ matrix.os }}-x64-$versionHash.sha256"
|
||||
cd $distPath && shasum -a 256 $exeName > $hashFile
|
||||
|
||||
- name: Upload Artifact
|
||||
uses: actions/upload-artifact@v4
|
||||
uses: actions/upload-artifact@v3
|
||||
with:
|
||||
name: AutoGGUF-${{ matrix.os }}-${{ matrix.arch }}-${{ github.event.inputs.build_type }}-${{ github.sha }}
|
||||
path: build/${{ github.event.inputs.build_type == 'RELEASE' && 'release' || 'dev' }}/dist
|
||||
path: |
|
||||
build/${{ github.event.inputs.build_type == 'RELEASE' && 'release' || 'dev' }}/dist
|
||||
!build/${{ github.event.inputs.build_type == 'RELEASE' && 'release' || 'dev' }}/dist/AutoGGUF-*.sha256
|
||||
|
||||
generate-checksums:
|
||||
needs: build
|
||||
runs-on: ubuntu-latest
|
||||
steps:
|
||||
- name: Download all artifacts
|
||||
uses: actions/download-artifact@v4
|
||||
with:
|
||||
path: ./artifacts
|
||||
|
||||
- name: Generate SHA256 checksums for all artifacts
|
||||
run: |
|
||||
cd artifacts
|
||||
versionHash=$(echo ${{ github.sha }} | cut -c1-7)
|
||||
echo "# AutoGGUF Build Checksums" > ../checksums.txt
|
||||
echo "Build: ${{ github.event.inputs.build_type }}" >> ../checksums.txt
|
||||
echo "Commit: ${{ github.sha }}" >> ../checksums.txt
|
||||
echo "Date: $(date -u)" >> ../checksums.txt
|
||||
echo "" >> ../checksums.txt
|
||||
|
||||
# Find all artifact directories and generate checksums of their zip equivalents
|
||||
for artifact_dir in AutoGGUF-*-${{ github.event.inputs.build_type }}-${{ github.sha }}; do
|
||||
if [ -d "$artifact_dir" ]; then
|
||||
echo "Processing $artifact_dir..."
|
||||
cd "$artifact_dir"
|
||||
|
||||
# Create a temporary zip to calculate hash (simulating what GitHub creates)
|
||||
zip -r "../temp_${artifact_dir}.zip" .
|
||||
cd ..
|
||||
|
||||
# Generate SHA256 of the zip file
|
||||
hash=$(sha256sum "temp_${artifact_dir}.zip" | cut -d' ' -f1)
|
||||
echo "${hash} ${artifact_dir}.zip" >> ../checksums.txt
|
||||
|
||||
# Clean up the temporary zip
|
||||
rm "temp_${artifact_dir}.zip"
|
||||
fi
|
||||
done
|
||||
|
||||
- name: Upload checksums
|
||||
uses: actions/upload-artifact@v4
|
||||
- name: Upload SHA256
|
||||
uses: actions/upload-artifact@v3
|
||||
with:
|
||||
name: AutoGGUF-${{ github.sha }}-SHA256
|
||||
path: checksums.txt
|
||||
path: build/${{ github.event.inputs.build_type == 'RELEASE' && 'release' || 'dev' }}/dist/AutoGGUF-*.sha256
|
||||
|
|
|
|||
|
|
@ -52,7 +52,7 @@ jobs:
|
|||
cat requirements.txt >> detailed_report.txt
|
||||
|
||||
- name: Upload audit results
|
||||
uses: actions/upload-artifact@v4
|
||||
uses: actions/upload-artifact@v3
|
||||
with:
|
||||
name: pip-audit-report
|
||||
path: detailed_report.txt
|
||||
|
|
|
|||
71
CHANGELOG.md
71
CHANGELOG.md
|
|
@ -1,50 +1,5 @@
|
|||
# Changelog
|
||||
|
||||
## [v2.0.1] - 2025-05-24
|
||||
|
||||
### Added
|
||||
- Human readable mappings from KV pairs into model properties
|
||||
- certifi library for backend download and update checking
|
||||
- Automated checksums in CI process
|
||||
|
||||
### Changed
|
||||
- Updated llama.cpp backend
|
||||
- Improved backend UI, logging, and task handling
|
||||
- Enhanced display of model properties and cleaner formatting of KV pairs
|
||||
- Updated tensor data formatting and removed redundant KV pairs property
|
||||
- Updated CUDA backend check for latest llama.cpp release format
|
||||
- Global urllib usage implementation
|
||||
- Updated README with more information about patches and updates
|
||||
- Edited quick start instructions
|
||||
- Small file formatting improvements
|
||||
|
||||
### Fixed
|
||||
- Type hints corrections
|
||||
- Build errors in CI
|
||||
- `@upload-artifact` updated to v4
|
||||
|
||||
## [v2.0.0] - 2025-01-27
|
||||
|
||||
### Added
|
||||
- Clipboard support for save/load preset functionality with shift-click option
|
||||
- Support for shift-clicking to get quantization command
|
||||
- AUTOGGUF_BACKEND_REPO environment variable for custom GitHub repository fetching
|
||||
- New HF to GGUF conversion types: `tq1_0` and `tq2_0`
|
||||
|
||||
### Changed
|
||||
- Updated multiple dependencies:
|
||||
- PySide6, PyTorch, Transformers, FastAPI, uvicorn, and other core libraries to their latest compatible versions
|
||||
- Adjusted monitoring intervals from 0.2s to 0.5s
|
||||
- Updated copyright year to 2025
|
||||
- Bundled llama.cpp licensing text in About menu
|
||||
- Removed x86 build matrix from CI
|
||||
- Removed Import Model confirmation dialog
|
||||
|
||||
### Fixed
|
||||
- Resolved PySide6 segfault issue
|
||||
- Fixed error when deleting models from list
|
||||
- Corrected incorrect menu bar name for Load Preset
|
||||
|
||||
## [v1.9.1] - 2024-10-13
|
||||
|
||||
### Added
|
||||
|
|
@ -260,7 +215,7 @@ ### Notes
|
|||
- Fast build: Higher unzipped size (97MB), smaller download (38MB)
|
||||
- Standard build: Created with PyInstaller, medium download and unzipped size (50MB), potentially slower
|
||||
|
||||
## [v1.6.0] - 2024-08-08
|
||||
## [1.6.0] - 2024-08-08
|
||||
|
||||
### Changed
|
||||
- Resolve licensing issues by using PySide6
|
||||
|
|
@ -268,7 +223,7 @@ ### Changed
|
|||
### Added
|
||||
- Add GPU monitoring support for NVIDIA GPUs
|
||||
|
||||
## [v1.5.1] - 2024-08-08
|
||||
## [1.5.1] - 2024-08-08
|
||||
|
||||
### Changed
|
||||
- Refactor localizations to use them in HF conversion area
|
||||
|
|
@ -280,7 +235,7 @@ ### Removed
|
|||
### Added
|
||||
- Support loading *.gguf file types
|
||||
|
||||
## [v1.5.0] - 2024-08-06
|
||||
## [1.5.0] - 2024-08-06
|
||||
|
||||
### Changed
|
||||
- Refactor localizations to use them in HF conversion area
|
||||
|
|
@ -293,7 +248,7 @@ ### Added
|
|||
### Fixed
|
||||
- Fix scaling on low resolution screens, interface now scrolls
|
||||
|
||||
## [v1.4.3] - 2024-08-05
|
||||
## [1.4.3] - 2024-08-05
|
||||
|
||||
### Changed
|
||||
- Updated src file in release to be Black formatted
|
||||
|
|
@ -306,7 +261,7 @@ ### Added
|
|||
- Added model sharding management support
|
||||
- Allow multiple quantization types to be selected and started simultaneously
|
||||
|
||||
## [v1.4.2] - 2024-08-04
|
||||
## [1.4.2] - 2024-08-04
|
||||
|
||||
### Fixed
|
||||
- Resolves bug where Base Model text was shown even when GGML type was selected
|
||||
|
|
@ -315,13 +270,13 @@ ### Fixed
|
|||
### Changed
|
||||
- Minor repository changes
|
||||
|
||||
## [v1.4.1] - 2024-08-04
|
||||
## [1.4.1] - 2024-08-04
|
||||
|
||||
### Added
|
||||
- Dynamic KV Overrides (see wiki: AutoGGUF/wiki/Dynamic-KV-Overrides)
|
||||
- Quantization commands are now printed and logged
|
||||
|
||||
## [v1.4.0] - 2024-08-04
|
||||
## [1.4.0] - 2024-08-04
|
||||
|
||||
### Added
|
||||
- LoRA Conversion:
|
||||
|
|
@ -345,7 +300,7 @@ ### Added
|
|||
- Currently includes src folder with conversion tools
|
||||
- No console window popup
|
||||
|
||||
## [v1.3.1] - 2024-08-04
|
||||
## [1.3.1] - 2024-08-04
|
||||
|
||||
### Added
|
||||
- AUTOGGUF_CHECK_BACKEND environment variable to disable backend check on start
|
||||
|
|
@ -353,7 +308,7 @@ ### Added
|
|||
### Changed
|
||||
- --onefile build with PyInstaller, _internal directory is no longer required
|
||||
|
||||
## [v1.3.0] - 2024-08-03
|
||||
## [1.3.0] - 2024-08-03
|
||||
|
||||
### Added
|
||||
- Support for new llama-imatrix parameters:
|
||||
|
|
@ -375,7 +330,7 @@ ### Fixed
|
|||
### Removed
|
||||
- Duplicated functions
|
||||
|
||||
## [v1.2.1] - 2024-08-03
|
||||
## [1.2.1] - 2024-08-03
|
||||
|
||||
### Added
|
||||
- Refresh Models button
|
||||
|
|
@ -384,13 +339,13 @@ ### Added
|
|||
### Fixed
|
||||
- iostream llama.cpp issue, quantized_models directory created on launch
|
||||
|
||||
## [v1.2.0] - 2024-08-03
|
||||
## [1.2.0] - 2024-08-03
|
||||
|
||||
### Added
|
||||
- More robust logging (find logs at latest_<timestamp>.log in logs folder)
|
||||
- Localizations with support for 28 languages (machine translated using Gemini Experimental 0801)
|
||||
|
||||
## [v1.1.0] - 2024-08-03
|
||||
## [1.1.0] - 2024-08-03
|
||||
|
||||
### Added
|
||||
- Dynamic KV override functionality
|
||||
|
|
@ -413,7 +368,7 @@ ### Added
|
|||
### Fixed
|
||||
- Issue where quantization errored with "AutoGGUF does not have x attribute"
|
||||
|
||||
## [v1.0.0] - 2024-08-02
|
||||
## [1.0.0] - 2024-08-02
|
||||
|
||||
### Added
|
||||
- Initial release
|
||||
|
|
|
|||
2
LICENSE
2
LICENSE
|
|
@ -186,7 +186,7 @@
|
|||
same "printed page" as the copyright notice for easier
|
||||
identification within third-party archives.
|
||||
|
||||
Copyright (c) 2024-2025 leafspark
|
||||
Copyright 2024 leafspark
|
||||
|
||||
Licensed under the Apache License, Version 2.0 (the "License");
|
||||
you may not use this file except in compliance with the License.
|
||||
|
|
|
|||
90
README.md
90
README.md
|
|
@ -9,55 +9,44 @@ # AutoGGUF - automated GGUF model quantizer
|
|||
|
||||
<!-- Project Info -->
|
||||
[](https://github.com/ggerganov/llama.cpp)
|
||||
|
||||
[]()
|
||||
[](https://github.com/leafspark/AutoGGUF/blob/main/LICENSE)
|
||||
|
||||
|
||||
<!-- Repository Stats -->
|
||||
|
||||
|
||||
|
||||
|
||||
<!--  -->
|
||||
|
||||
|
||||
<!-- Contribution -->
|
||||
[](https://github.com/leafspark/AutoGGUF/issues)
|
||||
[](https://github.com/psf/black)
|
||||
[](https://github.com/leafspark/AutoGGUF/issues)
|
||||
[](https://github.com/leafspark/AutoGGUF/pulls)
|
||||
|
||||
The most comprehensive GUI tool for GGUF model quantization. Stop wrestling with command lines - quantize, merge, and optimize your models with just a few clicks.
|
||||
AutoGGUF provides a graphical user interface for quantizing GGUF models using the llama.cpp library. It allows users to download different versions of llama.cpp, manage multiple backends, and perform quantization tasks with various options.
|
||||
|
||||
## Features
|
||||
|
||||
- 📩 Update and manage llama.cpp backends
|
||||
- 🗃️ Download and quantize GGUF/safetensors models
|
||||
- 📐 Configure quantization parameters
|
||||
- 💻 Monitor system resources in real time during quantization
|
||||
- ⏳ Parallel quantization + imatrix generation
|
||||
- 🎉 LoRA conversion and merging
|
||||
- 📁 Preset saving and loading
|
||||
- 8️⃣ AutoFP8 quantization
|
||||
- 🪓 GGUF splitting and merging
|
||||
- 🌐 HTTP API for automation and monitoring
|
||||
- Download and manage llama.cpp backends
|
||||
- Select and quantize GGUF models
|
||||
- Configure quantization parameters
|
||||
- Monitor system resources during quantization
|
||||
- Parallel quantization + imatrix generation
|
||||
- LoRA conversion and merging
|
||||
- Preset saving and loading
|
||||
- AutoFP8 quantization
|
||||
- GGUF splitting and merging
|
||||
|
||||
## Why AutoGGUF?
|
||||
- Fast: Saves time on manual configuration
|
||||
- Simple: Clean UI, no terminal needed
|
||||
- Powerful: Handles models up to infinite size, only limited by your RAM
|
||||
- Resource-aware: Optimized memory management and efficient UI library
|
||||
## Usage
|
||||
|
||||
|
||||
|
||||
## Quick Start
|
||||
|
||||
### Cross-platform (recommended)
|
||||
1. `git clone https://github.com/leafspark/AutoGGUF`
|
||||
2. `cd AutoGGUF`
|
||||
3. Install dependencies:
|
||||
### Cross-platform
|
||||
1. Install dependencies:
|
||||
```
|
||||
pip install -r requirements.txt
|
||||
```
|
||||
4. Run the application:
|
||||
2. Run the application:
|
||||
```
|
||||
python src/main.py
|
||||
```
|
||||
|
|
@ -65,7 +54,7 @@ ### Cross-platform (recommended)
|
|||
|
||||
macOS and Ubuntu builds are provided with GitHub Actions, you may download the binaries in the releases section.
|
||||
|
||||
### Windows (for the impatient)
|
||||
### Windows
|
||||
Standard builds:
|
||||
1. Download the latest release
|
||||
2. Extract all files to a folder
|
||||
|
|
@ -73,13 +62,13 @@ ### Windows (for the impatient)
|
|||
4. Any necessary folders will be automatically created
|
||||
|
||||
Setup builds:
|
||||
1. Download the setup variant of latest release
|
||||
1. Download setup variant of latest release
|
||||
2. Extract all files to a folder
|
||||
3. Run the setup program
|
||||
4. The .gguf extension will be registered with the program automatically
|
||||
4. The .GGUF extension will be registered with the program automatically
|
||||
5. Run the program from the Start Menu or desktop shortcuts
|
||||
|
||||
After launching the program, you may access its local server at port 7001 (set `AUTOGGUF_SERVER` to "enabled" first).
|
||||
After launching the program, you may access its local server at port 7001 (set `AUTOGGUF_SERVER` to "enabled" first)
|
||||
|
||||
### Verifying Releases
|
||||
|
||||
|
|
@ -121,50 +110,47 @@ ### Windows
|
|||
pip install -U pyinstaller
|
||||
build RELEASE | DEV
|
||||
```
|
||||
Find the executable in `build/<type>/dist/AutoGGUF-x64.exe`.
|
||||
Find the executable in `build/<type>/dist/AutoGGUF.exe`.
|
||||
|
||||
You can also use Nuitka, which may result in a slower build but a faster output executable:
|
||||
```bash
|
||||
build_optimized RELEASE | DEV
|
||||
```
|
||||
|
||||
## Dependencies
|
||||
|
||||
Find them in `requirements.txt`.
|
||||
|
||||
## Localizations
|
||||
|
||||
View the list of supported languages at [AutoGGUF/wiki/Installation#configuration](https://github.com/leafspark/AutoGGUF/wiki/Installation#configuration) (LLM translated, except for English).
|
||||
|
||||
Languages will be updated as soon as possible after an update, or as a part of the update.
|
||||
|
||||
To use a specific language, set the `AUTOGGUF_LANGUAGE` environment variable to one of the listed language codes (note: some languages may not be fully supported yet, in which the UI elements will fall back to English).
|
||||
To use a specific language, set the `AUTOGGUF_LANGUAGE` environment variable to one of the listed language codes (note: some languages may not be fully supported yet, those will fall back to English).
|
||||
|
||||
## Issues
|
||||
|
||||
- Some inconsistent logging and signal handling
|
||||
- Missing or duplicated translations (priority)
|
||||
- Buggy/incomplete API interfaces
|
||||
- Code review and formatting (priority)
|
||||
- Some inconsistent logging
|
||||
|
||||
## Planned Features
|
||||
|
||||
- [ ] Time estimation for quantization
|
||||
- [ ] Quantization file size estimate
|
||||
- [ ] Perplexity testing
|
||||
- [ ] bitsandbytes support
|
||||
- Time estimation for quantization
|
||||
- Quantization file size estimate
|
||||
- Perplexity testing
|
||||
- HuggingFace upload/download (coming in the next release)
|
||||
- bitsandbytes (coming soon)
|
||||
|
||||
#### Project Status
|
||||
|
||||
AutoGGUF has now entered maintenance mode. It's considered stable and feature-complete for most use cases, so I'm not actively developing new features, but I’ll continue to publish occasional builds, update dependencies regularly, and fix critical bugs as needed. If you encounter issues or have suggestions, feel free to open an issue.
|
||||
|
||||
## Support
|
||||
## Troubleshooting
|
||||
|
||||
- SSL module cannot be found error: Install OpenSSL or run from source using `python src/main.py` with the `run.bat` script (`pip install requests`)
|
||||
- Check out the [Wiki](https://github.com/leafspark/AutoGGUF/wiki) for advanced usage and configuration
|
||||
|
||||
## Contributing
|
||||
|
||||
Fork the repo, make your changes, and ensure you have the latest commits when merging. Include a changelog of new features in your pull request description. Read `CONTRIBUTING.md` for more information.
|
||||
|
||||
## User Interface
|
||||
|
||||
|
||||
|
||||
## Stargazers
|
||||
|
||||
[](https://star-history.com/#leafspark/AutoGGUF&Date)
|
||||
|
||||
`Last Updated: May 24, 2025`
|
||||
|
|
|
|||
|
|
@ -4,10 +4,10 @@ ## Supported Versions
|
|||
|
||||
| Version | Supported |
|
||||
|-----------------|--------------------|
|
||||
| stable (v2.0.x) | :white_check_mark: |
|
||||
| stable (v1.9.x) | :white_check_mark: |
|
||||
|
||||
Beta versions are not officially supported and may contain unknown security vulnerabilities. Use them at your own risk.
|
||||
|
||||
## Reporting a Vulnerability
|
||||
|
||||
Use the Issues tab, or for severe vulnerabilities, please contact the maintainers via email.
|
||||
Use the Issues tab, or for severe vulnerabilities please contact the maintainers via email.
|
||||
|
|
|
|||
|
|
@ -1,14 +1,13 @@
|
|||
PyYAML~=6.0.2
|
||||
psutil~=7.0.0
|
||||
pynvml~=12.0.0
|
||||
PySide6~=6.9.1
|
||||
safetensors~=0.5.3
|
||||
psutil~=6.0.0
|
||||
pynvml~=11.5.3
|
||||
PySide6~=6.8.0
|
||||
safetensors~=0.4.5
|
||||
numpy<2.0.0
|
||||
torch~=2.7.0
|
||||
torch~=2.4.1
|
||||
sentencepiece~=0.2.0
|
||||
setuptools~=80.7.1
|
||||
huggingface-hub~=0.33.1
|
||||
transformers~=4.51.3
|
||||
fastapi~=0.115.12
|
||||
uvicorn~=0.34.2
|
||||
certifi~=2025.4.26
|
||||
setuptools~=75.1.0
|
||||
huggingface-hub~=0.25.2
|
||||
transformers~=4.45.1
|
||||
fastapi~=0.115.2
|
||||
uvicorn~=0.31.1
|
||||
|
|
|
|||
4
setup.py
4
setup.py
|
|
@ -5,12 +5,12 @@
|
|||
|
||||
setup(
|
||||
name="AutoGGUF",
|
||||
version="v2.0.1",
|
||||
version="v1.9.0",
|
||||
packages=[""],
|
||||
url="https://github.com/leafspark/AutoGGUF",
|
||||
license="apache-2.0",
|
||||
author="leafspark",
|
||||
author_email="leafspark@proton.me",
|
||||
author_email="",
|
||||
description="automatically quant GGUF models",
|
||||
install_requires=required,
|
||||
entry_points={"console_scripts": ["autogguf-gui = main:main"]},
|
||||
|
|
|
|||
119
src/AutoGGUF.py
119
src/AutoGGUF.py
|
|
@ -1,9 +1,8 @@
|
|||
import json
|
||||
import os
|
||||
import shutil
|
||||
import urllib.error
|
||||
import urllib.request
|
||||
import certifi
|
||||
import ssl
|
||||
from datetime import datetime
|
||||
from functools import partial, wraps
|
||||
from typing import List
|
||||
|
|
@ -72,7 +71,7 @@ def __init__(self, args: List[str]) -> None:
|
|||
|
||||
self.parse_resolution = ui_update.parse_resolution.__get__(self)
|
||||
|
||||
self.log_dir_name = str(os.environ.get("AUTOGGUF_LOG_DIR_NAME", "logs"))
|
||||
self.log_dir_name = os.environ.get("AUTOGGUF_LOG_DIR_NAME", "logs")
|
||||
|
||||
width, height = self.parse_resolution()
|
||||
self.logger = Logger("AutoGGUF", self.log_dir_name)
|
||||
|
|
@ -195,7 +194,7 @@ def __init__(self, args: List[str]) -> None:
|
|||
save_preset_action = QAction(f"&{SAVE_PRESET}", self)
|
||||
save_preset_action.setShortcut(QKeySequence("Ctrl+S"))
|
||||
save_preset_action.triggered.connect(self.save_preset)
|
||||
load_preset_action = QAction(f"&{LOAD_PRESET}", self)
|
||||
load_preset_action = QAction(f"&{SAVE_PRESET}", self)
|
||||
load_preset_action.setShortcut(QKeySequence("Ctrl+S"))
|
||||
load_preset_action.triggered.connect(self.load_preset)
|
||||
file_menu.addAction(close_action)
|
||||
|
|
@ -340,15 +339,15 @@ def __init__(self, args: List[str]) -> None:
|
|||
output_layout.addWidget(output_button)
|
||||
self.merge_gguf_layout.addLayout(output_layout)
|
||||
|
||||
# Merge button
|
||||
merge_button = QPushButton(MERGE_GGUF)
|
||||
merge_button.clicked.connect(
|
||||
# Split button
|
||||
split_button = QPushButton(MERGE_GGUF)
|
||||
split_button.clicked.connect(
|
||||
lambda: self.merge_gguf(
|
||||
self.merge_gguf_input.text(),
|
||||
self.merge_gguf_output.text(),
|
||||
)
|
||||
)
|
||||
self.merge_gguf_layout.addWidget(merge_button)
|
||||
self.merge_gguf_layout.addWidget(split_button)
|
||||
self.merge_gguf_dialog.setLayout(self.merge_gguf_layout)
|
||||
|
||||
# HF Upload Window
|
||||
|
|
@ -501,7 +500,7 @@ def __init__(self, args: List[str]) -> None:
|
|||
# Timer for updating system info
|
||||
self.timer = QTimer()
|
||||
self.timer.timeout.connect(self.update_system_info)
|
||||
self.timer.start(500)
|
||||
self.timer.start(200)
|
||||
|
||||
# Backend selection
|
||||
backend_layout = QHBoxLayout()
|
||||
|
|
@ -764,7 +763,7 @@ def __init__(self, args: List[str]) -> None:
|
|||
|
||||
self.extra_arguments = QLineEdit()
|
||||
quant_options_layout.addRow(
|
||||
self.create_label(EXTRA_ARGUMENTS, EXTRA_ARGUMENTS_LABEL),
|
||||
self.create_label(EXTRA_ARGUMENTS, EXTRA_COMMAND_ARGUMENTS),
|
||||
self.extra_arguments,
|
||||
)
|
||||
|
||||
|
|
@ -776,7 +775,7 @@ def __init__(self, args: List[str]) -> None:
|
|||
# Quantize button layout
|
||||
quantize_layout = QHBoxLayout()
|
||||
quantize_button = QPushButton(QUANTIZE_MODEL)
|
||||
quantize_button.clicked[bool].connect(self.quantize_model_handler)
|
||||
quantize_button.clicked.connect(self.quantize_model)
|
||||
save_preset_button = QPushButton(SAVE_PRESET)
|
||||
save_preset_button.clicked.connect(self.save_preset)
|
||||
load_preset_button = QPushButton(LOAD_PRESET)
|
||||
|
|
@ -1024,9 +1023,7 @@ def __init__(self, args: List[str]) -> None:
|
|||
hf_to_gguf_layout.addRow(OUTPUT_FILE, hf_outfile_layout)
|
||||
|
||||
self.hf_outtype = QComboBox()
|
||||
self.hf_outtype.addItems(
|
||||
["f32", "f16", "bf16", "q8_0", "tq1_0", "tq2_0", "auto"]
|
||||
)
|
||||
self.hf_outtype.addItems(["f32", "f16", "bf16", "q8_0", "auto"])
|
||||
hf_to_gguf_layout.addRow(OUTPUT_TYPE, self.hf_outtype)
|
||||
|
||||
self.hf_vocab_only = QCheckBox(VOCAB_ONLY)
|
||||
|
|
@ -1104,20 +1101,6 @@ def __init__(self, args: List[str]) -> None:
|
|||
self.logger.info(AUTOGGUF_INITIALIZATION_COMPLETE)
|
||||
self.logger.info(STARTUP_ELASPED_TIME.format(init_timer.elapsed()))
|
||||
|
||||
def quantize_model_handler(self) -> None:
|
||||
if QApplication.keyboardModifiers() == Qt.ShiftModifier and self.quantize_model(
|
||||
return_command=True
|
||||
):
|
||||
QApplication.clipboard().setText(self.quantize_model(return_command=True))
|
||||
QMessageBox.information(
|
||||
None,
|
||||
INFO,
|
||||
f"{COPIED_COMMAND_TO_CLIPBOARD} "
|
||||
+ f"<code style='font-family: monospace; white-space: pre;'>{self.quantize_model(return_command=True)}</code>",
|
||||
)
|
||||
else:
|
||||
self.quantize_model()
|
||||
|
||||
def resizeEvent(self, event) -> None:
|
||||
super().resizeEvent(event)
|
||||
path = QPainterPath()
|
||||
|
|
@ -1130,7 +1113,7 @@ def delete_model(self, item):
|
|||
reply = QMessageBox.question(
|
||||
self,
|
||||
CONFIRM_DELETE,
|
||||
DELETE_MODEL_WARNING.format(model_name),
|
||||
DELETE_WARNING.format(model_name),
|
||||
QMessageBox.StandardButton.Yes | QMessageBox.StandardButton.No,
|
||||
QMessageBox.StandardButton.No,
|
||||
)
|
||||
|
|
@ -1143,17 +1126,14 @@ def delete_model(self, item):
|
|||
)
|
||||
self.logger.info(MODEL_DELETED_SUCCESSFULLY.format(model_name))
|
||||
except Exception as e:
|
||||
show_error(self.logger, ERROR_DELETING_MODEL.format(e))
|
||||
show_error(self.logger, f"Error deleting model: {e}")
|
||||
|
||||
def check_for_updates(self) -> None:
|
||||
try:
|
||||
url = "https://api.github.com/repos/leafspark/AutoGGUF/releases/latest"
|
||||
req = urllib.request.Request(url)
|
||||
|
||||
# Create SSL context with certifi certificates
|
||||
ssl_context = ssl.create_default_context(cafile=certifi.where())
|
||||
|
||||
with urllib.request.urlopen(req, context=ssl_context) as response:
|
||||
with urllib.request.urlopen(req) as response:
|
||||
if response.status != 200:
|
||||
raise urllib.error.HTTPError(
|
||||
url, response.status, "HTTP Error", response.headers, None
|
||||
|
|
@ -1206,25 +1186,15 @@ def refresh_backends(self) -> None:
|
|||
and "cudart-llama" not in item.lower()
|
||||
]
|
||||
|
||||
def extract_b_val(name: str) -> int:
|
||||
match = re.search(r"b(\d+)", name)
|
||||
return int(match.group(1)) if match else -1
|
||||
|
||||
if valid_backends:
|
||||
# Sort by newest version
|
||||
valid_backends.sort(key=lambda x: extract_b_val(x[0]), reverse=True)
|
||||
|
||||
for name, path in valid_backends:
|
||||
self.backend_combo.addItem(name, userData=path)
|
||||
|
||||
self.backend_combo.setEnabled(True)
|
||||
|
||||
# Selects the newest version (now at index 0)
|
||||
self.backend_combo.setCurrentIndex(0)
|
||||
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(len(valid_backends)))
|
||||
|
||||
def save_task_preset(self, task_item) -> None:
|
||||
|
|
@ -1266,13 +1236,13 @@ def download_finished(self, extract_dir) -> None:
|
|||
)
|
||||
else:
|
||||
QMessageBox.warning(
|
||||
self, CUDA_EXTRACTION_FAILED, NO_SUITABLE_CUDA_BACKEND_EXTRACTION
|
||||
self, CUDA_EXTRACTION_FAILED, NO_SUITABLE_CUDA_BACKEND_FOUND
|
||||
)
|
||||
else:
|
||||
QMessageBox.information(
|
||||
self,
|
||||
DOWNLOAD_COMPLETE,
|
||||
LLAMACPP_DOWNLOADED_AND_EXTRACTED.format(extract_dir),
|
||||
LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED.format(extract_dir),
|
||||
)
|
||||
|
||||
self.refresh_backends() # Refresh the backends after successful download
|
||||
|
|
@ -1284,6 +1254,23 @@ def download_finished(self, extract_dir) -> None:
|
|||
if index >= 0:
|
||||
self.backend_combo.setCurrentIndex(index)
|
||||
|
||||
def validate_quantization_inputs(self) -> None:
|
||||
self.logger.debug(VALIDATING_QUANTIZATION_INPUTS)
|
||||
errors = []
|
||||
if not self.backend_combo.currentData():
|
||||
errors.append(NO_BACKEND_SELECTED)
|
||||
if not self.models_input.text():
|
||||
errors.append(MODELS_PATH_REQUIRED)
|
||||
if not self.output_input.text():
|
||||
errors.append(OUTPUT_PATH_REQUIRED)
|
||||
if not self.logs_input.text():
|
||||
errors.append(LOGS_PATH_REQUIRED)
|
||||
if not self.model_tree.currentItem():
|
||||
errors.append(NO_MODEL_SELECTED)
|
||||
|
||||
if errors:
|
||||
raise ValueError("\n".join(errors))
|
||||
|
||||
def load_models(self) -> None:
|
||||
self.logger.info(LOADING_MODELS)
|
||||
models_dir = self.models_input.text()
|
||||
|
|
@ -1711,9 +1698,10 @@ def merge_gguf(self, model_dir: str, output_dir: str) -> None:
|
|||
show_error(self.logger, "Error starting merge GGUF task: {}".format(e))
|
||||
self.logger.info("Split GGUF task finished.")
|
||||
|
||||
def quantize_model(self, return_command=False) -> str:
|
||||
def quantize_model(self) -> None:
|
||||
self.logger.info(STARTING_MODEL_QUANTIZATION)
|
||||
try:
|
||||
self.validate_quantization_inputs()
|
||||
selected_item = self.model_tree.currentItem()
|
||||
if not selected_item:
|
||||
raise ValueError(NO_MODEL_SELECTED)
|
||||
|
|
@ -1834,12 +1822,6 @@ def quantize_model(self, return_command=False) -> str:
|
|||
if self.extra_arguments.text():
|
||||
command.extend(self.extra_arguments.text().split())
|
||||
|
||||
if return_command:
|
||||
self.logger.info(
|
||||
f"{QUANTIZATION_COMMAND}: {str(' '.join(command))}"
|
||||
)
|
||||
return str(" ".join(command))
|
||||
|
||||
logs_path = self.logs_input.text()
|
||||
ensure_directory(logs_path)
|
||||
|
||||
|
|
@ -1920,25 +1902,12 @@ def show_task_details(self, item) -> None:
|
|||
# Load existing content
|
||||
if os.path.exists(task_item.log_file):
|
||||
with open_file_safe(task_item.log_file, "r") as f:
|
||||
content = f.read().rstrip("\n") # Remove trailing newlines
|
||||
log_text.setPlainText(content)
|
||||
|
||||
# Scroll to the end
|
||||
log_text.moveCursor(QTextCursor.End)
|
||||
log_text.setPlainText(f.read())
|
||||
|
||||
# Connect to the thread if it's still running
|
||||
for thread in self.quant_threads:
|
||||
if thread.log_file == task_item.log_file:
|
||||
# Create a local slot function that updates the text
|
||||
def update_log(text):
|
||||
log_text.appendPlainText(text)
|
||||
log_text.moveCursor(QTextCursor.End)
|
||||
|
||||
thread.output_signal.connect(update_log)
|
||||
# Disconnect the signal when the dialog is destroyed
|
||||
log_dialog.destroyed.connect(
|
||||
lambda: thread.output_signal.disconnect(update_log)
|
||||
)
|
||||
thread.output_signal.connect(log_text.appendPlainText)
|
||||
break
|
||||
|
||||
log_dialog.exec()
|
||||
|
|
@ -1956,6 +1925,14 @@ def import_model(self) -> None:
|
|||
show_error(self.logger, INVALID_GGUF_FILE.format(file_name))
|
||||
return
|
||||
|
||||
reply = QMessageBox.question(
|
||||
self,
|
||||
CONFIRM_IMPORT,
|
||||
IMPORT_MODEL_CONFIRMATION.format(file_name),
|
||||
QMessageBox.StandardButton.Yes | QMessageBox.StandardButton.No,
|
||||
QMessageBox.StandardButton.No,
|
||||
)
|
||||
if reply == QMessageBox.StandardButton.Yes:
|
||||
self.imported_models.append(file_path)
|
||||
self.load_models()
|
||||
self.logger.info(MODEL_IMPORTED_SUCCESSFULLY.format(file_name))
|
||||
|
|
|
|||
|
|
@ -98,7 +98,7 @@ def mouseMoveEvent(self, event) -> None:
|
|||
def mouseReleaseEvent(self, event) -> None:
|
||||
self.pressing = False
|
||||
|
||||
def toggle_maximize(self) -> None:
|
||||
def toggle_maximize(self):
|
||||
if self.isMaximized:
|
||||
self.parent.showNormal()
|
||||
if self.normal_size:
|
||||
|
|
|
|||
|
|
@ -2,8 +2,6 @@
|
|||
import urllib.request
|
||||
import urllib.error
|
||||
import zipfile
|
||||
import ssl
|
||||
import certifi
|
||||
from PySide6.QtCore import QThread, Signal
|
||||
|
||||
|
||||
|
|
@ -21,10 +19,7 @@ def run(self) -> None:
|
|||
try:
|
||||
req = urllib.request.Request(self.url)
|
||||
|
||||
# Create SSL context with certifi certificates
|
||||
ssl_context = ssl.create_default_context(cafile=certifi.where())
|
||||
|
||||
with urllib.request.urlopen(req, context=ssl_context) as response:
|
||||
with urllib.request.urlopen(req) as response:
|
||||
if response.status != 200:
|
||||
raise urllib.error.HTTPError(
|
||||
self.url, response.status, "HTTP Error", response.headers, None
|
||||
|
|
|
|||
|
|
@ -95,7 +95,7 @@ def __init__(self, parent=None) -> None:
|
|||
|
||||
self.timer = QTimer(self)
|
||||
self.timer.timeout.connect(self.update_gpu_info)
|
||||
self.timer.start(500) # Update every 0.5 seconds
|
||||
self.timer.start(200) # Update every 0.2 seconds
|
||||
|
||||
self.gpu_data = []
|
||||
self.vram_data = []
|
||||
|
|
@ -192,7 +192,7 @@ def update_graph_data() -> None:
|
|||
|
||||
timer = QTimer(dialog)
|
||||
timer.timeout.connect(update_graph_data)
|
||||
timer.start(500) # Update every 0.5 seconds
|
||||
timer.start(200) # Update every 0.2 seconds
|
||||
|
||||
dialog.exec()
|
||||
|
||||
|
|
@ -227,7 +227,7 @@ def update_graph_data() -> None:
|
|||
|
||||
timer = QTimer(dialog)
|
||||
timer.timeout.connect(update_graph_data)
|
||||
timer.start(500) # Update every 0.5 seconds
|
||||
timer.start(200) # Update every 0.2 seconds
|
||||
|
||||
tab_widget.addTab(gpu_graph, GPU_USAGE_OVER_TIME)
|
||||
tab_widget.addTab(vram_graph, VRAM_USAGE_OVER_TIME)
|
||||
|
|
|
|||
|
|
@ -22,7 +22,6 @@ def __init__(self, parent=None) -> None:
|
|||
|
||||
self.key_input = QLineEdit()
|
||||
self.key_input.setPlaceholderText("Key")
|
||||
|
||||
# Set validator for key input (letters and dots only)
|
||||
key_validator = QRegularExpressionValidator(QRegularExpression(r"[A-Za-z.]+"))
|
||||
self.key_input.setValidator(key_validator)
|
||||
|
|
|
|||
|
|
@ -1,7 +1,7 @@
|
|||
import os
|
||||
import re
|
||||
|
||||
AUTOGGUF_VERSION = "v2.0.1"
|
||||
AUTOGGUF_VERSION = "v2.0.0"
|
||||
|
||||
|
||||
class _Localization:
|
||||
|
|
@ -53,11 +53,13 @@ def __init__(self):
|
|||
self.QUANTIZE_TO_FP8_DYNAMIC = "Quantize to FP8 Dynamic"
|
||||
self.OPEN_MODEL_FOLDER = "Open Model Folder"
|
||||
self.QUANTIZE = "Quantize"
|
||||
self.OPEN_MODEL_FOLDER = "Open Model Folder"
|
||||
self.INPUT_MODEL = "Input Model:"
|
||||
|
||||
# GGUF Verification
|
||||
self.INVALID_GGUF_FILE = "Invalid GGUF file: {}"
|
||||
self.SHARDED_MODEL_NAME = "{} (Sharded)"
|
||||
self.IMPORTED_MODEL_TOOLTIP = "Imported model: {}"
|
||||
self.CONCATENATED_FILE_WARNING = "This is a concatenated file part. It will not work with llama-quantize; please concat the file first."
|
||||
self.CONCATENATED_FILES_FOUND = (
|
||||
"Found {} concatenated file parts. Please concat the files first."
|
||||
|
|
@ -248,6 +250,12 @@ def __init__(self):
|
|||
self.LLAMACPP_DOWNLOADED_AND_EXTRACTED = (
|
||||
"llama.cpp binary downloaded and extracted to {0}"
|
||||
)
|
||||
self.NO_SUITABLE_CUDA_BACKEND_FOUND = (
|
||||
"No suitable CUDA backend found for extraction"
|
||||
)
|
||||
self.LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED = (
|
||||
"llama.cpp binary downloaded and extracted to {0}"
|
||||
)
|
||||
self.REFRESHING_LLAMACPP_RELEASES = "Refreshing llama.cpp releases"
|
||||
self.UPDATING_ASSET_LIST = "Updating asset list"
|
||||
self.UPDATING_CUDA_OPTIONS = "Updating CUDA options"
|
||||
|
|
@ -365,7 +373,7 @@ def __init__(self):
|
|||
self.ADDING_LORA_ADAPTER = "Adding LoRA Adapter..."
|
||||
self.DELETING_LORA_ADAPTER = "Deleting LoRA Adapter..."
|
||||
self.SELECT_LORA_ADAPTER_FILE = "Select LoRA Adapter File"
|
||||
self.STARTING_LORA_EXPORT = "Starting LoRA export"
|
||||
self.STARTING_LORA_EXPORT = "Starting LoRA export..."
|
||||
self.SELECT_OUTPUT_TYPE = "Select Output Type (GGUF or GGML)"
|
||||
self.BASE_MODEL = "Base Model"
|
||||
self.SELECT_BASE_MODEL_FILE = "Select Base Model File (GGUF)"
|
||||
|
|
@ -407,7 +415,6 @@ def __init__(self):
|
|||
# Model actions
|
||||
self.CONFIRM_DELETE = "Confirm Delete"
|
||||
self.DELETE_MODEL_WARNING = "Are you sure you want to delete the model: {}?"
|
||||
self.ERROR_DELETING_MODEL = "Error deleting model: {}"
|
||||
self.MODEL_RENAMED_SUCCESSFULLY = "Model renamed successfully."
|
||||
self.MODEL_DELETED_SUCCESSFULLY = "Model deleted successfully."
|
||||
|
||||
|
|
@ -444,15 +451,8 @@ def __init__(self):
|
|||
self.HF_REPOSITORY_TYPE = "Repository Type"
|
||||
self.UPLOAD_TYPE = "Upload Type"
|
||||
self.UPLOAD = "Upload"
|
||||
self.INFO = "Info"
|
||||
|
||||
self.COPIED_COMMAND_TO_CLIPBOARD = "Copied command to clipboard:"
|
||||
|
||||
# Repository
|
||||
self.INVALID_REPOSITORY_FORMAT = (
|
||||
"Invalid repository format. Must be 'owner/repo'"
|
||||
)
|
||||
self.REPO_CANNOT_BE_EMPTY = "Owner and repository name cannot be empty"
|
||||
self.EXTRA_COMMAND_ARGUMENTS = "Additional command-line arguments"
|
||||
|
||||
|
||||
class _French(_Localization):
|
||||
|
|
@ -5797,339 +5797,187 @@ def __init__(self):
|
|||
class _Czech(_Localization):
|
||||
def __init__(self):
|
||||
super().__init__()
|
||||
|
||||
# General UI
|
||||
self.WINDOW_TITLE = (
|
||||
"AutoGGUF (automatizovaný nástroj pro kvantizaci GGUF modelů)"
|
||||
)
|
||||
self.WINDOW_TITLE = "AutoGGUF (Automatický kvantizátor modelů GGUF)"
|
||||
self.RAM_USAGE = "Využití RAM:"
|
||||
self.CPU_USAGE = "Využití CPU:"
|
||||
self.BACKEND = "Backend Llama.cpp:"
|
||||
self.REFRESH_BACKENDS = "Obnovit backendy"
|
||||
self.MODELS_PATH = "Cesta k modelům:"
|
||||
self.OUTPUT_PATH = "Cesta pro výstup:"
|
||||
self.OUTPUT_PATH = "Výstupní cesta:"
|
||||
self.LOGS_PATH = "Cesta k logům:"
|
||||
self.BROWSE = "Procházet"
|
||||
self.AVAILABLE_MODELS = "Dostupné modely:"
|
||||
self.REFRESH_MODELS = "Obnovit modely"
|
||||
self.STARTUP_ELASPED_TIME = "Inicializace trvala {0} ms"
|
||||
|
||||
# Usage Graphs
|
||||
self.CPU_USAGE_OVER_TIME = "Využití CPU v čase"
|
||||
self.RAM_USAGE_OVER_TIME = "Využití RAM v čase"
|
||||
|
||||
# Environment variables
|
||||
self.DOTENV_FILE_NOT_FOUND = "Soubor .env nenalezen."
|
||||
self.COULD_NOT_PARSE_LINE = "Nelze zpracovat řádek: {0}"
|
||||
self.ERROR_LOADING_DOTENV = "Chyba při načítání .env: {0}"
|
||||
|
||||
# Model Import
|
||||
self.IMPORT_MODEL = "Importovat model"
|
||||
self.SELECT_MODEL_TO_IMPORT = "Vyberte model k importu"
|
||||
self.CONFIRM_IMPORT = "Potvrdit import"
|
||||
self.IMPORT_MODEL_CONFIRMATION = "Chcete importovat model {}?"
|
||||
self.MODEL_IMPORTED_SUCCESSFULLY = "Model {} byl úspěšně importován"
|
||||
self.IMPORTING_MODEL = "Importuji model"
|
||||
self.IMPORTED_MODEL_TOOLTIP = "Importovaný model: {}"
|
||||
|
||||
# AutoFP8 Quantization
|
||||
self.AUTOFP8_QUANTIZATION_TASK_STARTED = (
|
||||
"Úloha kvantizace AutoFP8 byla spuštěna"
|
||||
)
|
||||
self.ERROR_STARTING_AUTOFP8_QUANTIZATION = (
|
||||
"Chyba při spuštění kvantizace AutoFP8"
|
||||
)
|
||||
self.QUANTIZING_WITH_AUTOFP8 = "Kvantizuji {0} pomocí AutoFP8"
|
||||
self.QUANTIZING_TO_WITH_AUTOFP8 = "Kvantizuji {0} na {1}"
|
||||
self.QUANTIZE_TO_FP8_DYNAMIC = "Kvantizovat na FP8 Dynamic"
|
||||
self.OPEN_MODEL_FOLDER = "Otevřít složku modelu"
|
||||
self.QUANTIZE = "Kvantizovat"
|
||||
self.OPEN_MODEL_FOLDER = "Otevřít složku modelu"
|
||||
self.INPUT_MODEL = "Vstupní model:"
|
||||
|
||||
# GGUF Verification
|
||||
self.INVALID_GGUF_FILE = "Neplatný GGUF soubor: {}"
|
||||
self.SHARDED_MODEL_NAME = "{} (Rozdělený)"
|
||||
self.IMPORTED_MODEL_TOOLTIP = "Importovaný model: {}"
|
||||
self.CONCATENATED_FILE_WARNING = "Toto je spojená část souboru. Nebude fungovat s llama-quantize; prosím, spojte soubor nejdříve."
|
||||
self.CONCATENATED_FILES_FOUND = (
|
||||
"Nalezeno {} spojených částí souboru. Prosím, spojte soubory nejdříve."
|
||||
)
|
||||
|
||||
# Plugins
|
||||
self.PLUGINS_DIR_NOT_EXIST = (
|
||||
"Adresář pluginů '{}' neexistuje. Žádné pluginy nebudou načteny."
|
||||
)
|
||||
self.PLUGINS_DIR_NOT_DIRECTORY = (
|
||||
"'{}' existuje, ale není to adresář. Žádné pluginy nebudou načteny."
|
||||
)
|
||||
self.PLUGIN_LOADED = "Načten plugin: {} {}"
|
||||
self.PLUGIN_INCOMPATIBLE = (
|
||||
"Plugin {} {} není kompatibilní s verzí AutoGGUF {}. Podporované verze: {}"
|
||||
)
|
||||
self.PLUGIN_LOAD_FAILED = "Nepodařilo se načíst plugin {}: {}"
|
||||
self.NO_PLUGINS_LOADED = "Žádné pluginy nebyly načteny."
|
||||
|
||||
# GPU Monitoring
|
||||
self.GPU_USAGE = "Využití GPU:"
|
||||
self.GPU_USAGE_FORMAT = "GPU: {:.1f}% | VRAM: {:.1f}% ({} MB / {} MB)"
|
||||
self.GPU_DETAILS = "Detaily GPU"
|
||||
self.GPU_USAGE_OVER_TIME = "Využití GPU v čase"
|
||||
self.VRAM_USAGE_OVER_TIME = "Využití VRAM v čase"
|
||||
self.PERCENTAGE = "Procento"
|
||||
self.TIME = "Čas (s)"
|
||||
self.NO_GPU_DETECTED = "Nebyla detekována žádná GPU"
|
||||
self.SELECT_GPU = "Vybrat GPU"
|
||||
self.AMD_GPU_NOT_SUPPORTED = "Byla detekována AMD GPU, ale není podporována"
|
||||
|
||||
# Quantization
|
||||
self.QUANTIZATION_TYPE = "Typ kvantizace:"
|
||||
self.ALLOW_REQUANTIZE = "Povolit rekvantizaci"
|
||||
self.LEAVE_OUTPUT_TENSOR = "Ponechat výstupní tensor"
|
||||
self.LEAVE_OUTPUT_TENSOR = "Ponechat výstupní tenzor"
|
||||
self.PURE = "Čistý"
|
||||
self.IMATRIX = "IMatrix:"
|
||||
self.INCLUDE_WEIGHTS = "Zahrnout váhy:"
|
||||
self.EXCLUDE_WEIGHTS = "Vyloučit váhy:"
|
||||
self.USE_OUTPUT_TENSOR_TYPE = "Použít typ výstupního tenzoru"
|
||||
self.USE_TOKEN_EMBEDDING_TYPE = "Použít typ pro token embeddings"
|
||||
self.USE_TOKEN_EMBEDDING_TYPE = "Použít typ vkládání tokenů"
|
||||
self.KEEP_SPLIT = "Zachovat rozdělení"
|
||||
self.KV_OVERRIDES = "Přepsání KV:"
|
||||
self.ADD_NEW_OVERRIDE = "Přidat nové přepsání"
|
||||
self.QUANTIZE_MODEL = "Kvantizovat model"
|
||||
self.EXTRA_ARGUMENTS = "Další argumenty:"
|
||||
self.EXTRA_ARGUMENTS_LABEL = "Další argumenty příkazové řádky"
|
||||
self.QUANTIZATION_COMMAND = "Příkaz pro kvantizaci"
|
||||
|
||||
# Presets
|
||||
self.SAVE_PRESET = "Uložit předvolbu"
|
||||
self.LOAD_PRESET = "Načíst předvolbu"
|
||||
|
||||
# Tasks
|
||||
self.TASKS = "Úlohy:"
|
||||
|
||||
# llama.cpp Download
|
||||
self.TASKS = "Úkoly:"
|
||||
self.DOWNLOAD_LLAMACPP = "Stáhnout llama.cpp"
|
||||
self.SELECT_RELEASE = "Vybrat verzi:"
|
||||
self.SELECT_ASSET = "Vybrat asset:"
|
||||
self.EXTRACT_CUDA_FILES = "Extrahovat CUDA soubory"
|
||||
self.SELECT_CUDA_BACKEND = "Vybrat CUDA backend:"
|
||||
self.SELECT_ASSET = "Vybrat aktivum:"
|
||||
self.EXTRACT_CUDA_FILES = "Extrahovat soubory CUDA"
|
||||
self.SELECT_CUDA_BACKEND = "Vybrat backend CUDA:"
|
||||
self.DOWNLOAD = "Stáhnout"
|
||||
self.REFRESH_RELEASES = "Obnovit verze"
|
||||
|
||||
# IMatrix Generation
|
||||
self.IMATRIX_GENERATION = "Generování IMatrix"
|
||||
self.DATA_FILE = "Datový soubor:"
|
||||
self.MODEL = "Model:"
|
||||
self.OUTPUT = "Výstup:"
|
||||
self.OUTPUT_FREQUENCY = "Frekvence výstupu:"
|
||||
self.GPU_OFFLOAD = "Využití GPU:"
|
||||
self.GPU_OFFLOAD = "Odlehčení GPU:"
|
||||
self.AUTO = "Automaticky"
|
||||
self.GENERATE_IMATRIX = "Generovat IMatrix"
|
||||
self.CONTEXT_SIZE = "Velikost kontextu:"
|
||||
self.CONTEXT_SIZE_FOR_IMATRIX = "Velikost kontextu pro generování IMatrix"
|
||||
self.THREADS = "Vlákna:"
|
||||
self.NUMBER_OF_THREADS_FOR_IMATRIX = "Počet vláken pro generování IMatrix"
|
||||
self.IMATRIX_GENERATION_COMMAND = "Příkaz pro generování IMatrix"
|
||||
|
||||
# LoRA Conversion
|
||||
self.LORA_CONVERSION = "Konverze LoRA"
|
||||
self.LORA_INPUT_PATH = "Vstupní cesta LoRA"
|
||||
self.LORA_OUTPUT_PATH = "Výstupní cesta LoRA"
|
||||
self.SELECT_LORA_INPUT_DIRECTORY = "Vybrat vstupní adresář LoRA"
|
||||
self.SELECT_LORA_OUTPUT_FILE = "Vybrat výstupní soubor LoRA"
|
||||
self.CONVERT_LORA = "Převést LoRA"
|
||||
self.LORA_CONVERSION_COMMAND = "Příkaz pro konverzi LoRA"
|
||||
|
||||
# LoRA Export
|
||||
self.EXPORT_LORA = "Exportovat LoRA"
|
||||
self.GGML_LORA_ADAPTERS = "GGML LoRA adaptéry"
|
||||
self.SELECT_LORA_ADAPTER_FILES = "Vybrat soubory LoRA adaptéru"
|
||||
self.ADD_ADAPTER = "Přidat adaptér"
|
||||
self.DELETE_ADAPTER = "Smazat"
|
||||
self.LORA_SCALE = "LoRA škála"
|
||||
self.ENTER_LORA_SCALE_VALUE = "Zadejte hodnotu LoRA škály (volitelné)"
|
||||
self.NUMBER_OF_THREADS_FOR_LORA_EXPORT = "Počet vláken pro export LoRA"
|
||||
self.LORA_EXPORT_COMMAND = "Příkaz pro export LoRA"
|
||||
|
||||
# HuggingFace to GGUF Conversion
|
||||
self.HF_TO_GGUF_CONVERSION = "Konverze HuggingFace na GGUF"
|
||||
self.MODEL_DIRECTORY = "Adresář modelu:"
|
||||
self.OUTPUT_FILE = "Výstupní soubor:"
|
||||
self.OUTPUT_TYPE = "Typ výstupu:"
|
||||
self.VOCAB_ONLY = "Pouze slovník"
|
||||
self.USE_TEMP_FILE = "Použít dočasný soubor"
|
||||
self.NO_LAZY_EVALUATION = "Bez líného vyhodnocování"
|
||||
self.MODEL_NAME = "Název modelu:"
|
||||
self.VERBOSE = "Podrobný výpis"
|
||||
self.SPLIT_MAX_SIZE = "Maximální velikost pro rozdělení:"
|
||||
self.DRY_RUN = "Zkušební běh"
|
||||
self.CONVERT_HF_TO_GGUF = "Převést HF na GGUF"
|
||||
self.SELECT_HF_MODEL_DIRECTORY = "Vybrat adresář modelu HuggingFace"
|
||||
self.BROWSE_FOR_HF_MODEL_DIRECTORY = "Procházení adresáře modelu HuggingFace"
|
||||
self.BROWSE_FOR_HF_TO_GGUF_OUTPUT = (
|
||||
"Procházení výstupního souboru pro konverzi HuggingFace na GGUF"
|
||||
)
|
||||
|
||||
# Update Checking
|
||||
self.UPDATE_AVAILABLE = "Aktualizace je k dispozici"
|
||||
self.NEW_VERSION_AVAILABLE = "Je k dispozici nová verze: {}"
|
||||
self.DOWNLOAD_NEW_VERSION = "Stáhnout?"
|
||||
self.ERROR_CHECKING_FOR_UPDATES = "Chyba při kontrole aktualizací:"
|
||||
self.CHECKING_FOR_UPDATES = "Kontrola aktualizací"
|
||||
|
||||
# General Messages
|
||||
self.ERROR = "Chyba"
|
||||
self.WARNING = "Varování"
|
||||
self.PROPERTIES = "Vlastnosti"
|
||||
self.CANCEL = "Zrušit"
|
||||
self.RESTART = "Restartovat"
|
||||
self.DELETE = "Smazat"
|
||||
self.RENAME = "Přejmenovat"
|
||||
self.CONFIRM_DELETION = "Opravdu chcete smazat tuto úlohu?"
|
||||
self.CONFIRM_DELETION = "Jste si jisti, že chcete smazat tento úkol?"
|
||||
self.TASK_RUNNING_WARNING = (
|
||||
"Některé úlohy stále běží. Opravdu chcete aplikaci ukončit?"
|
||||
"Některé úkoly stále běží. Jste si jisti, že chcete ukončit?"
|
||||
)
|
||||
self.YES = "Ano"
|
||||
self.NO = "Ne"
|
||||
self.COMPLETED = "Dokončeno"
|
||||
|
||||
# File Types
|
||||
self.ALL_FILES = "Všechny soubory (*)"
|
||||
self.GGUF_FILES = "GGUF soubory (*.gguf)"
|
||||
self.DAT_FILES = "DAT soubory (*.dat)"
|
||||
self.JSON_FILES = "JSON soubory (*.json)"
|
||||
self.BIN_FILES = "Binární soubory (*.bin)"
|
||||
self.LORA_FILES = "LoRA soubory (*.bin *.gguf)"
|
||||
self.GGUF_AND_BIN_FILES = "GGUF a binární soubory (*.gguf *.bin)"
|
||||
self.SHARDED = "rozdělený"
|
||||
|
||||
# Status Messages
|
||||
self.DOWNLOAD_COMPLETE = "Stahování dokončeno"
|
||||
self.CUDA_EXTRACTION_FAILED = "Extrakce CUDA selhala"
|
||||
self.CUDA_EXTRACTION_FAILED = "Extrahování CUDA se nezdařilo"
|
||||
self.PRESET_SAVED = "Předvolba uložena"
|
||||
self.PRESET_LOADED = "Předvolba načtena"
|
||||
self.NO_ASSET_SELECTED = "Nebyl vybrán žádný asset"
|
||||
self.DOWNLOAD_FAILED = "Stahování selhalo"
|
||||
self.NO_ASSET_SELECTED = "Nebylo vybráno žádné aktivum"
|
||||
self.DOWNLOAD_FAILED = "Stahování se nezdařilo"
|
||||
self.NO_BACKEND_SELECTED = "Nebyl vybrán žádný backend"
|
||||
self.NO_MODEL_SELECTED = "Nebyl vybrán žádný model"
|
||||
self.NO_SUITABLE_CUDA_BACKENDS = "Nebyly nalezeny žádné vhodné CUDA backendy"
|
||||
self.IN_PROGRESS = "Probíhá"
|
||||
self.LLAMACPP_DOWNLOADED_EXTRACTED = (
|
||||
"Binární soubor llama.cpp byl stažen a extrahován do {0}"
|
||||
)
|
||||
self.CUDA_FILES_EXTRACTED = "CUDA soubory byly extrahovány do"
|
||||
self.REFRESH_RELEASES = "Obnovit verze"
|
||||
self.NO_SUITABLE_CUDA_BACKENDS = "Nebyly nalezeny žádné vhodné backendy CUDA"
|
||||
self.LLAMACPP_DOWNLOADED_EXTRACTED = "Binární soubor llama.cpp byl stažen a extrahován do {0}\nSoubory CUDA extrahovány do {1}"
|
||||
self.CUDA_FILES_EXTRACTED = "Soubory CUDA extrahovány do"
|
||||
self.NO_SUITABLE_CUDA_BACKEND_EXTRACTION = (
|
||||
"Nebyl nalezen žádný vhodný CUDA backend pro extrakci"
|
||||
"Nebyl nalezen žádný vhodný backend CUDA pro extrakci"
|
||||
)
|
||||
self.ERROR_FETCHING_RELEASES = "Chyba při načítání verzí: {0}"
|
||||
self.CONFIRM_DELETION_TITLE = "Potvrdit smazání"
|
||||
self.LOG_FOR = "Log pro {0}"
|
||||
self.FAILED_TO_LOAD_PRESET = "Nepodařilo se načíst předvolbu: {0}"
|
||||
self.ALL_FILES = "Všechny soubory (*)"
|
||||
self.GGUF_FILES = "Soubory GGUF (*.gguf)"
|
||||
self.DAT_FILES = "Soubory DAT (*.dat)"
|
||||
self.JSON_FILES = "Soubory JSON (*.json)"
|
||||
self.FAILED_LOAD_PRESET = "Nepodařilo se načíst předvolbu: {0}"
|
||||
self.INITIALIZING_AUTOGGUF = "Inicializace aplikace AutoGGUF"
|
||||
self.AUTOGGUF_INITIALIZATION_COMPLETE = "Inicializace AutoGGUF dokončena"
|
||||
self.REFRESHING_BACKENDS = "Obnovuji backendy"
|
||||
self.NO_BACKENDS_AVAILABLE = "Nejsou dostupné žádné backendy"
|
||||
self.REFRESHING_BACKENDS = "Obnovování backendů"
|
||||
self.NO_BACKENDS_AVAILABLE = "Žádné dostupné backendy"
|
||||
self.FOUND_VALID_BACKENDS = "Nalezeno {0} platných backendů"
|
||||
self.SAVING_PRESET = "Ukládám předvolbu"
|
||||
self.SAVING_PRESET = "Ukládání předvolby"
|
||||
self.PRESET_SAVED_TO = "Předvolba uložena do {0}"
|
||||
self.LOADING_PRESET = "Načítám předvolbu"
|
||||
self.LOADING_PRESET = "Načítání předvolby"
|
||||
self.PRESET_LOADED_FROM = "Předvolba načtena z {0}"
|
||||
self.ADDING_KV_OVERRIDE = "Přidávám přepsání KV: {0}"
|
||||
self.SAVING_TASK_PRESET = "Ukládám předvolbu úlohy pro {0}"
|
||||
self.TASK_PRESET_SAVED = "Předvolba úlohy uložena"
|
||||
self.TASK_PRESET_SAVED_TO = "Předvolba úlohy uložena do {0}"
|
||||
self.RESTARTING_TASK = "Restartuji úlohu: {0}"
|
||||
self.ADDING_KV_OVERRIDE = "Přidávání přepsání KV: {0}"
|
||||
self.SAVING_TASK_PRESET = "Ukládání předvolby úkolu pro {0}"
|
||||
self.TASK_PRESET_SAVED = "Předvolba úkolu uložena"
|
||||
self.TASK_PRESET_SAVED_TO = "Předvolba úkolu uložena do {0}"
|
||||
self.RESTARTING_TASK = "Restartování úkolu: {0}"
|
||||
self.IN_PROGRESS = "Probíhá"
|
||||
self.DOWNLOAD_FINISHED_EXTRACTED_TO = "Stahování dokončeno. Extrahováno do: {0}"
|
||||
self.LLAMACPP_DOWNLOADED_AND_EXTRACTED = (
|
||||
"Binární soubor llama.cpp byl stažen a extrahován do {0}"
|
||||
)
|
||||
self.LLAMACPP_DOWNLOADED_AND_EXTRACTED = "Binární soubor llama.cpp byl stažen a extrahován do {0}\nSoubory CUDA extrahovány do {1}"
|
||||
self.NO_SUITABLE_CUDA_BACKEND_FOUND = (
|
||||
"Nebyl nalezen žádný vhodný CUDA backend pro extrakci"
|
||||
"Nebyl nalezen žádný vhodný backend CUDA pro extrakci"
|
||||
)
|
||||
self.LLAMACPP_BINARY_DOWNLOADED_AND_EXTRACTED = (
|
||||
"Binární soubor llama.cpp byl stažen a extrahován do {0}"
|
||||
)
|
||||
self.REFRESHING_LLAMACPP_RELEASES = "Obnovuji verze llama.cpp"
|
||||
self.UPDATING_ASSET_LIST = "Aktualizuji seznam assetů"
|
||||
self.UPDATING_CUDA_OPTIONS = "Aktualizuji možnosti CUDA"
|
||||
self.STARTING_LLAMACPP_DOWNLOAD = "Spouštím stahování llama.cpp"
|
||||
self.UPDATING_CUDA_BACKENDS = "Aktualizuji CUDA backendy"
|
||||
self.NO_CUDA_BACKEND_SELECTED = "Nebyl vybrán žádný CUDA backend pro extrakci"
|
||||
self.EXTRACTING_CUDA_FILES = "Extrahuji CUDA soubory z {0} do {1}"
|
||||
self.REFRESHING_LLAMACPP_RELEASES = "Obnovování verzí llama.cpp"
|
||||
self.UPDATING_ASSET_LIST = "Aktualizace seznamu aktiv"
|
||||
self.UPDATING_CUDA_OPTIONS = "Aktualizace možností CUDA"
|
||||
self.STARTING_LLAMACPP_DOWNLOAD = "Zahájení stahování llama.cpp"
|
||||
self.UPDATING_CUDA_BACKENDS = "Aktualizace backendů CUDA"
|
||||
self.NO_CUDA_BACKEND_SELECTED = "Nebyl vybrán žádný backend CUDA pro extrakci"
|
||||
self.EXTRACTING_CUDA_FILES = "Extrahování souborů CUDA z {0} do {1}"
|
||||
self.DOWNLOAD_ERROR = "Chyba stahování: {0}"
|
||||
self.SHOWING_TASK_CONTEXT_MENU = "Zobrazuji kontextové menu úlohy"
|
||||
self.SHOWING_PROPERTIES_FOR_TASK = "Zobrazuji vlastnosti pro úlohu: {0}"
|
||||
self.CANCELLING_TASK = "Ruším úlohu: {0}"
|
||||
self.SHOWING_TASK_CONTEXT_MENU = "Zobrazení kontextové nabídky úkolu"
|
||||
self.SHOWING_PROPERTIES_FOR_TASK = "Zobrazení vlastností úkolu: {0}"
|
||||
self.CANCELLING_TASK = "Zrušení úkolu: {0}"
|
||||
self.CANCELED = "Zrušeno"
|
||||
self.DELETING_TASK = "Mažu úlohu: {0}"
|
||||
self.LOADING_MODELS = "Načítám modely"
|
||||
self.DELETING_TASK = "Mazání úkolu: {0}"
|
||||
self.LOADING_MODELS = "Načítání modelů"
|
||||
self.LOADED_MODELS = "Načteno {0} modelů"
|
||||
self.BROWSING_FOR_MODELS_DIRECTORY = "Procházení adresáře modelů"
|
||||
self.SELECT_MODELS_DIRECTORY = "Vybrat adresář modelů"
|
||||
self.SELECT_MODELS_DIRECTORY = "Vyberte adresář modelů"
|
||||
self.BROWSING_FOR_OUTPUT_DIRECTORY = "Procházení výstupního adresáře"
|
||||
self.SELECT_OUTPUT_DIRECTORY = "Vybrat výstupní adresář"
|
||||
self.SELECT_OUTPUT_DIRECTORY = "Vyberte výstupní adresář"
|
||||
self.BROWSING_FOR_LOGS_DIRECTORY = "Procházení adresáře logů"
|
||||
self.SELECT_LOGS_DIRECTORY = "Vybrat adresář logů"
|
||||
self.SELECT_LOGS_DIRECTORY = "Vyberte adresář logů"
|
||||
self.BROWSING_FOR_IMATRIX_FILE = "Procházení souboru IMatrix"
|
||||
self.SELECT_IMATRIX_FILE = "Vybrat soubor IMatrix"
|
||||
self.SELECT_IMATRIX_FILE = "Vyberte soubor IMatrix"
|
||||
self.RAM_USAGE_FORMAT = "{0:.1f}% ({1} MB / {2} MB)"
|
||||
self.CPU_USAGE_FORMAT = "Využití CPU: {0:.1f}%"
|
||||
self.VALIDATING_QUANTIZATION_INPUTS = "Validuji vstupy pro kvantizaci"
|
||||
self.MODELS_PATH_REQUIRED = "Je vyžadována cesta k modelům"
|
||||
self.OUTPUT_PATH_REQUIRED = "Je vyžadována cesta pro výstup"
|
||||
self.LOGS_PATH_REQUIRED = "Je vyžadována cesta k logům"
|
||||
self.STARTING_MODEL_QUANTIZATION = "Spouštím kvantizaci modelu"
|
||||
self.VALIDATING_QUANTIZATION_INPUTS = "Ověřování vstupů kvantizace"
|
||||
self.MODELS_PATH_REQUIRED = "Cesta k modelům je vyžadována"
|
||||
self.OUTPUT_PATH_REQUIRED = "Výstupní cesta je vyžadována"
|
||||
self.LOGS_PATH_REQUIRED = "Cesta k logům je vyžadována"
|
||||
self.STARTING_MODEL_QUANTIZATION = "Spuštění kvantizace modelu"
|
||||
self.INPUT_FILE_NOT_EXIST = "Vstupní soubor '{0}' neexistuje."
|
||||
self.QUANTIZING_MODEL_TO = "Kvantizuji {0} na {1}"
|
||||
self.QUANTIZATION_TASK_STARTED = "Úloha kvantizace spuštěna pro {0}"
|
||||
self.QUANTIZING_MODEL_TO = "Kvantizace {0} na {1}"
|
||||
self.QUANTIZATION_TASK_STARTED = "Úkol kvantizace spuštěn pro {0}"
|
||||
self.ERROR_STARTING_QUANTIZATION = "Chyba při spuštění kvantizace: {0}"
|
||||
self.UPDATING_MODEL_INFO = "Aktualizuji informace o modelu: {0}"
|
||||
self.TASK_FINISHED = "Úloha dokončena: {0}"
|
||||
self.SHOWING_TASK_DETAILS_FOR = "Zobrazuji detaily úlohy pro: {0}"
|
||||
self.UPDATING_MODEL_INFO = "Aktualizace informací o modelu: {0}"
|
||||
self.TASK_FINISHED = "Úkol dokončen: {0}"
|
||||
self.SHOWING_TASK_DETAILS_FOR = "Zobrazení detailů úkolu pro: {0}"
|
||||
self.BROWSING_FOR_IMATRIX_DATA_FILE = "Procházení datového souboru IMatrix"
|
||||
self.SELECT_DATA_FILE = "Vybrat datový soubor"
|
||||
self.SELECT_DATA_FILE = "Vyberte datový soubor"
|
||||
self.BROWSING_FOR_IMATRIX_MODEL_FILE = "Procházení souboru modelu IMatrix"
|
||||
self.SELECT_MODEL_FILE = "Vybrat soubor modelu"
|
||||
self.SELECT_MODEL_FILE = "Vyberte soubor modelu"
|
||||
self.BROWSING_FOR_IMATRIX_OUTPUT_FILE = "Procházení výstupního souboru IMatrix"
|
||||
self.SELECT_OUTPUT_FILE = "Vybrat výstupní soubor"
|
||||
self.STARTING_IMATRIX_GENERATION = "Spouštím generování IMatrix"
|
||||
self.BACKEND_PATH_NOT_EXIST = "Cesta k backendu neexistuje: {0}"
|
||||
self.GENERATING_IMATRIX = "Generuji IMatrix"
|
||||
self.SELECT_OUTPUT_FILE = "Vyberte výstupní soubor"
|
||||
self.STARTING_IMATRIX_GENERATION = "Spuštění generování IMatrix"
|
||||
self.BACKEND_PATH_NOT_EXIST = "Cesta backendu neexistuje: {0}"
|
||||
self.GENERATING_IMATRIX = "Generování IMatrix"
|
||||
self.ERROR_STARTING_IMATRIX_GENERATION = (
|
||||
"Chyba při spuštění generování IMatrix: {0}"
|
||||
)
|
||||
self.IMATRIX_GENERATION_TASK_STARTED = "Úloha generování IMatrix spuštěna"
|
||||
self.IMATRIX_GENERATION_TASK_STARTED = "Úkol generování IMatrix spuštěn"
|
||||
self.ERROR_MESSAGE = "Chyba: {0}"
|
||||
self.TASK_ERROR = "Chyba úlohy: {0}"
|
||||
self.APPLICATION_CLOSING = "Ukončuji aplikaci"
|
||||
self.APPLICATION_CLOSED = "Aplikace ukončena"
|
||||
self.TASK_ERROR = "Chyba úkolu: {0}"
|
||||
self.APPLICATION_CLOSING = "Zavírání aplikace"
|
||||
self.APPLICATION_CLOSED = "Aplikace zavřena"
|
||||
self.SELECT_QUANTIZATION_TYPE = "Vyberte typ kvantizace"
|
||||
self.ALLOWS_REQUANTIZING = (
|
||||
"Povoluje rekvantizaci tenzorů, které již byly kvantizovány"
|
||||
"Umožňuje rekvantizovat tenzory, které již byly kvantizovány"
|
||||
)
|
||||
self.LEAVE_OUTPUT_WEIGHT = (
|
||||
"Ponechá output.weight nekvantizovaný (nebo rekvantizovaný)"
|
||||
)
|
||||
self.LEAVE_OUTPUT_WEIGHT = "Ponechá output.weight ne(re)kvantizovaný"
|
||||
self.DISABLE_K_QUANT_MIXTURES = (
|
||||
"Zakáže k-kvantové směsi a kvantizuje všechny tenzory na stejný typ"
|
||||
"Zakázat k-kvantové směsi a kvantizovat všechny tenzory na stejný typ"
|
||||
)
|
||||
self.USE_DATA_AS_IMPORTANCE_MATRIX = (
|
||||
"Použije data v souboru jako matici důležitosti pro optimalizace kvantizace"
|
||||
"Použít data v souboru jako matici důležitosti pro optimalizace kvantizace"
|
||||
)
|
||||
self.USE_IMPORTANCE_MATRIX_FOR_TENSORS = (
|
||||
"Použije matici důležitosti pro tyto tenzory"
|
||||
"Použít matici důležitosti pro tyto tenzory"
|
||||
)
|
||||
self.DONT_USE_IMPORTANCE_MATRIX_FOR_TENSORS = (
|
||||
"Nepoužije matici důležitosti pro tyto tenzory"
|
||||
"Nepoužívat matici důležitosti pro tyto tenzory"
|
||||
)
|
||||
self.OUTPUT_TENSOR_TYPE = "Typ výstupního tenzoru:"
|
||||
self.USE_THIS_TYPE_FOR_OUTPUT_WEIGHT = (
|
||||
"Použije tento typ pro tenzor output.weight"
|
||||
"Použít tento typ pro tenzor output.weight"
|
||||
)
|
||||
self.TOKEN_EMBEDDING_TYPE = "Typ token embeddings:"
|
||||
self.TOKEN_EMBEDDING_TYPE = "Typ vkládání tokenů:"
|
||||
self.USE_THIS_TYPE_FOR_TOKEN_EMBEDDINGS = (
|
||||
"Použije tento typ pro tenzor token embeddings"
|
||||
"Použít tento typ pro tenzor vkládání tokenů"
|
||||
)
|
||||
self.WILL_GENERATE_QUANTIZED_MODEL_IN_SAME_SHARDS = (
|
||||
"Vygeneruje kvantizovaný model ve stejných sharded souborech jako vstup"
|
||||
"Vygeneruje kvantizovaný model ve stejných fragmentech jako vstup"
|
||||
)
|
||||
self.OVERRIDE_MODEL_METADATA = "Přepsat metadata modelu"
|
||||
self.INPUT_DATA_FILE_FOR_IMATRIX = (
|
||||
|
|
@ -6138,123 +5986,9 @@ def __init__(self):
|
|||
self.MODEL_TO_BE_QUANTIZED = "Model, který má být kvantizován"
|
||||
self.OUTPUT_PATH_FOR_GENERATED_IMATRIX = "Výstupní cesta pro generovaný IMatrix"
|
||||
self.HOW_OFTEN_TO_SAVE_IMATRIX = "Jak často ukládat IMatrix"
|
||||
self.SET_GPU_OFFLOAD_VALUE = "Nastavit hodnotu pro využití GPU (-ngl)"
|
||||
self.STARTING_LORA_CONVERSION = "Spouštím konverzi LoRA"
|
||||
self.LORA_INPUT_PATH_REQUIRED = "Je vyžadována vstupní cesta LoRA."
|
||||
self.LORA_OUTPUT_PATH_REQUIRED = "Je vyžadována výstupní cesta LoRA."
|
||||
self.ERROR_STARTING_LORA_CONVERSION = "Chyba při spuštění konverze LoRA: {}"
|
||||
self.LORA_CONVERSION_TASK_STARTED = "Úloha konverze LoRA spuštěna."
|
||||
self.BROWSING_FOR_LORA_INPUT_DIRECTORY = "Procházení vstupního adresáře LoRA..."
|
||||
self.BROWSING_FOR_LORA_OUTPUT_FILE = "Procházení výstupního souboru LoRA..."
|
||||
self.CONVERTING_LORA = "Konverze LoRA"
|
||||
self.LORA_CONVERSION_FINISHED = "Konverze LoRA dokončena."
|
||||
self.LORA_FILE_MOVED = "Soubor LoRA přesunut z {} do {}."
|
||||
self.LORA_FILE_NOT_FOUND = "Soubor LoRA nenalezen: {}."
|
||||
self.ERROR_MOVING_LORA_FILE = "Chyba při přesouvání souboru LoRA: {}"
|
||||
self.MODEL_PATH_REQUIRED = "Je vyžadována cesta k modelu."
|
||||
self.AT_LEAST_ONE_LORA_ADAPTER_REQUIRED = (
|
||||
"Je vyžadován alespoň jeden LoRA adaptér."
|
||||
)
|
||||
self.INVALID_LORA_SCALE_VALUE = "Neplatná hodnota LoRA škály."
|
||||
self.ERROR_STARTING_LORA_EXPORT = "Chyba při spuštění exportu LoRA: {}"
|
||||
self.LORA_EXPORT_TASK_STARTED = "Úloha exportu LoRA spuštěna."
|
||||
self.EXPORTING_LORA = "Exportuji LoRA..."
|
||||
self.BROWSING_FOR_EXPORT_LORA_MODEL_FILE = (
|
||||
"Procházení souboru modelu pro export LoRA..."
|
||||
)
|
||||
self.BROWSING_FOR_EXPORT_LORA_OUTPUT_FILE = (
|
||||
"Procházení výstupního souboru pro export LoRA..."
|
||||
)
|
||||
self.ADDING_LORA_ADAPTER = "Přidávám LoRA adaptér..."
|
||||
self.DELETING_LORA_ADAPTER = "Mažu LoRA adaptér..."
|
||||
self.SELECT_LORA_ADAPTER_FILE = "Vybrat soubor LoRA adaptéru"
|
||||
self.STARTING_LORA_EXPORT = "Spouštím export LoRA..."
|
||||
self.SELECT_OUTPUT_TYPE = "Vyberte typ výstupu (GGUF nebo GGML)"
|
||||
self.BASE_MODEL = "Základní model"
|
||||
self.SELECT_BASE_MODEL_FILE = "Vybrat soubor základního modelu (GGUF)"
|
||||
self.BASE_MODEL_PATH_REQUIRED = (
|
||||
"Pro výstup GGUF je vyžadována cesta k základnímu modelu."
|
||||
)
|
||||
self.BROWSING_FOR_BASE_MODEL_FILE = "Procházení souboru základního modelu..."
|
||||
self.SELECT_BASE_MODEL_FOLDER = (
|
||||
"Vybrat složku základního modelu (obsahující safetensors)"
|
||||
)
|
||||
self.BROWSING_FOR_BASE_MODEL_FOLDER = "Procházení složky základního modelu..."
|
||||
self.LORA_CONVERSION_FROM_TO = "Konverze LoRA z {} na {}"
|
||||
self.GENERATING_IMATRIX_FOR = "Generuji IMatrix pro {}"
|
||||
self.MODEL_PATH_REQUIRED_FOR_IMATRIX = (
|
||||
"Pro generování IMatrix je vyžadována cesta k modelu."
|
||||
)
|
||||
self.NO_ASSET_SELECTED_FOR_CUDA_CHECK = (
|
||||
"Nebyl vybrán žádný asset pro kontrolu CUDA"
|
||||
)
|
||||
self.NO_QUANTIZATION_TYPE_SELECTED = "Nebyl vybrán žádný typ kvantizace. Prosím, vyberte alespoň jeden typ kvantizace."
|
||||
self.STARTING_HF_TO_GGUF_CONVERSION = "Spouštím konverzi HuggingFace na GGUF"
|
||||
self.MODEL_DIRECTORY_REQUIRED = "Je vyžadován adresář modelu"
|
||||
self.HF_TO_GGUF_CONVERSION_COMMAND = "Příkaz pro konverzi HF na GGUF: {}"
|
||||
self.CONVERTING_TO_GGUF = "Převádím {} na GGUF"
|
||||
self.ERROR_STARTING_HF_TO_GGUF_CONVERSION = (
|
||||
"Chyba při spuštění konverze HuggingFace na GGUF: {}"
|
||||
)
|
||||
self.HF_TO_GGUF_CONVERSION_TASK_STARTED = (
|
||||
"Úloha konverze HuggingFace na GGUF spuštěna"
|
||||
)
|
||||
|
||||
# Split GGUF
|
||||
self.SPLIT_GGUF = "Rozdělit GGUF"
|
||||
self.SPLIT_MAX_SIZE = "Maximální velikost pro rozdělení"
|
||||
self.SPLIT_MAX_TENSORS = "Maximální počet tensorů pro rozdělení"
|
||||
self.SPLIT_GGUF_TASK_STARTED = "Úloha rozdělení GGUF spuštěna"
|
||||
self.SPLIT_GGUF_TASK_FINISHED = "Úloha rozdělení GGUF dokončena"
|
||||
self.SPLIT_GGUF_COMMAND = "Příkaz pro rozdělení GGUF"
|
||||
self.SPLIT_GGUF_ERROR = "Chyba při spuštění rozdělení GGUF"
|
||||
self.NUMBER_OF_TENSORS = "Počet tensorů"
|
||||
self.SIZE_IN_UNITS = "Velikost v G/M"
|
||||
|
||||
# Model actions
|
||||
self.CONFIRM_DELETE = "Potvrdit smazání"
|
||||
self.DELETE_MODEL_WARNING = "Opravdu chcete smazat model: {}?"
|
||||
self.MODEL_RENAMED_SUCCESSFULLY = "Model byl úspěšně přejmenován."
|
||||
self.MODEL_DELETED_SUCCESSFULLY = "Model byl úspěšně smazán."
|
||||
|
||||
# HuggingFace Transfer
|
||||
self.ALL_FIELDS_REQUIRED = "Všechna pole jsou vyžadována."
|
||||
self.HUGGINGFACE_UPLOAD_COMMAND = "Příkaz pro nahrávání na HuggingFace: "
|
||||
self.UPLOADING = "Nahrávám"
|
||||
self.UPLOADING_FOLDER = "Nahrávám složku"
|
||||
self.HF_TRANSFER_TASK_NAME = "{} {} do {} z {}"
|
||||
self.ERROR_STARTING_HF_TRANSFER = "Chyba při spuštění přenosu na HF: {}"
|
||||
self.STARTED_HUGGINGFACE_TRANSFER = "Spuštěna operace {} na HuggingFace."
|
||||
self.SELECT_FOLDER = "Vybrat složku"
|
||||
self.SELECT_FILE = "Vybrat soubor"
|
||||
|
||||
# Menubar
|
||||
self.CLOSE = "Zavřít"
|
||||
self.FILE = "Soubor"
|
||||
self.FOLDER = "Složka"
|
||||
self.HELP = "Nápověda"
|
||||
self.ABOUT = "O aplikaci"
|
||||
|
||||
self.AUTOFP8 = "AutoFP8"
|
||||
self.TOOLS = "Nástroje"
|
||||
self.HF_TRANSFER = "Přenos HF"
|
||||
self.MERGE_GGUF = "Sloučit GGUF"
|
||||
|
||||
self.HF_UPLOAD = "Nahrát na HF"
|
||||
self.HF_REPOSITORY = "Repozitář:"
|
||||
self.HF_REMOTE_PATH = "Vzdálená cesta:"
|
||||
self.HF_LOCAL_PATH = "Lokální cesta:"
|
||||
self.MODEL = "Model"
|
||||
self.DATASET = "Dataset"
|
||||
self.SPACE = "Space"
|
||||
self.HF_REPOSITORY_TYPE = "Typ repozitáře"
|
||||
self.UPLOAD_TYPE = "Typ nahrávání"
|
||||
self.UPLOAD = "Nahrát"
|
||||
|
||||
self.EXTRA_COMMAND_ARGUMENTS = "Další argumenty příkazové řádky"
|
||||
|
||||
self.INFO = "Info"
|
||||
self.COPIED_COMMAND_TO_CLIPBOARD = "Zkopírován příkaz do schránky:"
|
||||
self.SET_GPU_OFFLOAD_VALUE = "Nastavit hodnotu odlehčení GPU (-ngl)"
|
||||
self.COMPLETED = "Dokončeno"
|
||||
self.REFRESH_MODELS = "Reîmprospătează modelele"
|
||||
|
||||
|
||||
class _CanadianFrench(_Localization):
|
||||
|
|
|
|||
|
|
@ -24,21 +24,8 @@ def __init__(self, model_info, parent=None) -> None:
|
|||
def format_model_info(self, model_info) -> str:
|
||||
html = "<h2>Model Information</h2>"
|
||||
html += f"<p><b>Architecture:</b> {model_info.get('architecture', 'N/A')}</p>"
|
||||
|
||||
# Format quantization types
|
||||
quant_types = model_info.get("quantization_type", [])
|
||||
if quant_types:
|
||||
# Clean up the format: remove "- type " prefix and join with " | "
|
||||
formatted_types = []
|
||||
for qtype in quant_types:
|
||||
# Remove "- type " prefix if present
|
||||
clean_type = qtype.replace("- type ", "").strip()
|
||||
formatted_types.append(clean_type)
|
||||
quant_display = " | ".join(formatted_types)
|
||||
else:
|
||||
quant_display = "N/A"
|
||||
|
||||
html += f"<p><b>Quantization Type:</b> {quant_display}</p>"
|
||||
html += f"<p><b>Quantization Type:</b> {model_info.get('quantization_type', 'N/A')}</p>"
|
||||
html += f"<p><b>KV Pairs:</b> {model_info.get('kv_pairs', 'N/A')}</p>"
|
||||
html += f"<p><b>Tensors:</b> {model_info.get('tensors', 'N/A')}</p>"
|
||||
|
||||
html += "<h3>Key-Value Pairs:</h3>"
|
||||
|
|
|
|||
|
|
@ -59,34 +59,6 @@ def run(self) -> None:
|
|||
self.error_signal.emit(str(e))
|
||||
|
||||
def parse_model_info(self, line) -> None:
|
||||
# Mapping of technical keys to human-readable names
|
||||
key_mappings = {
|
||||
"general.architecture": "Architecture",
|
||||
"general.name": "Model Name",
|
||||
"general.file_type": "File Type",
|
||||
"general.quantization_version": "Quantization Version",
|
||||
"llama.block_count": "Layers",
|
||||
"llama.context_length": "Context Length",
|
||||
"llama.embedding_length": "Embedding Size",
|
||||
"llama.feed_forward_length": "Feed Forward Length",
|
||||
"llama.attention.head_count": "Attention Heads",
|
||||
"llama.attention.head_count_kv": "Key-Value Heads",
|
||||
"llama.attention.layer_norm_rms_epsilon": "RMS Norm Epsilon",
|
||||
"llama.rope.freq_base": "RoPE Frequency Base",
|
||||
"llama.rope.dimension_count": "RoPE Dimensions",
|
||||
"llama.vocab_size": "Vocabulary Size",
|
||||
"tokenizer.ggml.model": "Tokenizer Model",
|
||||
"tokenizer.ggml.pre": "Tokenizer Preprocessing",
|
||||
"tokenizer.ggml.tokens": "Tokens",
|
||||
"tokenizer.ggml.token_type": "Token Types",
|
||||
"tokenizer.ggml.merges": "BPE Merges",
|
||||
"tokenizer.ggml.bos_token_id": "Begin of Sequence Token ID",
|
||||
"tokenizer.ggml.eos_token_id": "End of Sequence Token ID",
|
||||
"tokenizer.chat_template": "Chat Template",
|
||||
"tokenizer.ggml.padding_token_id": "Padding Token ID",
|
||||
"tokenizer.ggml.unk_token_id": "Unknown Token ID",
|
||||
}
|
||||
|
||||
# Parse output for model information
|
||||
if "llama_model_loader: loaded meta data with" in line:
|
||||
parts = line.split()
|
||||
|
|
@ -94,25 +66,10 @@ def parse_model_info(self, line) -> None:
|
|||
self.model_info["tensors"] = parts[9]
|
||||
elif "general.architecture" in line:
|
||||
self.model_info["architecture"] = line.split("=")[-1].strip()
|
||||
elif line.startswith("llama_model_loader: - kv") and "=" in line:
|
||||
# Split on '=' and take the parts
|
||||
parts = line.split("=", 1) # Split only on first '='
|
||||
left_part = parts[0].strip()
|
||||
value = parts[1].strip()
|
||||
|
||||
# Extract key and type from left part
|
||||
# Format: "llama_model_loader: - kv N: key type"
|
||||
kv_parts = left_part.split(":")
|
||||
if len(kv_parts) >= 3:
|
||||
key_type_part = kv_parts[2].strip() # This is "key type"
|
||||
key = key_type_part.rsplit(" ", 1)[
|
||||
0
|
||||
] # Everything except last word (type)
|
||||
|
||||
# Use human-readable name if available, otherwise use original key
|
||||
display_key = key_mappings.get(key, key)
|
||||
|
||||
self.model_info.setdefault("kv_data", {})[display_key] = value
|
||||
elif line.startswith("llama_model_loader: - kv"):
|
||||
key = line.split(":")[2].strip()
|
||||
value = line.split("=")[-1].strip()
|
||||
self.model_info.setdefault("kv_data", {})[key] = value
|
||||
elif line.startswith("llama_model_loader: - type"):
|
||||
parts = line.split(":")
|
||||
if len(parts) > 1:
|
||||
|
|
|
|||
|
|
@ -95,41 +95,29 @@ def show_task_context_menu(self, position) -> None:
|
|||
|
||||
def show_task_properties(self, item) -> None:
|
||||
self.logger.debug(SHOWING_PROPERTIES_FOR_TASK.format(item.text()))
|
||||
task_item = self.task_list.itemWidget(item)
|
||||
for thread in self.quant_threads:
|
||||
if thread.log_file == task_item.log_file:
|
||||
model_info_dialog = ModelInfoDialog(thread.model_info, self)
|
||||
|
||||
model_info_dialog.exec()
|
||||
break
|
||||
|
||||
def cancel_task(self, item) -> None:
|
||||
# TODO: fix possibly buggy signal behavior
|
||||
self.logger.info(CANCELLING_TASK.format(item.text()))
|
||||
task_item = self.task_list.itemWidget(item)
|
||||
if task_item:
|
||||
task_name = task_item.task_name # Store the name before any changes
|
||||
self.logger.info(CANCELLING_TASK.format(task_name))
|
||||
|
||||
# Find the thread and disconnect signals before terminating
|
||||
for thread in self.quant_threads:
|
||||
if thread.log_file == task_item.log_file:
|
||||
# Disconnect all signals from this thread first
|
||||
try:
|
||||
thread.error_signal.disconnect() # Disconnect all error signal connections
|
||||
thread.output_signal.disconnect() # Disconnect all output signal connections
|
||||
except TypeError:
|
||||
# No connections to disconnect
|
||||
pass
|
||||
|
||||
# Now terminate the thread
|
||||
thread.terminate()
|
||||
task_item.update_status(CANCELED)
|
||||
self.quant_threads.remove(thread)
|
||||
break
|
||||
|
||||
def delete_task(self, item) -> None:
|
||||
task_item = self.task_list.itemWidget(item)
|
||||
if not task_item:
|
||||
return
|
||||
self.logger.info(DELETING_TASK.format(item.text()))
|
||||
|
||||
task_name = task_item.task_name # Store task_name before deletion
|
||||
self.logger.info(DELETING_TASK.format(task_name))
|
||||
# Cancel the task first
|
||||
self.cancel_task(item)
|
||||
|
||||
reply = QMessageBox.question(
|
||||
self,
|
||||
|
|
@ -138,16 +126,12 @@ def delete_task(self, item) -> None:
|
|||
QMessageBox.StandardButton.Yes | QMessageBox.StandardButton.No,
|
||||
QMessageBox.StandardButton.No,
|
||||
)
|
||||
|
||||
if reply == QMessageBox.StandardButton.Yes:
|
||||
# Cancel the task first (which disconnects signals)
|
||||
self.cancel_task(item)
|
||||
|
||||
# Now remove from list and delete
|
||||
task_item = self.task_list.itemWidget(item)
|
||||
row = self.task_list.row(item)
|
||||
self.task_list.takeItem(row)
|
||||
|
||||
# Delete the widget after removing from list
|
||||
if task_item:
|
||||
task_item.deleteLater()
|
||||
|
||||
def update_status(self, status) -> None:
|
||||
|
|
|
|||
File diff suppressed because it is too large
Load Diff
|
|
@ -18,16 +18,15 @@
|
|||
SupportsIndex,
|
||||
cast,
|
||||
)
|
||||
from transformers import AutoConfig
|
||||
|
||||
import torch
|
||||
|
||||
if TYPE_CHECKING:
|
||||
from torch import Tensor
|
||||
import gguf
|
||||
|
||||
# reuse model definitions from convert_hf_to_gguf.py
|
||||
from convert_hf_to_gguf import LazyTorchTensor, ModelBase
|
||||
from gguf.constants import *
|
||||
|
||||
from convert_hf_to_gguf import LazyTorchTensor, Model
|
||||
|
||||
logger = logging.getLogger("lora-to-gguf")
|
||||
|
||||
|
|
@ -38,10 +37,9 @@ class PartialLoraTensor:
|
|||
B: Tensor | None = None
|
||||
|
||||
|
||||
# magic to support tensor shape modifications and splitting
|
||||
class LoraTorchTensor:
|
||||
_lora_A: Tensor # (n_rank, row_size)
|
||||
_lora_B: Tensor # (col_size, n_rank)
|
||||
_lora_A: Tensor
|
||||
_lora_B: Tensor
|
||||
_rank: int
|
||||
|
||||
def __init__(self, A: Tensor, B: Tensor):
|
||||
|
|
@ -59,20 +57,14 @@ def get_lora_A_B(self) -> tuple[Tensor, Tensor]:
|
|||
|
||||
def __getitem__(
|
||||
self,
|
||||
indices: (
|
||||
SupportsIndex
|
||||
| slice
|
||||
| tuple[
|
||||
SupportsIndex | slice | Tensor, ...
|
||||
] # TODO: add ellipsis in the type signature
|
||||
),
|
||||
indices: SupportsIndex | slice | tuple[SupportsIndex | slice | Tensor, ...],
|
||||
) -> LoraTorchTensor:
|
||||
shape = self.shape
|
||||
if isinstance(indices, SupportsIndex):
|
||||
if len(shape) > 2:
|
||||
return LoraTorchTensor(self._lora_A[indices], self._lora_B[indices])
|
||||
else:
|
||||
raise NotImplementedError # can't return a vector
|
||||
raise NotImplementedError
|
||||
elif isinstance(indices, slice):
|
||||
if len(shape) > 2:
|
||||
return LoraTorchTensor(self._lora_A[indices], self._lora_B[indices])
|
||||
|
|
@ -82,7 +74,7 @@ def __getitem__(
|
|||
assert len(indices) > 0
|
||||
if indices[-1] is Ellipsis:
|
||||
return self[indices[:-1]]
|
||||
# expand ellipsis
|
||||
|
||||
indices = tuple(
|
||||
u
|
||||
for v in (
|
||||
|
|
@ -102,7 +94,6 @@ def __getitem__(
|
|||
*(slice(None, None) for _ in range(len(indices), len(shape))),
|
||||
)
|
||||
|
||||
# TODO: make sure this is correct
|
||||
indices_A = (
|
||||
*(
|
||||
(
|
||||
|
|
@ -118,7 +109,7 @@ def __getitem__(
|
|||
indices_B = indices[:-1]
|
||||
return LoraTorchTensor(self._lora_A[indices_A], self._lora_B[indices_B])
|
||||
else:
|
||||
raise NotImplementedError # unknown indice type
|
||||
raise NotImplementedError
|
||||
|
||||
@property
|
||||
def dtype(self) -> torch.dtype:
|
||||
|
|
@ -141,9 +132,8 @@ def reshape(self, *shape: int | tuple[int, ...]) -> LoraTorchTensor:
|
|||
new_shape = cast(tuple[int, ...], shape)
|
||||
orig_shape = self.shape
|
||||
if len(new_shape) < 2:
|
||||
raise NotImplementedError # can't become a vector
|
||||
raise NotImplementedError
|
||||
|
||||
# expand -1 in the shape
|
||||
if any(dim == -1 for dim in new_shape):
|
||||
n_elems = prod(orig_shape)
|
||||
n_new_elems = prod(dim if dim != -1 else 1 for dim in new_shape)
|
||||
|
|
@ -153,7 +143,7 @@ def reshape(self, *shape: int | tuple[int, ...]) -> LoraTorchTensor:
|
|||
)
|
||||
|
||||
if new_shape[-1] != orig_shape[-1]:
|
||||
raise NotImplementedError # can't reshape the row size trivially
|
||||
raise NotImplementedError
|
||||
|
||||
shape_A = (*(1 for _ in new_shape[:-2]), self._rank, orig_shape[-1])
|
||||
shape_B = (*new_shape[:-1], self._rank)
|
||||
|
|
@ -172,7 +162,7 @@ def permute(self, *dims: int) -> LoraTorchTensor:
|
|||
shape = self.shape
|
||||
dims = tuple(dim - len(shape) if dim >= 0 else dim for dim in dims)
|
||||
if dims[-1] == -1:
|
||||
# TODO: support higher dimensional A shapes bigger than 1
|
||||
|
||||
assert all(dim == 1 for dim in self._lora_A.shape[:-2])
|
||||
return LoraTorchTensor(self._lora_A, self._lora_B.permute(*dims))
|
||||
if len(shape) == 2 and dims[-1] == -2 and dims[-2] == -1:
|
||||
|
|
@ -180,7 +170,7 @@ def permute(self, *dims: int) -> LoraTorchTensor:
|
|||
self._lora_B.permute(*dims), self._lora_A.permute(*dims)
|
||||
)
|
||||
else:
|
||||
# TODO: compose the above two
|
||||
|
||||
raise NotImplementedError
|
||||
|
||||
def transpose(self, dim0: int, dim1: int) -> LoraTorchTensor:
|
||||
|
|
@ -199,7 +189,7 @@ def to(self, *args, **kwargs):
|
|||
|
||||
@classmethod
|
||||
def __torch_function__(cls, func: Callable, types, args=(), kwargs=None):
|
||||
del types # unused
|
||||
del types
|
||||
|
||||
if kwargs is None:
|
||||
kwargs = {}
|
||||
|
|
@ -240,73 +230,28 @@ def get_base_tensor_name(lora_tensor_name: str) -> str:
|
|||
base_name = lora_tensor_name.replace("base_model.model.", "")
|
||||
base_name = base_name.replace(".lora_A.weight", ".weight")
|
||||
base_name = base_name.replace(".lora_B.weight", ".weight")
|
||||
# models produced by mergekit-extract-lora have token embeddings in the adapter
|
||||
base_name = base_name.replace(".lora_embedding_A", ".weight")
|
||||
base_name = base_name.replace(".lora_embedding_B", ".weight")
|
||||
return base_name
|
||||
|
||||
|
||||
def parse_args() -> argparse.Namespace:
|
||||
parser = argparse.ArgumentParser(
|
||||
description="Convert a Hugging Face PEFT LoRA adapter to a GGUF file"
|
||||
)
|
||||
parser.add_argument(
|
||||
"--outfile",
|
||||
type=Path,
|
||||
help="path to write to; default: based on input. {ftype} will be replaced by the outtype.",
|
||||
)
|
||||
parser = argparse.ArgumentParser()
|
||||
parser.add_argument("--outfile", type=Path)
|
||||
parser.add_argument(
|
||||
"--outtype",
|
||||
type=str,
|
||||
choices=["f32", "f16", "bf16", "q8_0", "auto"],
|
||||
default="f16",
|
||||
help="output format - use f32 for float32, f16 for float16, bf16 for bfloat16, q8_0 for Q8_0, auto for the highest-fidelity 16-bit float type depending on the first loaded tensor type",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--bigendian",
|
||||
action="store_true",
|
||||
help="model is executed on big endian machine",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--no-lazy",
|
||||
action="store_true",
|
||||
help="use more RAM by computing all outputs before writing (use in case lazy evaluation is broken)",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--verbose",
|
||||
action="store_true",
|
||||
help="increase output verbosity",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--dry-run",
|
||||
action="store_true",
|
||||
help="only print out what will be done, without writing any new files",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--base",
|
||||
type=Path,
|
||||
help="directory containing Hugging Face model config files (config.json, tokenizer.json) for the base model that the adapter is based on - only config is needed, actual model weights are not required. If base model is unspecified, it will be loaded from Hugging Face hub based on the adapter config",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--base-model-id",
|
||||
type=str,
|
||||
help="the model ID of the base model, if it is not available locally or in the adapter config. If specified, it will ignore --base and load the base model config from the Hugging Face hub (Example: 'meta-llama/Llama-3.2-1B-Instruct')",
|
||||
)
|
||||
parser.add_argument(
|
||||
"lora_path",
|
||||
type=Path,
|
||||
help="directory containing Hugging Face PEFT LoRA config (adapter_model.json) and weights (adapter_model.safetensors or adapter_model.bin)",
|
||||
)
|
||||
parser.add_argument("--bigendian", action="store_true")
|
||||
parser.add_argument("--no-lazy", action="store_true")
|
||||
parser.add_argument("--verbose", action="store_true")
|
||||
parser.add_argument("--dry-run", action="store_true")
|
||||
parser.add_argument("--base", type=Path, required=True)
|
||||
parser.add_argument("lora_path", type=Path)
|
||||
|
||||
return parser.parse_args()
|
||||
|
||||
|
||||
def load_hparams_from_hf(hf_model_id: str) -> dict[str, Any]:
|
||||
# normally, adapter does not come with base model config, we need to load it from AutoConfig
|
||||
config = AutoConfig.from_pretrained(hf_model_id)
|
||||
return config.to_dict()
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
args = parse_args()
|
||||
logging.basicConfig(level=logging.DEBUG if args.verbose else logging.INFO)
|
||||
|
|
@ -321,20 +266,19 @@ def load_hparams_from_hf(hf_model_id: str) -> dict[str, Any]:
|
|||
|
||||
ftype = ftype_map[args.outtype]
|
||||
|
||||
dir_base_model: Path | None = args.base
|
||||
dir_base_model: Path = args.base
|
||||
dir_lora: Path = args.lora_path
|
||||
base_model_id: str | None = args.base_model_id
|
||||
lora_config = dir_lora / "adapter_config.json"
|
||||
input_model = dir_lora / "adapter_model.safetensors"
|
||||
|
||||
if args.outfile is not None:
|
||||
fname_out = args.outfile
|
||||
else:
|
||||
# output in the same directory as the model by default
|
||||
|
||||
fname_out = dir_lora
|
||||
|
||||
if os.path.exists(input_model):
|
||||
# lazy import load_file only if lora is in safetensors format.
|
||||
|
||||
from safetensors.torch import load_file
|
||||
|
||||
lora_model = load_file(input_model, device="cpu")
|
||||
|
|
@ -342,41 +286,11 @@ def load_hparams_from_hf(hf_model_id: str) -> dict[str, Any]:
|
|||
input_model = os.path.join(dir_lora, "adapter_model.bin")
|
||||
lora_model = torch.load(input_model, map_location="cpu", weights_only=True)
|
||||
|
||||
# load LoRA config
|
||||
with open(lora_config, "r") as f:
|
||||
lparams: dict[str, Any] = json.load(f)
|
||||
|
||||
# load base model
|
||||
if base_model_id is not None:
|
||||
logger.info(f"Loading base model from Hugging Face: {base_model_id}")
|
||||
hparams = load_hparams_from_hf(base_model_id)
|
||||
elif dir_base_model is None:
|
||||
if "base_model_name_or_path" in lparams:
|
||||
model_id = lparams["base_model_name_or_path"]
|
||||
logger.info(f"Loading base model from Hugging Face: {model_id}")
|
||||
try:
|
||||
hparams = load_hparams_from_hf(model_id)
|
||||
except OSError as e:
|
||||
logger.error(f"Failed to load base model config: {e}")
|
||||
logger.error(
|
||||
"Please try downloading the base model and add its path to --base"
|
||||
)
|
||||
sys.exit(1)
|
||||
else:
|
||||
logger.error(
|
||||
"'base_model_name_or_path' is not found in adapter_config.json"
|
||||
)
|
||||
logger.error(
|
||||
"Base model config is required. Please download the base model and add its path to --base"
|
||||
)
|
||||
sys.exit(1)
|
||||
else:
|
||||
logger.info(f"Loading base model: {dir_base_model.name}")
|
||||
hparams = ModelBase.load_hparams(dir_base_model)
|
||||
|
||||
hparams = Model.load_hparams(dir_base_model)
|
||||
with torch.inference_mode():
|
||||
try:
|
||||
model_class = ModelBase.from_model_architecture(hparams["architectures"][0])
|
||||
model_class = Model.from_model_architecture(hparams["architectures"][0])
|
||||
except NotImplementedError:
|
||||
logger.error(f"Model {hparams['architectures'][0]} is not supported")
|
||||
sys.exit(1)
|
||||
|
|
@ -395,9 +309,6 @@ def __init__(
|
|||
self.dir_model_card = dir_lora_model
|
||||
self.lora_alpha = float(lora_alpha)
|
||||
|
||||
def set_vocab(self):
|
||||
pass
|
||||
|
||||
def set_type(self):
|
||||
self.gguf_writer.add_type(gguf.GGUFType.ADAPTER)
|
||||
self.gguf_writer.add_string(gguf.Keys.Adapter.TYPE, "lora")
|
||||
|
|
@ -406,10 +317,7 @@ def set_gguf_parameters(self):
|
|||
self.gguf_writer.add_float32(
|
||||
gguf.Keys.Adapter.LORA_ALPHA, self.lora_alpha
|
||||
)
|
||||
|
||||
def generate_extra_tensors(self) -> Iterable[tuple[str, Tensor]]:
|
||||
# Never add extra tensors (e.g. rope_freqs) for LoRA adapters
|
||||
return ()
|
||||
super().set_gguf_parameters()
|
||||
|
||||
def get_tensors(self) -> Iterator[tuple[str, Tensor]]:
|
||||
tensor_map: dict[str, PartialLoraTensor] = {}
|
||||
|
|
@ -418,26 +326,14 @@ def get_tensors(self) -> Iterator[tuple[str, Tensor]]:
|
|||
if self.lazy:
|
||||
tensor = LazyTorchTensor.from_eager(tensor)
|
||||
base_name = get_base_tensor_name(name)
|
||||
# note: mergekit-extract-lora also adds token embeddings to the adapter
|
||||
is_lora_a = ".lora_A.weight" in name or ".lora_embedding_A" in name
|
||||
is_lora_b = ".lora_B.weight" in name or ".lora_embedding_B" in name
|
||||
is_lora_a = ".lora_A.weight" in name
|
||||
is_lora_b = ".lora_B.weight" in name
|
||||
if not is_lora_a and not is_lora_b:
|
||||
if ".base_layer.weight" in name:
|
||||
continue
|
||||
# mergekit-extract-lora add these layernorm to the adapter, we need to keep them
|
||||
if "_layernorm" in name or ".norm" in name:
|
||||
yield (base_name, tensor)
|
||||
continue
|
||||
logger.error(
|
||||
f"Unexpected name '{name}': Not a lora_A or lora_B tensor"
|
||||
)
|
||||
if ".embed_tokens.weight" in name or ".lm_head.weight" in name:
|
||||
logger.error(
|
||||
"Embeddings is present in the adapter. This can be due to new tokens added during fine tuning"
|
||||
)
|
||||
logger.error(
|
||||
"Please refer to https://github.com/ggml-org/llama.cpp/pull/9948"
|
||||
)
|
||||
sys.exit(1)
|
||||
|
||||
if base_name in tensor_map:
|
||||
|
|
@ -462,34 +358,17 @@ def get_tensors(self) -> Iterator[tuple[str, Tensor]]:
|
|||
def modify_tensors(
|
||||
self, data_torch: Tensor, name: str, bid: int | None
|
||||
) -> Iterable[tuple[str, Tensor]]:
|
||||
dest = list(super().modify_tensors(data_torch, name, bid))
|
||||
# some archs may have the same tensor for lm_head and output (tie word embeddings)
|
||||
# in this case, adapters targeting lm_head will fail when using llama-export-lora
|
||||
# therefore, we ignore them for now
|
||||
# see: https://github.com/ggml-org/llama.cpp/issues/9065
|
||||
if name == "lm_head.weight" and len(dest) == 0:
|
||||
raise ValueError(
|
||||
"lm_head is present in adapter, but is ignored in base model"
|
||||
)
|
||||
dest = super().modify_tensors(data_torch, name, bid)
|
||||
for dest_name, dest_data in dest:
|
||||
# mergekit-extract-lora add these layernorm to the adapter
|
||||
if "_norm" in dest_name:
|
||||
assert dest_data.dim() == 1
|
||||
yield (dest_name, dest_data)
|
||||
continue
|
||||
|
||||
# otherwise, we must get the lora_A and lora_B tensors
|
||||
assert isinstance(dest_data, LoraTorchTensor)
|
||||
lora_a, lora_b = dest_data.get_lora_A_B()
|
||||
|
||||
# note: mergekit-extract-lora flip and transpose A and B
|
||||
# here we only need to transpose token_embd.lora_a, see llm_build_inp_embd()
|
||||
if "token_embd.weight" in dest_name:
|
||||
lora_a = lora_a.T
|
||||
|
||||
yield (dest_name + ".lora_a", lora_a)
|
||||
yield (dest_name + ".lora_b", lora_b)
|
||||
|
||||
with open(lora_config, "r") as f:
|
||||
lparams: dict[str, Any] = json.load(f)
|
||||
|
||||
alpha: float = lparams["lora_alpha"]
|
||||
|
||||
model_instance = LoraModel(
|
||||
|
|
@ -502,7 +381,7 @@ def modify_tensors(
|
|||
dry_run=args.dry_run,
|
||||
dir_lora_model=dir_lora,
|
||||
lora_alpha=alpha,
|
||||
hparams=hparams,
|
||||
is_lora=True,
|
||||
)
|
||||
|
||||
logger.info("Exporting model...")
|
||||
|
|
|
|||
File diff suppressed because it is too large
Load Diff
|
|
@ -1,11 +0,0 @@
|
|||
import importlib
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
sys.path.insert(0, str(Path(__file__).parent.parent))
|
||||
|
||||
# Compatibility for people trying to import gguf/gguf.py directly instead of as a package.
|
||||
importlib.invalidate_caches()
|
||||
import gguf # noqa: E402
|
||||
|
||||
importlib.reload(gguf)
|
||||
|
|
@ -1,8 +1,11 @@
|
|||
#
|
||||
# GGUF file reading/modification support. For API usage information,
|
||||
# please see the files scripts/ for some fairly simple examples.
|
||||
#
|
||||
from __future__ import annotations
|
||||
|
||||
import logging
|
||||
import os
|
||||
import sys
|
||||
from collections import OrderedDict
|
||||
from typing import Any, Literal, NamedTuple, TypeVar, Union
|
||||
|
||||
|
|
@ -12,6 +15,7 @@
|
|||
from .quants import quant_shape_to_byte_shape
|
||||
|
||||
if __name__ == "__main__":
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
# Allow running file in package as a script.
|
||||
|
|
@ -24,7 +28,6 @@
|
|||
GGUF_VERSION,
|
||||
GGMLQuantizationType,
|
||||
GGUFValueType,
|
||||
GGUFEndian,
|
||||
)
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
|
@ -50,52 +53,6 @@ class ReaderField(NamedTuple):
|
|||
|
||||
types: list[GGUFValueType] = []
|
||||
|
||||
def contents(self, index_or_slice: int | slice = slice(None)) -> Any:
|
||||
if self.types:
|
||||
to_string = lambda x: str(x.tobytes(), encoding="utf-8") # noqa: E731
|
||||
main_type = self.types[0]
|
||||
|
||||
if main_type == GGUFValueType.ARRAY:
|
||||
sub_type = self.types[-1]
|
||||
|
||||
if sub_type == GGUFValueType.STRING:
|
||||
indices = self.data[index_or_slice]
|
||||
|
||||
if isinstance(index_or_slice, int):
|
||||
return to_string(self.parts[indices]) # type: ignore
|
||||
else:
|
||||
return [to_string(self.parts[idx]) for idx in indices] # type: ignore
|
||||
else:
|
||||
# FIXME: When/if _get_field_parts() support multi-dimensional arrays, this must do so too
|
||||
|
||||
# Check if it's unsafe to perform slice optimization on data
|
||||
# if any(True for idx in self.data if len(self.parts[idx]) != 1):
|
||||
# optim_slice = slice(None)
|
||||
# else:
|
||||
# optim_slice = index_or_slice
|
||||
# index_or_slice = slice(None)
|
||||
|
||||
# if isinstance(optim_slice, int):
|
||||
# return self.parts[self.data[optim_slice]].tolist()[0]
|
||||
# else:
|
||||
# return [pv for idx in self.data[optim_slice] for pv in self.parts[idx].tolist()][index_or_slice]
|
||||
|
||||
if isinstance(index_or_slice, int):
|
||||
return self.parts[self.data[index_or_slice]].tolist()[0]
|
||||
else:
|
||||
return [
|
||||
pv
|
||||
for idx in self.data[index_or_slice]
|
||||
for pv in self.parts[idx].tolist()
|
||||
]
|
||||
|
||||
if main_type == GGUFValueType.STRING:
|
||||
return to_string(self.parts[-1])
|
||||
else:
|
||||
return self.parts[-1].tolist()[0]
|
||||
|
||||
return None
|
||||
|
||||
|
||||
class ReaderTensor(NamedTuple):
|
||||
name: str
|
||||
|
|
@ -146,23 +103,12 @@ def __init__(
|
|||
# If we get 0 here that means it's (probably) a GGUF file created for
|
||||
# the opposite byte order of the machine this script is running on.
|
||||
self.byte_order = "S"
|
||||
temp_version = temp_version.view(
|
||||
temp_version.dtype.newbyteorder(self.byte_order)
|
||||
)
|
||||
temp_version = temp_version.newbyteorder(self.byte_order)
|
||||
version = temp_version[0]
|
||||
if version not in READER_SUPPORTED_VERSIONS:
|
||||
raise ValueError(
|
||||
f"Sorry, file appears to be version {version} which we cannot handle"
|
||||
)
|
||||
if sys.byteorder == "little":
|
||||
# Host is little endian
|
||||
host_endian = GGUFEndian.LITTLE
|
||||
swapped_endian = GGUFEndian.BIG
|
||||
else:
|
||||
# Sorry PDP or other weird systems that don't use BE or LE.
|
||||
host_endian = GGUFEndian.BIG
|
||||
swapped_endian = GGUFEndian.LITTLE
|
||||
self.endianess = swapped_endian if self.byte_order == "S" else host_endian
|
||||
self.fields: OrderedDict[str, ReaderField] = OrderedDict()
|
||||
self.tensors: list[ReaderTensor] = []
|
||||
offs += self._push_field(
|
||||
|
|
@ -223,11 +169,10 @@ def _get(
|
|||
count = int(count)
|
||||
itemsize = int(np.empty([], dtype=dtype).itemsize)
|
||||
end_offs = offset + itemsize * count
|
||||
arr = self.data[offset:end_offs].view(dtype=dtype)[:count]
|
||||
return arr.view(
|
||||
arr.dtype.newbyteorder(
|
||||
self.byte_order if override_order is None else override_order
|
||||
)
|
||||
return (
|
||||
self.data[offset:end_offs]
|
||||
.view(dtype=dtype)[:count]
|
||||
.newbyteorder(override_order or self.byte_order)
|
||||
)
|
||||
|
||||
def _push_field(self, field: ReaderField, skip_sum: bool = False) -> int:
|
||||
|
|
@ -274,7 +219,6 @@ def _get_field_parts(
|
|||
offs += int(alen.nbytes)
|
||||
aparts: list[npt.NDArray[Any]] = [raw_itype, alen]
|
||||
data_idxs: list[int] = []
|
||||
# FIXME: Handle multi-dimensional arrays properly instead of flattening
|
||||
for idx in range(alen[0]):
|
||||
curr_size, curr_parts, curr_idxs, curr_types = self._get_field_parts(
|
||||
offs, raw_itype[0]
|
||||
|
|
|
|||
|
|
@ -26,14 +26,12 @@
|
|||
RopeScalingType,
|
||||
PoolingType,
|
||||
TokenType,
|
||||
ExpertGatingFuncType,
|
||||
)
|
||||
|
||||
from .quants import quant_shape_from_byte_shape
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
SHARD_NAME_FORMAT = "{:s}-{:05d}-of-{:05d}.gguf"
|
||||
|
||||
|
||||
|
|
@ -137,7 +135,7 @@ def get_total_parameter_count(self) -> tuple[int, int, int, int]:
|
|||
continue
|
||||
elif name.endswith(".lora_b"):
|
||||
if last_lora_a is None or last_lora_a[0] != name[:-1] + "a":
|
||||
# Bail when the LoRA pair can't be found trivially
|
||||
|
||||
logger.warning(
|
||||
"can't measure LoRA size correctly, tensor order is unusual"
|
||||
)
|
||||
|
|
@ -156,14 +154,11 @@ def get_total_parameter_count(self) -> tuple[int, int, int, int]:
|
|||
|
||||
total_params += size
|
||||
|
||||
# Hopefully this should work even for variable-expert-count models
|
||||
expert_count = (expert_sum // n_expert_tensors) if n_expert_tensors > 0 else 0
|
||||
|
||||
# Negate the total to signal it's likely not exact
|
||||
if last_lora_a is not None:
|
||||
total_params = -total_params
|
||||
|
||||
# NOTE: keep the output in the same order as accepted by 'size_label' in gguf-py/gguf/utility.py
|
||||
return total_params, shared_params, expert_params, expert_count
|
||||
|
||||
def format_shard_names(self, path: Path) -> list[Path]:
|
||||
|
|
@ -182,7 +177,7 @@ def open_output_file(self, path: Path | None = None) -> None:
|
|||
and self.fout is not None
|
||||
and (path is None or path == self.path)
|
||||
):
|
||||
# allow calling this multiple times as long as the path is the same
|
||||
|
||||
return
|
||||
|
||||
if self.state is not WriterState.NO_FILE:
|
||||
|
|
@ -211,7 +206,7 @@ def print_plan(self) -> list[Path]:
|
|||
if self.dry_run:
|
||||
logger.info("Dry run, not writing files")
|
||||
for name in filenames:
|
||||
print(name) # noqa: NP100
|
||||
print(name)
|
||||
exit()
|
||||
|
||||
return filenames
|
||||
|
|
@ -395,12 +390,11 @@ def add_tensor_info(
|
|||
if tensor_dtype == np.uint8:
|
||||
tensor_shape = quant_shape_from_byte_shape(tensor_shape, raw_dtype)
|
||||
|
||||
# make sure there is at least one tensor before splitting
|
||||
if len(self.tensors[-1]) > 0:
|
||||
if ( # split when over tensor limit
|
||||
if (
|
||||
self.split_max_tensors != 0
|
||||
and len(self.tensors[-1]) >= self.split_max_tensors
|
||||
) or ( # split when over size limit
|
||||
) or (
|
||||
self.split_max_size != 0
|
||||
and sum(ti.nbytes for ti in self.tensors[-1].values()) + tensor_nbytes
|
||||
> self.split_max_size
|
||||
|
|
@ -466,8 +460,6 @@ def write_tensor_data(self, tensor: np.ndarray[Any, Any]) -> None:
|
|||
|
||||
fout = self.fout[file_id]
|
||||
|
||||
# pop the first tensor info
|
||||
# TODO: cleaner way to get the first key
|
||||
first_tensor_name = [
|
||||
name for name, _ in zip(self.tensors[file_id].keys(), range(1))
|
||||
][0]
|
||||
|
|
@ -514,11 +506,8 @@ def write_tensors_to_file(self, *, progress: bool = False) -> None:
|
|||
total = sum(ti.nbytes for ti in tensors.values())
|
||||
shard_bar.reset(total=(total if total > 0 else None))
|
||||
|
||||
# relying on the fact that Python dicts preserve insertion order (since 3.7)
|
||||
for ti in tensors.values():
|
||||
assert (
|
||||
ti.tensor is not None
|
||||
) # can only iterate once over the tensors
|
||||
assert ti.tensor is not None
|
||||
assert ti.tensor.nbytes == ti.nbytes
|
||||
ti.tensor.tofile(fout)
|
||||
if shard_bar is not None:
|
||||
|
|
@ -642,11 +631,6 @@ def add_base_model_organization(self, source_id: int, organization: str) -> None
|
|||
Keys.General.BASE_MODEL_ORGANIZATION.format(id=source_id), organization
|
||||
)
|
||||
|
||||
def add_base_model_description(self, source_id: int, description: str) -> None:
|
||||
self.add_string(
|
||||
Keys.General.BASE_MODEL_DESCRIPTION.format(id=source_id), description
|
||||
)
|
||||
|
||||
def add_base_model_url(self, source_id: int, url: str) -> None:
|
||||
self.add_string(Keys.General.BASE_MODEL_URL.format(id=source_id), url)
|
||||
|
||||
|
|
@ -659,46 +643,15 @@ def add_base_model_uuid(self, source_id: int, uuid: str) -> None:
|
|||
def add_base_model_repo_url(self, source_id: int, repo_url: str) -> None:
|
||||
self.add_string(Keys.General.BASE_MODEL_REPO_URL.format(id=source_id), repo_url)
|
||||
|
||||
def add_dataset_count(self, source_count: int) -> None:
|
||||
self.add_uint32(Keys.General.DATASET_COUNT, source_count)
|
||||
|
||||
def add_dataset_name(self, source_id: int, name: str) -> None:
|
||||
self.add_string(Keys.General.DATASET_NAME.format(id=source_id), name)
|
||||
|
||||
def add_dataset_author(self, source_id: int, author: str) -> None:
|
||||
self.add_string(Keys.General.DATASET_AUTHOR.format(id=source_id), author)
|
||||
|
||||
def add_dataset_version(self, source_id: int, version: str) -> None:
|
||||
self.add_string(Keys.General.DATASET_VERSION.format(id=source_id), version)
|
||||
|
||||
def add_dataset_organization(self, source_id: int, organization: str) -> None:
|
||||
self.add_string(
|
||||
Keys.General.DATASET_ORGANIZATION.format(id=source_id), organization
|
||||
)
|
||||
|
||||
def add_dataset_description(self, source_id: int, description: str) -> None:
|
||||
self.add_string(
|
||||
Keys.General.DATASET_DESCRIPTION.format(id=source_id), description
|
||||
)
|
||||
|
||||
def add_dataset_url(self, source_id: int, url: str) -> None:
|
||||
self.add_string(Keys.General.DATASET_URL.format(id=source_id), url)
|
||||
|
||||
def add_dataset_doi(self, source_id: int, doi: str) -> None:
|
||||
self.add_string(Keys.General.DATASET_DOI.format(id=source_id), doi)
|
||||
|
||||
def add_dataset_uuid(self, source_id: int, uuid: str) -> None:
|
||||
self.add_string(Keys.General.DATASET_UUID.format(id=source_id), uuid)
|
||||
|
||||
def add_dataset_repo_url(self, source_id: int, repo_url: str) -> None:
|
||||
self.add_string(Keys.General.DATASET_REPO_URL.format(id=source_id), repo_url)
|
||||
|
||||
def add_tags(self, tags: Sequence[str]) -> None:
|
||||
self.add_array(Keys.General.TAGS, tags)
|
||||
|
||||
def add_languages(self, languages: Sequence[str]) -> None:
|
||||
self.add_array(Keys.General.LANGUAGES, languages)
|
||||
|
||||
def add_datasets(self, datasets: Sequence[str]) -> None:
|
||||
self.add_array(Keys.General.DATASETS, datasets)
|
||||
|
||||
def add_tensor_data_layout(self, layout: str) -> None:
|
||||
self.add_string(Keys.LLM.TENSOR_DATA_LAYOUT.format(arch=self.arch), layout)
|
||||
|
||||
|
|
@ -711,21 +664,6 @@ def add_context_length(self, length: int) -> None:
|
|||
def add_embedding_length(self, length: int) -> None:
|
||||
self.add_uint32(Keys.LLM.EMBEDDING_LENGTH.format(arch=self.arch), length)
|
||||
|
||||
def add_features_length(self, length: int) -> None:
|
||||
self.add_uint32(Keys.LLM.FEATURES_LENGTH.format(arch=self.arch), length)
|
||||
|
||||
def add_posnet_embedding_length(self, length: int) -> None:
|
||||
self.add_uint32(Keys.PosNet.EMBEDDING_LENGTH.format(arch=self.arch), length)
|
||||
|
||||
def add_posnet_block_count(self, length: int) -> None:
|
||||
self.add_uint32(Keys.PosNet.BLOCK_COUNT.format(arch=self.arch), length)
|
||||
|
||||
def add_convnext_embedding_length(self, length: int) -> None:
|
||||
self.add_uint32(Keys.ConvNext.EMBEDDING_LENGTH.format(arch=self.arch), length)
|
||||
|
||||
def add_convnext_block_count(self, length: int) -> None:
|
||||
self.add_uint32(Keys.ConvNext.BLOCK_COUNT.format(arch=self.arch), length)
|
||||
|
||||
def add_block_count(self, length: int) -> None:
|
||||
self.add_uint32(Keys.LLM.BLOCK_COUNT.format(arch=self.arch), length)
|
||||
|
||||
|
|
@ -774,12 +712,6 @@ def add_key_length(self, length: int) -> None:
|
|||
def add_value_length(self, length: int) -> None:
|
||||
self.add_uint32(Keys.Attention.VALUE_LENGTH.format(arch=self.arch), length)
|
||||
|
||||
def add_key_length_mla(self, length: int) -> None:
|
||||
self.add_uint32(Keys.Attention.KEY_LENGTH_MLA.format(arch=self.arch), length)
|
||||
|
||||
def add_value_length_mla(self, length: int) -> None:
|
||||
self.add_uint32(Keys.Attention.VALUE_LENGTH_MLA.format(arch=self.arch), length)
|
||||
|
||||
def add_max_alibi_bias(self, bias: float) -> None:
|
||||
self.add_float32(Keys.Attention.MAX_ALIBI_BIAS.format(arch=self.arch), bias)
|
||||
|
||||
|
|
@ -807,18 +739,6 @@ def add_expert_shared_count(self, count: int) -> None:
|
|||
def add_expert_weights_scale(self, value: float) -> None:
|
||||
self.add_float32(Keys.LLM.EXPERT_WEIGHTS_SCALE.format(arch=self.arch), value)
|
||||
|
||||
def add_expert_weights_norm(self, value: bool) -> None:
|
||||
self.add_bool(Keys.LLM.EXPERT_WEIGHTS_NORM.format(arch=self.arch), value)
|
||||
|
||||
def add_expert_gating_func(self, value: ExpertGatingFuncType) -> None:
|
||||
self.add_uint32(Keys.LLM.EXPERT_GATING_FUNC.format(arch=self.arch), value.value)
|
||||
|
||||
def add_moe_every_n_layers(self, value: int) -> None:
|
||||
self.add_uint32(Keys.LLM.MOE_EVERY_N_LAYERS.format(arch=self.arch), value)
|
||||
|
||||
def add_swin_norm(self, value: bool) -> None:
|
||||
self.add_bool(Keys.LLM.SWIN_NORM.format(arch=self.arch), value)
|
||||
|
||||
def add_rescale_every_n_layers(self, count: int) -> None:
|
||||
self.add_uint32(Keys.LLM.RESCALE_EVERY_N_LAYERS.format(arch=self.arch), count)
|
||||
|
||||
|
|
@ -837,26 +757,12 @@ def add_embedding_scale(self, value: float) -> None:
|
|||
def add_wkv_head_size(self, size: int) -> None:
|
||||
self.add_uint32(Keys.WKV.HEAD_SIZE.format(arch=self.arch), size)
|
||||
|
||||
def add_token_shift_count(self, count: int) -> None:
|
||||
self.add_uint32(Keys.LLM.TOKEN_SHIFT_COUNT.format(arch=self.arch), count)
|
||||
|
||||
def add_interleave_moe_layer_step(self, value: int) -> None:
|
||||
self.add_uint32(
|
||||
Keys.LLM.INTERLEAVE_MOE_LAYER_STEP.format(arch=self.arch), value
|
||||
)
|
||||
|
||||
def add_layer_norm_eps(self, value: float) -> None:
|
||||
self.add_float32(Keys.Attention.LAYERNORM_EPS.format(arch=self.arch), value)
|
||||
|
||||
def add_layer_norm_rms_eps(self, value: float) -> None:
|
||||
self.add_float32(Keys.Attention.LAYERNORM_RMS_EPS.format(arch=self.arch), value)
|
||||
|
||||
def add_group_norm_eps(self, value: float) -> None:
|
||||
self.add_float32(Keys.Attention.GROUPNORM_EPS.format(arch=self.arch), value)
|
||||
|
||||
def add_group_norm_groups(self, value: int) -> None:
|
||||
self.add_uint32(Keys.Attention.GROUPNORM_GROUPS.format(arch=self.arch), value)
|
||||
|
||||
def add_causal_attention(self, value: bool) -> None:
|
||||
self.add_bool(Keys.Attention.CAUSAL.format(arch=self.arch), value)
|
||||
|
||||
|
|
@ -866,20 +772,6 @@ def add_q_lora_rank(self, length: int) -> None:
|
|||
def add_kv_lora_rank(self, length: int) -> None:
|
||||
self.add_uint32(Keys.Attention.KV_LORA_RANK.format(arch=self.arch), length)
|
||||
|
||||
def add_decay_lora_rank(self, length: int) -> None:
|
||||
self.add_uint32(Keys.Attention.DECAY_LORA_RANK.format(arch=self.arch), length)
|
||||
|
||||
def add_iclr_lora_rank(self, length: int) -> None:
|
||||
self.add_uint32(Keys.Attention.ICLR_LORA_RANK.format(arch=self.arch), length)
|
||||
|
||||
def add_value_residual_mix_lora_rank(self, length: int) -> None:
|
||||
self.add_uint32(
|
||||
Keys.Attention.VALUE_RESIDUAL_MIX_LORA_RANK.format(arch=self.arch), length
|
||||
)
|
||||
|
||||
def add_gate_lora_rank(self, length: int) -> None:
|
||||
self.add_uint32(Keys.Attention.GATE_LORA_RANK.format(arch=self.arch), length)
|
||||
|
||||
def add_relative_attn_buckets_count(self, value: int) -> None:
|
||||
self.add_uint32(Keys.Attention.REL_BUCKETS_COUNT.format(arch=self.arch), value)
|
||||
|
||||
|
|
@ -895,9 +787,6 @@ def add_pooling_type(self, value: PoolingType) -> None:
|
|||
def add_rope_dimension_count(self, count: int) -> None:
|
||||
self.add_uint32(Keys.Rope.DIMENSION_COUNT.format(arch=self.arch), count)
|
||||
|
||||
def add_rope_dimension_sections(self, dims: Sequence[int]) -> None:
|
||||
self.add_array(Keys.Rope.DIMENSION_SECTIONS.format(arch=self.arch), dims)
|
||||
|
||||
def add_rope_freq_base(self, value: float) -> None:
|
||||
self.add_float32(Keys.Rope.FREQ_BASE.format(arch=self.arch), value)
|
||||
|
||||
|
|
@ -974,6 +863,9 @@ def add_sep_token_id(self, id: int) -> None:
|
|||
def add_pad_token_id(self, id: int) -> None:
|
||||
self.add_uint32(Keys.Tokenizer.PAD_ID, id)
|
||||
|
||||
def add_cls_token_id(self, id: int) -> None:
|
||||
self.add_uint32(Keys.Tokenizer.CLS_ID, id)
|
||||
|
||||
def add_mask_token_id(self, id: int) -> None:
|
||||
self.add_uint32(Keys.Tokenizer.MASK_ID, id)
|
||||
|
||||
|
|
@ -1001,7 +893,6 @@ def add_chat_template(self, value: str | Sequence[Mapping[str, str]]) -> None:
|
|||
name = choice.get("name", "")
|
||||
template = choice.get("template")
|
||||
|
||||
# Allowing non-alphanumerical characters in template name is probably not a good idea, so filter it
|
||||
name = "".join(
|
||||
(c if c in ascii_letters + digits else "_" for c in name)
|
||||
)
|
||||
|
|
@ -1025,65 +916,21 @@ def add_chat_template(self, value: str | Sequence[Mapping[str, str]]) -> None:
|
|||
|
||||
self.add_string(Keys.Tokenizer.CHAT_TEMPLATE, value)
|
||||
|
||||
def add_prefix_token_id(self, id: int) -> None:
|
||||
self.add_uint32(Keys.Tokenizer.PREFIX_ID, id)
|
||||
|
||||
def add_suffix_token_id(self, id: int) -> None:
|
||||
self.add_uint32(Keys.Tokenizer.SUFFIX_ID, id)
|
||||
|
||||
def add_middle_token_id(self, id: int) -> None:
|
||||
self.add_uint32(Keys.Tokenizer.MIDDLE_ID, id)
|
||||
|
||||
def add_eot_token_id(self, id: int) -> None:
|
||||
self.add_uint32(Keys.Tokenizer.EOT_ID, id)
|
||||
|
||||
def add_eom_token_id(self, id: int) -> None:
|
||||
self.add_uint32(Keys.Tokenizer.EOM_ID, id)
|
||||
|
||||
# for vision models
|
||||
|
||||
def add_vision_projection_dim(self, value: int) -> None:
|
||||
self.add_uint32(Keys.ClipVision.PROJECTION_DIM, value)
|
||||
|
||||
def add_vision_has_vision_encoder(self, value: bool) -> None:
|
||||
self.add_bool(Keys.ClipVision.HAS_VISION_ENCODER, value)
|
||||
|
||||
def add_vision_patch_size(self, value: int) -> None:
|
||||
self.add_uint32(Keys.ClipVision.PATCH_SIZE, value)
|
||||
|
||||
def add_vision_embedding_length(self, value: int) -> None:
|
||||
self.add_uint32(Keys.ClipVision.EMBEDDING_LENGTH, value)
|
||||
|
||||
def add_vision_feed_forward_length(self, value: int) -> None:
|
||||
self.add_uint32(Keys.ClipVision.FEED_FORWARD_LENGTH, value)
|
||||
|
||||
def add_vision_block_count(self, value: int) -> None:
|
||||
self.add_uint32(Keys.ClipVision.BLOCK_COUNT, value)
|
||||
|
||||
def add_vision_head_count(self, value: int) -> None:
|
||||
self.add_uint32(Keys.ClipVision.Attention.HEAD_COUNT, value)
|
||||
|
||||
def add_vision_projector_type(self, value: str) -> None:
|
||||
self.add_string(Keys.ClipVision.PROJECTOR_TYPE, value)
|
||||
|
||||
def add_vision_attention_layernorm_eps(self, value: float) -> None:
|
||||
self.add_float32(Keys.ClipVision.Attention.LAYERNORM_EPS, value)
|
||||
|
||||
def add_vision_image_size(self, value: int) -> None:
|
||||
self.add_uint32(Keys.ClipVision.IMAGE_SIZE, value)
|
||||
|
||||
def add_vision_image_mean(self, values: Sequence[float]) -> None:
|
||||
self.add_array(Keys.ClipVision.IMAGE_MEAN, values)
|
||||
|
||||
def add_vision_image_std(self, values: Sequence[float]) -> None:
|
||||
self.add_array(Keys.ClipVision.IMAGE_STD, values)
|
||||
|
||||
def add_vision_spatial_merge_size(self, value: int) -> None:
|
||||
self.add_uint32(Keys.ClipVision.SPATIAL_MERGE_SIZE, value)
|
||||
|
||||
def add_vision_use_gelu(self, value: bool) -> None:
|
||||
self.add_bool(Keys.ClipVision.USE_GELU, value)
|
||||
|
||||
def add_vision_use_silu(self, value: bool) -> None:
|
||||
self.add_bool(Keys.ClipVision.USE_SILU, value)
|
||||
|
||||
def add_vision_projector_scale_factor(self, value: int) -> None:
|
||||
self.add_uint32(Keys.ClipVision.Projector.SCALE_FACTOR, value)
|
||||
|
||||
def add_vision_n_wa_pattern(self, value: int) -> None:
|
||||
self.add_uint32(Keys.ClipVision.N_WA_PATTERN, value)
|
||||
|
||||
def _pack(self, fmt: str, value: Any, skip_pack_prefix: bool = False) -> bytes:
|
||||
pack_prefix = ""
|
||||
if not skip_pack_prefix:
|
||||
|
|
|
|||
|
|
@ -12,7 +12,6 @@
|
|||
|
||||
|
||||
class LazyMeta(ABCMeta):
|
||||
|
||||
def __new__(
|
||||
cls, name: str, bases: tuple[type, ...], namespace: dict[str, Any], **kwargs
|
||||
):
|
||||
|
|
@ -35,7 +34,7 @@ def __getattr__(self, name: str) -> Any:
|
|||
|
||||
# need to make a builder for the wrapped wrapper to copy the name,
|
||||
# or else it fails with very cryptic error messages,
|
||||
# because somehow the same string would end up in every closures
|
||||
# because somehow the same string would end up in every closure
|
||||
def mk_wrap(op_name: str, *, meta_noop: bool = False):
|
||||
# need to wrap the wrapper to get self
|
||||
def wrapped_special_op(self, *args, **kwargs):
|
||||
|
|
@ -201,27 +200,6 @@ def wrapped_fn(*args, **kwargs):
|
|||
return cls(
|
||||
meta=cls.eager_to_meta(res), args=args, kwargs=kwargs, func=fn
|
||||
)
|
||||
elif isinstance(res, tuple) and all(
|
||||
isinstance(t, cls._tensor_type) for t in res
|
||||
):
|
||||
# share the evaluation between lazy tuple elements
|
||||
shared_args: list = [args, None]
|
||||
|
||||
def eager_tuple_element(a: list[Any], i: int = 0, /, **kw) -> LazyBase:
|
||||
assert len(a) == 2
|
||||
if a[1] is None:
|
||||
a[1] = fn(*a[0], **kw)
|
||||
return a[1][i]
|
||||
|
||||
return tuple(
|
||||
cls(
|
||||
meta=cls.eager_to_meta(res[i]),
|
||||
args=(shared_args, i),
|
||||
kwargs=kwargs,
|
||||
func=eager_tuple_element,
|
||||
)
|
||||
for i in range(len(res))
|
||||
)
|
||||
else:
|
||||
del res # not needed
|
||||
# non-tensor return likely relies on the contents of the args
|
||||
|
|
@ -276,8 +254,6 @@ def from_eager(cls, t: Any) -> Any:
|
|||
class LazyNumpyTensor(LazyBase):
|
||||
_tensor_type = np.ndarray
|
||||
|
||||
shape: tuple[int, ...] # Makes the type checker happy in quants.py
|
||||
|
||||
@classmethod
|
||||
def meta_with_dtype_and_shape(
|
||||
cls, dtype: DTypeLike, shape: tuple[int, ...]
|
||||
|
|
|
|||
|
|
@ -41,7 +41,7 @@ class Metadata:
|
|||
base_models: Optional[list[dict]] = None
|
||||
tags: Optional[list[str]] = None
|
||||
languages: Optional[list[str]] = None
|
||||
datasets: Optional[list[dict]] = None
|
||||
datasets: Optional[list[str]] = None
|
||||
|
||||
@staticmethod
|
||||
def load(
|
||||
|
|
@ -50,7 +50,7 @@ def load(
|
|||
model_name: Optional[str] = None,
|
||||
total_params: int = 0,
|
||||
) -> Metadata:
|
||||
# This grabs as many contextual authorship metadata as possible from the model repository
|
||||
# This grabs as much contextual authorship metadata as possible from the model repository
|
||||
# making any conversion as required to match the gguf kv store metadata format
|
||||
# as well as giving users the ability to override any authorship metadata that may be incorrect
|
||||
|
||||
|
|
@ -126,13 +126,13 @@ def load(
|
|||
"general.base_models", metadata.base_models
|
||||
)
|
||||
|
||||
# Datasets is received here as an array of datasets
|
||||
metadata.datasets = metadata_override.get("general.datasets", metadata.datasets)
|
||||
|
||||
metadata.tags = metadata_override.get(Keys.General.TAGS, metadata.tags)
|
||||
metadata.languages = metadata_override.get(
|
||||
Keys.General.LANGUAGES, metadata.languages
|
||||
)
|
||||
metadata.datasets = metadata_override.get(
|
||||
Keys.General.DATASETS, metadata.datasets
|
||||
)
|
||||
|
||||
# Direct Metadata Override (via direct cli argument)
|
||||
if model_name is not None:
|
||||
|
|
@ -160,32 +160,12 @@ def load_model_card(model_path: Optional[Path] = None) -> dict[str, Any]:
|
|||
if not model_card_path.is_file():
|
||||
return {}
|
||||
|
||||
# The model card metadata is assumed to always be in YAML (frontmatter)
|
||||
# The model card metadata is assumed to always be in YAML
|
||||
# ref: https://github.com/huggingface/transformers/blob/a5c642fe7a1f25d3bdcd76991443ba6ff7ee34b2/src/transformers/modelcard.py#L468-L473
|
||||
yaml_content: str = ""
|
||||
with open(model_card_path, "r", encoding="utf-8") as f:
|
||||
content = f.read()
|
||||
lines = content.splitlines()
|
||||
lines_yaml = []
|
||||
if len(lines) == 0:
|
||||
# Empty file
|
||||
return {}
|
||||
if len(lines) > 0 and lines[0] != "---":
|
||||
# No frontmatter
|
||||
return {}
|
||||
for line in lines[1:]:
|
||||
if line == "---":
|
||||
break # End of frontmatter
|
||||
else:
|
||||
lines_yaml.append(line)
|
||||
yaml_content = "\n".join(lines_yaml) + "\n"
|
||||
|
||||
# Quick hack to fix the Norway problem
|
||||
# https://hitchdev.com/strictyaml/why/implicit-typing-removed/
|
||||
yaml_content = yaml_content.replace("- no\n", '- "no"\n')
|
||||
|
||||
if yaml_content:
|
||||
data = yaml.safe_load(yaml_content)
|
||||
if f.readline() == "---\n":
|
||||
raw = f.read().partition("---\n")[0]
|
||||
data = yaml.safe_load(raw)
|
||||
if isinstance(data, dict):
|
||||
return data
|
||||
else:
|
||||
|
|
@ -248,11 +228,7 @@ def get_model_id_components(
|
|||
org_component, model_full_name_component = None, model_id
|
||||
|
||||
# Check if we erroneously matched against './' or '../' etc...
|
||||
if (
|
||||
org_component is not None
|
||||
and len(org_component) > 0
|
||||
and org_component[0] == "."
|
||||
):
|
||||
if org_component is not None and org_component[0] == ".":
|
||||
org_component = None
|
||||
|
||||
name_parts: list[str] = model_full_name_component.split("-")
|
||||
|
|
@ -411,86 +387,27 @@ def apply_metadata_heuristic(
|
|||
########################
|
||||
if model_card is not None:
|
||||
|
||||
def use_model_card_metadata(metadata_key: str, model_card_key: str):
|
||||
if (
|
||||
model_card_key in model_card
|
||||
and getattr(metadata, metadata_key, None) is None
|
||||
):
|
||||
setattr(metadata, metadata_key, model_card.get(model_card_key))
|
||||
|
||||
def use_array_model_card_metadata(metadata_key: str, model_card_key: str):
|
||||
# Note: Will append rather than replace if already exist
|
||||
tags_value = model_card.get(model_card_key, None)
|
||||
if tags_value is None:
|
||||
return
|
||||
|
||||
current_value = getattr(metadata, metadata_key, None)
|
||||
if current_value is None:
|
||||
current_value = []
|
||||
|
||||
if isinstance(tags_value, str):
|
||||
current_value.append(tags_value)
|
||||
elif isinstance(tags_value, list):
|
||||
current_value.extend(tags_value)
|
||||
|
||||
setattr(metadata, metadata_key, current_value)
|
||||
|
||||
# LLAMA.cpp's direct internal convention
|
||||
# (Definitely not part of hugging face formal/informal standard)
|
||||
#########################################
|
||||
use_model_card_metadata("name", "name")
|
||||
use_model_card_metadata("author", "author")
|
||||
use_model_card_metadata("version", "version")
|
||||
use_model_card_metadata("organization", "organization")
|
||||
use_model_card_metadata("description", "description")
|
||||
use_model_card_metadata("finetune", "finetune")
|
||||
use_model_card_metadata("basename", "basename")
|
||||
use_model_card_metadata("size_label", "size_label")
|
||||
use_model_card_metadata("source_url", "url")
|
||||
use_model_card_metadata("source_doi", "doi")
|
||||
use_model_card_metadata("source_uuid", "uuid")
|
||||
use_model_card_metadata("source_repo_url", "repo_url")
|
||||
|
||||
# LLAMA.cpp's huggingface style convention
|
||||
# (Definitely not part of hugging face formal/informal standard... but with model_ appended to match their style)
|
||||
###########################################
|
||||
use_model_card_metadata("name", "model_name")
|
||||
use_model_card_metadata("author", "model_author")
|
||||
use_model_card_metadata("version", "model_version")
|
||||
use_model_card_metadata("organization", "model_organization")
|
||||
use_model_card_metadata("description", "model_description")
|
||||
use_model_card_metadata("finetune", "model_finetune")
|
||||
use_model_card_metadata("basename", "model_basename")
|
||||
use_model_card_metadata("size_label", "model_size_label")
|
||||
use_model_card_metadata("source_url", "model_url")
|
||||
use_model_card_metadata("source_doi", "model_doi")
|
||||
use_model_card_metadata("source_uuid", "model_uuid")
|
||||
use_model_card_metadata("source_repo_url", "model_repo_url")
|
||||
|
||||
# Hugging Face Direct Convention
|
||||
#################################
|
||||
|
||||
if "model_name" in model_card and metadata.name is None:
|
||||
# Not part of huggingface model card standard but notice some model creator using it
|
||||
# such as TheBloke in 'TheBloke/Mistral-7B-Instruct-v0.2-GGUF'
|
||||
use_model_card_metadata("name", "model_name")
|
||||
use_model_card_metadata("author", "model_creator")
|
||||
use_model_card_metadata("basename", "model_type")
|
||||
metadata.name = model_card.get("model_name")
|
||||
|
||||
if (
|
||||
"base_model" in model_card
|
||||
or "base_models" in model_card
|
||||
or "base_model_sources" in model_card
|
||||
):
|
||||
if "model_creator" in model_card and metadata.author is None:
|
||||
# Not part of huggingface model card standard but notice some model creator using it
|
||||
# such as TheBloke in 'TheBloke/Mistral-7B-Instruct-v0.2-GGUF'
|
||||
metadata.author = model_card.get("model_creator")
|
||||
|
||||
if "model_type" in model_card and metadata.basename is None:
|
||||
# Not part of huggingface model card standard but notice some model creator using it
|
||||
# such as TheBloke in 'TheBloke/Mistral-7B-Instruct-v0.2-GGUF'
|
||||
metadata.basename = model_card.get("model_type")
|
||||
|
||||
if "base_model" in model_card:
|
||||
# This represents the parent models that this is based on
|
||||
# Example: stabilityai/stable-diffusion-xl-base-1.0. Can also be a list (for merges)
|
||||
# Example of merges: https://huggingface.co/EmbeddedLLM/Mistral-7B-Merge-14-v0.1/blob/main/README.md
|
||||
metadata_base_models = []
|
||||
base_model_value = model_card.get(
|
||||
"base_model",
|
||||
model_card.get(
|
||||
"base_models", model_card.get("base_model_sources", None)
|
||||
),
|
||||
)
|
||||
base_model_value = model_card.get("base_model", None)
|
||||
|
||||
if base_model_value is not None:
|
||||
if isinstance(base_model_value, str):
|
||||
|
|
@ -503,47 +420,6 @@ def use_array_model_card_metadata(metadata_key: str, model_card_key: str):
|
|||
|
||||
for model_id in metadata_base_models:
|
||||
# NOTE: model size of base model is assumed to be similar to the size of the current model
|
||||
base_model = {}
|
||||
if isinstance(model_id, str):
|
||||
if (
|
||||
model_id.startswith("http://")
|
||||
or model_id.startswith("https://")
|
||||
or model_id.startswith("ssh://")
|
||||
):
|
||||
base_model["repo_url"] = model_id
|
||||
|
||||
# Check if Hugging Face ID is present in URL
|
||||
if "huggingface.co" in model_id:
|
||||
match = re.match(
|
||||
r"https?://huggingface.co/([^/]+/[^/]+)$", model_id
|
||||
)
|
||||
if match:
|
||||
model_id_component = match.group(1)
|
||||
(
|
||||
model_full_name_component,
|
||||
org_component,
|
||||
basename,
|
||||
finetune,
|
||||
version,
|
||||
size_label,
|
||||
) = Metadata.get_model_id_components(
|
||||
model_id_component, total_params
|
||||
)
|
||||
|
||||
# Populate model dictionary with extracted components
|
||||
if model_full_name_component is not None:
|
||||
base_model["name"] = Metadata.id_to_title(
|
||||
model_full_name_component
|
||||
)
|
||||
if org_component is not None:
|
||||
base_model["organization"] = (
|
||||
Metadata.id_to_title(org_component)
|
||||
)
|
||||
if version is not None:
|
||||
base_model["version"] = version
|
||||
|
||||
else:
|
||||
# Likely a Hugging Face ID
|
||||
(
|
||||
model_full_name_component,
|
||||
org_component,
|
||||
|
|
@ -552,16 +428,13 @@ def use_array_model_card_metadata(metadata_key: str, model_card_key: str):
|
|||
version,
|
||||
size_label,
|
||||
) = Metadata.get_model_id_components(model_id, total_params)
|
||||
|
||||
# Populate model dictionary with extracted components
|
||||
base_model = {}
|
||||
if model_full_name_component is not None:
|
||||
base_model["name"] = Metadata.id_to_title(
|
||||
model_full_name_component
|
||||
)
|
||||
if org_component is not None:
|
||||
base_model["organization"] = Metadata.id_to_title(
|
||||
org_component
|
||||
)
|
||||
base_model["organization"] = Metadata.id_to_title(org_component)
|
||||
if version is not None:
|
||||
base_model["version"] = version
|
||||
if (
|
||||
|
|
@ -571,127 +444,62 @@ def use_array_model_card_metadata(metadata_key: str, model_card_key: str):
|
|||
base_model["repo_url"] = (
|
||||
f"https://huggingface.co/{org_component}/{model_full_name_component}"
|
||||
)
|
||||
|
||||
elif isinstance(model_id, dict):
|
||||
base_model = model_id
|
||||
|
||||
else:
|
||||
logger.error(
|
||||
f"base model entry '{str(model_id)}' not in a known format"
|
||||
)
|
||||
|
||||
metadata.base_models.append(base_model)
|
||||
|
||||
if (
|
||||
"datasets" in model_card
|
||||
or "dataset" in model_card
|
||||
or "dataset_sources" in model_card
|
||||
):
|
||||
# This represents the datasets that this was trained from
|
||||
metadata_datasets = []
|
||||
dataset_value = model_card.get(
|
||||
"datasets",
|
||||
model_card.get("dataset", model_card.get("dataset_sources", None)),
|
||||
)
|
||||
if "license" in model_card and metadata.license is None:
|
||||
metadata.license = model_card.get("license")
|
||||
|
||||
if "license_name" in model_card and metadata.license_name is None:
|
||||
metadata.license_name = model_card.get("license_name")
|
||||
|
||||
if "license_link" in model_card and metadata.license_link is None:
|
||||
metadata.license_link = model_card.get("license_link")
|
||||
|
||||
tags_value = model_card.get("tags", None)
|
||||
if tags_value is not None:
|
||||
|
||||
if metadata.tags is None:
|
||||
metadata.tags = []
|
||||
|
||||
if isinstance(tags_value, str):
|
||||
metadata.tags.append(tags_value)
|
||||
elif isinstance(tags_value, list):
|
||||
metadata.tags.extend(tags_value)
|
||||
|
||||
pipeline_tags_value = model_card.get("pipeline_tag", None)
|
||||
if pipeline_tags_value is not None:
|
||||
|
||||
if metadata.tags is None:
|
||||
metadata.tags = []
|
||||
|
||||
if isinstance(pipeline_tags_value, str):
|
||||
metadata.tags.append(pipeline_tags_value)
|
||||
elif isinstance(pipeline_tags_value, list):
|
||||
metadata.tags.extend(pipeline_tags_value)
|
||||
|
||||
language_value = model_card.get(
|
||||
"languages", model_card.get("language", None)
|
||||
)
|
||||
if language_value is not None:
|
||||
|
||||
if metadata.languages is None:
|
||||
metadata.languages = []
|
||||
|
||||
if isinstance(language_value, str):
|
||||
metadata.languages.append(language_value)
|
||||
elif isinstance(language_value, list):
|
||||
metadata.languages.extend(language_value)
|
||||
|
||||
dataset_value = model_card.get("datasets", model_card.get("dataset", None))
|
||||
if dataset_value is not None:
|
||||
if isinstance(dataset_value, str):
|
||||
metadata_datasets.append(dataset_value)
|
||||
elif isinstance(dataset_value, list):
|
||||
metadata_datasets.extend(dataset_value)
|
||||
|
||||
if metadata.datasets is None:
|
||||
metadata.datasets = []
|
||||
|
||||
for dataset_id in metadata_datasets:
|
||||
# NOTE: model size of base model is assumed to be similar to the size of the current model
|
||||
dataset = {}
|
||||
if isinstance(dataset_id, str):
|
||||
if dataset_id.startswith(("http://", "https://", "ssh://")):
|
||||
dataset["repo_url"] = dataset_id
|
||||
|
||||
# Check if Hugging Face ID is present in URL
|
||||
if "huggingface.co" in dataset_id:
|
||||
match = re.match(
|
||||
r"https?://huggingface.co/([^/]+/[^/]+)$",
|
||||
dataset_id,
|
||||
)
|
||||
if match:
|
||||
dataset_id_component = match.group(1)
|
||||
(
|
||||
dataset_name_component,
|
||||
org_component,
|
||||
basename,
|
||||
finetune,
|
||||
version,
|
||||
size_label,
|
||||
) = Metadata.get_model_id_components(
|
||||
dataset_id_component, total_params
|
||||
)
|
||||
|
||||
# Populate dataset dictionary with extracted components
|
||||
if dataset_name_component is not None:
|
||||
dataset["name"] = Metadata.id_to_title(
|
||||
dataset_name_component
|
||||
)
|
||||
if org_component is not None:
|
||||
dataset["organization"] = Metadata.id_to_title(
|
||||
org_component
|
||||
)
|
||||
if version is not None:
|
||||
dataset["version"] = version
|
||||
|
||||
else:
|
||||
# Likely a Hugging Face ID
|
||||
(
|
||||
dataset_name_component,
|
||||
org_component,
|
||||
basename,
|
||||
finetune,
|
||||
version,
|
||||
size_label,
|
||||
) = Metadata.get_model_id_components(
|
||||
dataset_id, total_params
|
||||
)
|
||||
|
||||
# Populate dataset dictionary with extracted components
|
||||
if dataset_name_component is not None:
|
||||
dataset["name"] = Metadata.id_to_title(
|
||||
dataset_name_component
|
||||
)
|
||||
if org_component is not None:
|
||||
dataset["organization"] = Metadata.id_to_title(
|
||||
org_component
|
||||
)
|
||||
if version is not None:
|
||||
dataset["version"] = version
|
||||
if (
|
||||
org_component is not None
|
||||
and dataset_name_component is not None
|
||||
):
|
||||
dataset["repo_url"] = (
|
||||
f"https://huggingface.co/{org_component}/{dataset_name_component}"
|
||||
)
|
||||
|
||||
elif isinstance(dataset_id, dict):
|
||||
dataset = dataset_id
|
||||
|
||||
else:
|
||||
logger.error(
|
||||
f"dataset entry '{str(dataset_id)}' not in a known format"
|
||||
)
|
||||
|
||||
metadata.datasets.append(dataset)
|
||||
|
||||
use_model_card_metadata("license", "license")
|
||||
use_model_card_metadata("license_name", "license_name")
|
||||
use_model_card_metadata("license_link", "license_link")
|
||||
|
||||
use_array_model_card_metadata("tags", "tags")
|
||||
use_array_model_card_metadata("tags", "pipeline_tag")
|
||||
|
||||
use_array_model_card_metadata("languages", "languages")
|
||||
use_array_model_card_metadata("languages", "language")
|
||||
if isinstance(dataset_value, str):
|
||||
metadata.datasets.append(dataset_value)
|
||||
elif isinstance(dataset_value, list):
|
||||
metadata.datasets.extend(dataset_value)
|
||||
|
||||
# Hugging Face Parameter Heuristics
|
||||
####################################
|
||||
|
|
@ -700,7 +508,7 @@ def use_array_model_card_metadata(metadata_key: str, model_card_key: str):
|
|||
|
||||
hf_name_or_path = hf_params.get("_name_or_path")
|
||||
if hf_name_or_path is not None and hf_name_or_path.count("/") <= 1:
|
||||
# Use _name_or_path only if its actually a model name and not some computer path
|
||||
# Use _name_or_path only if it's actually a model name and not some computer path
|
||||
# e.g. 'meta-llama/Llama-2-7b-hf'
|
||||
model_id = hf_name_or_path
|
||||
(
|
||||
|
|
@ -776,9 +584,6 @@ def set_gguf_meta_model(self, gguf_writer: gguf.GGUFWriter):
|
|||
gguf_writer.add_size_label(self.size_label)
|
||||
|
||||
if self.license is not None:
|
||||
if isinstance(self.license, list):
|
||||
gguf_writer.add_license(",".join(self.license))
|
||||
else:
|
||||
gguf_writer.add_license(self.license)
|
||||
if self.license_name is not None:
|
||||
gguf_writer.add_license_name(self.license_name)
|
||||
|
|
@ -816,10 +621,6 @@ def set_gguf_meta_model(self, gguf_writer: gguf.GGUFWriter):
|
|||
gguf_writer.add_base_model_organization(
|
||||
key, base_model_entry["organization"]
|
||||
)
|
||||
if "description" in base_model_entry:
|
||||
gguf_writer.add_base_model_description(
|
||||
key, base_model_entry["description"]
|
||||
)
|
||||
if "url" in base_model_entry:
|
||||
gguf_writer.add_base_model_url(key, base_model_entry["url"])
|
||||
if "doi" in base_model_entry:
|
||||
|
|
@ -831,33 +632,9 @@ def set_gguf_meta_model(self, gguf_writer: gguf.GGUFWriter):
|
|||
key, base_model_entry["repo_url"]
|
||||
)
|
||||
|
||||
if self.datasets is not None:
|
||||
gguf_writer.add_dataset_count(len(self.datasets))
|
||||
for key, dataset_entry in enumerate(self.datasets):
|
||||
if "name" in dataset_entry:
|
||||
gguf_writer.add_dataset_name(key, dataset_entry["name"])
|
||||
if "author" in dataset_entry:
|
||||
gguf_writer.add_dataset_author(key, dataset_entry["author"])
|
||||
if "version" in dataset_entry:
|
||||
gguf_writer.add_dataset_version(key, dataset_entry["version"])
|
||||
if "organization" in dataset_entry:
|
||||
gguf_writer.add_dataset_organization(
|
||||
key, dataset_entry["organization"]
|
||||
)
|
||||
if "description" in dataset_entry:
|
||||
gguf_writer.add_dataset_description(
|
||||
key, dataset_entry["description"]
|
||||
)
|
||||
if "url" in dataset_entry:
|
||||
gguf_writer.add_dataset_url(key, dataset_entry["url"])
|
||||
if "doi" in dataset_entry:
|
||||
gguf_writer.add_dataset_doi(key, dataset_entry["doi"])
|
||||
if "uuid" in dataset_entry:
|
||||
gguf_writer.add_dataset_uuid(key, dataset_entry["uuid"])
|
||||
if "repo_url" in dataset_entry:
|
||||
gguf_writer.add_dataset_repo_url(key, dataset_entry["repo_url"])
|
||||
|
||||
if self.tags is not None:
|
||||
gguf_writer.add_tags(self.tags)
|
||||
if self.languages is not None:
|
||||
gguf_writer.add_languages(self.languages)
|
||||
if self.datasets is not None:
|
||||
gguf_writer.add_datasets(self.datasets)
|
||||
|
|
|
|||
1443
src/gguf/quants.py
1443
src/gguf/quants.py
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
|
|
@ -1,11 +1,7 @@
|
|||
from __future__ import annotations
|
||||
|
||||
from dataclasses import dataclass
|
||||
from typing import Literal
|
||||
|
||||
import os
|
||||
import json
|
||||
|
||||
|
||||
def fill_templated_filename(filename: str, output_type: str | None) -> str:
|
||||
# Given a file name fill in any type templates e.g. 'some-model-name.{ftype}.gguf'
|
||||
|
|
@ -71,7 +67,7 @@ def naming_convention(
|
|||
output_type: str | None,
|
||||
model_type: Literal["vocab", "LoRA"] | None = None,
|
||||
) -> str:
|
||||
# Reference: https://github.com/ggml-org/ggml/blob/master/docs/gguf.md#gguf-naming-convention
|
||||
# Reference: https://github.com/ggerganov/ggml/blob/master/docs/gguf.md#gguf-naming-convention
|
||||
|
||||
if base_name is not None:
|
||||
name = base_name.strip().replace(" ", "-").replace("/", "-")
|
||||
|
|
@ -103,214 +99,3 @@ def naming_convention(
|
|||
kind = f"-{model_type.strip().replace(' ', '-')}" if model_type is not None else ""
|
||||
|
||||
return f"{name}{parameters}{finetune}{version}{encoding}{kind}"
|
||||
|
||||
|
||||
@dataclass
|
||||
class RemoteTensor:
|
||||
dtype: str
|
||||
shape: tuple[int, ...]
|
||||
offset_start: int
|
||||
size: int
|
||||
url: str
|
||||
|
||||
def data(self) -> bytearray:
|
||||
# TODO: handle request errors (maybe with limited retries?)
|
||||
# NOTE: using a bytearray, otherwise PyTorch complains the buffer is not writeable
|
||||
data = bytearray(
|
||||
SafetensorRemote.get_data_by_range(
|
||||
url=self.url, start=self.offset_start, size=self.size
|
||||
)
|
||||
)
|
||||
return data
|
||||
|
||||
|
||||
class SafetensorRemote:
|
||||
"""
|
||||
Uility class to handle remote safetensor files.
|
||||
This class is designed to work with Hugging Face model repositories.
|
||||
|
||||
Example (one model has single safetensor file, the other has multiple):
|
||||
for model_id in ["ngxson/TEST-Tiny-Llama4", "Qwen/Qwen2.5-7B-Instruct"]:
|
||||
tensors = SafetensorRemote.get_list_tensors_hf_model(model_id)
|
||||
print(tensors)
|
||||
|
||||
Example reading tensor data:
|
||||
tensors = SafetensorRemote.get_list_tensors_hf_model(model_id)
|
||||
for name, meta in tensors.items():
|
||||
dtype, shape, offset_start, size, remote_safetensor_url = meta
|
||||
# read the tensor data
|
||||
data = SafetensorRemote.get_data_by_range(remote_safetensor_url, offset_start, size)
|
||||
print(data)
|
||||
"""
|
||||
|
||||
BASE_DOMAIN = "https://huggingface.co"
|
||||
ALIGNMENT = 8 # bytes
|
||||
|
||||
@classmethod
|
||||
def get_list_tensors_hf_model(cls, model_id: str) -> dict[str, RemoteTensor]:
|
||||
"""
|
||||
Get list of tensors from a Hugging Face model repository.
|
||||
|
||||
Returns a dictionary of tensor names and their metadata.
|
||||
Each tensor is represented as a tuple of (dtype, shape, offset_start, size, remote_safetensor_url)
|
||||
"""
|
||||
# case 1: model has only one single model.safetensor file
|
||||
is_single_file = cls.check_file_exist(
|
||||
f"{cls.BASE_DOMAIN}/{model_id}/resolve/main/model.safetensors"
|
||||
)
|
||||
if is_single_file:
|
||||
url = f"{cls.BASE_DOMAIN}/{model_id}/resolve/main/model.safetensors"
|
||||
return cls.get_list_tensors(url)
|
||||
|
||||
# case 2: model has multiple files
|
||||
index_url = (
|
||||
f"{cls.BASE_DOMAIN}/{model_id}/resolve/main/model.safetensors.index.json"
|
||||
)
|
||||
is_multiple_files = cls.check_file_exist(index_url)
|
||||
if is_multiple_files:
|
||||
# read the index file
|
||||
index_data = cls.get_data_by_range(index_url, 0)
|
||||
index_str = index_data.decode("utf-8")
|
||||
index_json = json.loads(index_str)
|
||||
assert (
|
||||
index_json.get("weight_map") is not None
|
||||
), "weight_map not found in index file"
|
||||
weight_map = index_json["weight_map"]
|
||||
# get the list of files
|
||||
all_files = list(set(weight_map.values()))
|
||||
all_files.sort() # make sure we load shard files in order
|
||||
# get the list of tensors
|
||||
tensors: dict[str, RemoteTensor] = {}
|
||||
for file in all_files:
|
||||
url = f"{cls.BASE_DOMAIN}/{model_id}/resolve/main/{file}"
|
||||
for key, val in cls.get_list_tensors(url).items():
|
||||
tensors[key] = val
|
||||
return tensors
|
||||
|
||||
raise ValueError(f"Model {model_id} does not have any safetensor files")
|
||||
|
||||
@classmethod
|
||||
def get_list_tensors(cls, url: str) -> dict[str, RemoteTensor]:
|
||||
"""
|
||||
Get list of tensors from a remote safetensor file.
|
||||
|
||||
Returns a dictionary of tensor names and their metadata.
|
||||
Each tensor is represented as a tuple of (dtype, shape, offset_start, size)
|
||||
"""
|
||||
metadata, data_start_offset = cls.get_metadata(url)
|
||||
res: dict[str, RemoteTensor] = {}
|
||||
|
||||
for name, meta in metadata.items():
|
||||
if name == "__metadata__":
|
||||
continue
|
||||
if not isinstance(meta, dict):
|
||||
raise ValueError(f"Invalid metadata for tensor '{name}': {meta}")
|
||||
try:
|
||||
dtype = meta["dtype"]
|
||||
shape = meta["shape"]
|
||||
offset_start_relative, offset_end_relative = meta["data_offsets"]
|
||||
size = offset_end_relative - offset_start_relative
|
||||
offset_start = data_start_offset + offset_start_relative
|
||||
res[name] = RemoteTensor(
|
||||
dtype=dtype,
|
||||
shape=tuple(shape),
|
||||
offset_start=offset_start,
|
||||
size=size,
|
||||
url=url,
|
||||
)
|
||||
except KeyError as e:
|
||||
raise ValueError(
|
||||
f"Missing key in metadata for tensor '{name}': {e}, meta = {meta}"
|
||||
)
|
||||
|
||||
return res
|
||||
|
||||
@classmethod
|
||||
def get_metadata(cls, url: str) -> tuple[dict, int]:
|
||||
"""
|
||||
Get JSON metadata from a remote safetensor file.
|
||||
|
||||
Returns tuple of (metadata, data_start_offset)
|
||||
"""
|
||||
# Request first 5MB of the file (hopefully enough for metadata)
|
||||
read_size = 5 * 1024 * 1024
|
||||
raw_data = cls.get_data_by_range(url, 0, read_size)
|
||||
|
||||
# Parse header
|
||||
# First 8 bytes contain the metadata length as u64 little-endian
|
||||
if len(raw_data) < 8:
|
||||
raise ValueError("Not enough data to read metadata size")
|
||||
metadata_length = int.from_bytes(raw_data[:8], byteorder="little")
|
||||
|
||||
# Calculate the data start offset
|
||||
data_start_offset = 8 + metadata_length
|
||||
alignment = SafetensorRemote.ALIGNMENT
|
||||
if data_start_offset % alignment != 0:
|
||||
data_start_offset += alignment - (data_start_offset % alignment)
|
||||
|
||||
# Check if we have enough data to read the metadata
|
||||
if len(raw_data) < 8 + metadata_length:
|
||||
raise ValueError(
|
||||
f"Could not read complete metadata. Need {8 + metadata_length} bytes, got {len(raw_data)}"
|
||||
)
|
||||
|
||||
# Extract metadata bytes and parse as JSON
|
||||
metadata_bytes = raw_data[8 : 8 + metadata_length]
|
||||
metadata_str = metadata_bytes.decode("utf-8")
|
||||
try:
|
||||
metadata = json.loads(metadata_str)
|
||||
return metadata, data_start_offset
|
||||
except json.JSONDecodeError as e:
|
||||
raise ValueError(f"Failed to parse safetensor metadata as JSON: {e}")
|
||||
|
||||
@classmethod
|
||||
def get_data_by_range(cls, url: str, start: int, size: int = -1) -> bytes:
|
||||
"""
|
||||
Get raw byte data from a remote file by range.
|
||||
If size is not specified, it will read the entire file.
|
||||
"""
|
||||
import requests
|
||||
from urllib.parse import urlparse
|
||||
|
||||
parsed_url = urlparse(url)
|
||||
if not parsed_url.scheme or not parsed_url.netloc:
|
||||
raise ValueError(f"Invalid URL: {url}")
|
||||
|
||||
headers = cls._get_request_headers()
|
||||
if size > -1:
|
||||
headers["Range"] = f"bytes={start}-{start + size}"
|
||||
response = requests.get(url, allow_redirects=True, headers=headers)
|
||||
response.raise_for_status()
|
||||
|
||||
# Get raw byte data
|
||||
return response.content[:size]
|
||||
|
||||
@classmethod
|
||||
def check_file_exist(cls, url: str) -> bool:
|
||||
"""
|
||||
Check if a file exists at the given URL.
|
||||
Returns True if the file exists, False otherwise.
|
||||
"""
|
||||
import requests
|
||||
from urllib.parse import urlparse
|
||||
|
||||
parsed_url = urlparse(url)
|
||||
if not parsed_url.scheme or not parsed_url.netloc:
|
||||
raise ValueError(f"Invalid URL: {url}")
|
||||
|
||||
try:
|
||||
headers = cls._get_request_headers()
|
||||
headers["Range"] = "bytes=0-0"
|
||||
response = requests.head(url, allow_redirects=True, headers=headers)
|
||||
# Success (2xx) or redirect (3xx)
|
||||
return 200 <= response.status_code < 400
|
||||
except requests.RequestException:
|
||||
return False
|
||||
|
||||
@classmethod
|
||||
def _get_request_headers(cls) -> dict[str, str]:
|
||||
"""Prepare common headers for requests."""
|
||||
headers = {"User-Agent": "convert_hf_to_gguf"}
|
||||
if os.environ.get("HF_TOKEN"):
|
||||
headers["Authorization"] = f"Bearer {os.environ['HF_TOKEN']}"
|
||||
return headers
|
||||
|
|
|
|||
|
|
@ -157,36 +157,8 @@ def _try_load_from_tokenizer_json(self, path: Path) -> bool:
|
|||
tokenizer = json.load(f)
|
||||
if self.load_merges:
|
||||
merges = tokenizer.get("model", {}).get("merges")
|
||||
if isinstance(merges, list) and merges:
|
||||
if isinstance(merges[0], str):
|
||||
if isinstance(merges, list) and merges and isinstance(merges[0], str):
|
||||
self.merges = merges
|
||||
elif (
|
||||
isinstance(merges[0], list)
|
||||
and len(merges[0]) == 2
|
||||
and isinstance(merges[0][0], str)
|
||||
):
|
||||
# New format since transformers 4.45 to support spaces in merges
|
||||
# ref: https://github.com/ggml-org/llama.cpp/issues/9692
|
||||
# TODO: internally store as the new format instead of converting to old
|
||||
if any(" " in s for pair in merges for s in pair):
|
||||
logger.warning(
|
||||
f'Spaces in merges detected, encoding as {chr(ord(" ") + 256)!r}'
|
||||
)
|
||||
self.merges = [
|
||||
" ".join(
|
||||
[
|
||||
# ensure the spaces are properly encoded
|
||||
"".join(
|
||||
chr(ord(c) + 256) if c == " " else c
|
||||
for c in part
|
||||
)
|
||||
for part in pair
|
||||
]
|
||||
)
|
||||
for pair in merges
|
||||
]
|
||||
else:
|
||||
raise ValueError("Unknown tokenizer merges format")
|
||||
added_tokens = tokenizer.get("added_tokens", {})
|
||||
else:
|
||||
added_tokens = {}
|
||||
|
|
@ -195,12 +167,7 @@ def _try_load_from_tokenizer_json(self, path: Path) -> bool:
|
|||
return True
|
||||
with open(tokenizer_config_file, encoding="utf-8") as f:
|
||||
tokenizer_config = json.load(f)
|
||||
chat_template_alt = None
|
||||
chat_template_file = path / "chat_template.json"
|
||||
if chat_template_file.is_file():
|
||||
with open(chat_template_file, encoding="utf-8") as f:
|
||||
chat_template_alt = json.load(f).get("chat_template")
|
||||
chat_template = tokenizer_config.get("chat_template", chat_template_alt)
|
||||
chat_template = tokenizer_config.get("chat_template")
|
||||
if chat_template is None or isinstance(chat_template, (str, list)):
|
||||
self.chat_template = chat_template
|
||||
else:
|
||||
|
|
@ -258,6 +225,7 @@ class Vocab(BaseVocab, Protocol):
|
|||
fname_tokenizer: Path
|
||||
|
||||
def __init__(self, base_path: Path): ...
|
||||
|
||||
def all_tokens(self) -> Iterable[tuple[bytes, float, gguf.TokenType]]: ...
|
||||
|
||||
|
||||
|
|
|
|||
|
|
@ -1,7 +1,7 @@
|
|||
import os
|
||||
import re
|
||||
import sys
|
||||
from typing import Any, IO, List, TextIO, Union
|
||||
from typing import Any, List, TextIO, Union
|
||||
|
||||
from PySide6.QtWidgets import (
|
||||
QMessageBox,
|
||||
|
|
@ -80,15 +80,11 @@ def load_dotenv(self=Any) -> None:
|
|||
|
||||
|
||||
def show_about(self) -> None:
|
||||
about_text = f"""AutoGGUF
|
||||
|
||||
Version: {AUTOGGUF_VERSION}
|
||||
|
||||
A tool for managing and converting GGUF models.
|
||||
This application is licensed under the Apache License 2.0.
|
||||
Copyright (c) 2024-2025 leafspark.
|
||||
It also utilizes llama.cpp, licensed under the MIT License.
|
||||
Copyright (c) 2023-2025 The ggml authors."""
|
||||
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)
|
||||
|
||||
|
||||
|
|
@ -97,7 +93,7 @@ def ensure_directory(path) -> None:
|
|||
os.makedirs(path)
|
||||
|
||||
|
||||
def open_file_safe(file_path, mode="r") -> IO[Any]:
|
||||
def open_file_safe(file_path, mode="r") -> TextIO:
|
||||
encodings = ["utf-8", "latin-1", "ascii", "utf-16"]
|
||||
for encoding in encodings:
|
||||
try:
|
||||
|
|
|
|||
|
|
@ -1,7 +1,7 @@
|
|||
import json
|
||||
|
||||
from PySide6.QtCore import Qt
|
||||
from PySide6.QtWidgets import QApplication, QFileDialog, QMessageBox
|
||||
from PySide6.QtWidgets import QFileDialog, QMessageBox
|
||||
from Localizations import (
|
||||
SAVING_PRESET,
|
||||
SAVE_PRESET,
|
||||
|
|
@ -36,40 +36,20 @@ def save_preset(self) -> None:
|
|||
"extra_arguments": self.extra_arguments.text(),
|
||||
}
|
||||
|
||||
if not QApplication.keyboardModifiers() & Qt.ShiftModifier:
|
||||
file_name, _ = QFileDialog.getSaveFileName(self, SAVE_PRESET, "", JSON_FILES)
|
||||
if file_name:
|
||||
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))
|
||||
else:
|
||||
clipboard = QApplication.clipboard()
|
||||
preset_str = json.dumps(preset, indent=1)
|
||||
clipboard.setText(preset_str)
|
||||
QMessageBox.information(self, PRESET_SAVED, "Preset copied to clipboard")
|
||||
self.logger.info("Preset copied to clipboard")
|
||||
|
||||
|
||||
def load_preset(self) -> None:
|
||||
self.logger.info(LOADING_PRESET)
|
||||
|
||||
try:
|
||||
if QApplication.keyboardModifiers() & Qt.ShiftModifier:
|
||||
clipboard = QApplication.clipboard()
|
||||
preset = json.loads(clipboard.text())
|
||||
source = "clipboard"
|
||||
else:
|
||||
file_name, _ = QFileDialog.getOpenFileName(
|
||||
self, LOAD_PRESET, "", JSON_FILES
|
||||
)
|
||||
if not file_name:
|
||||
return
|
||||
file_name, _ = QFileDialog.getOpenFileName(self, LOAD_PRESET, "", JSON_FILES)
|
||||
if file_name:
|
||||
with open(file_name, "r") as f:
|
||||
preset = json.load(f)
|
||||
source = file_name
|
||||
|
||||
self.quant_type.clearSelection()
|
||||
for quant_type in preset.get("quant_types", []):
|
||||
|
|
@ -100,19 +80,6 @@ def load_preset(self) -> None:
|
|||
self.add_kv_override(override)
|
||||
|
||||
QMessageBox.information(
|
||||
self,
|
||||
PRESET_LOADED,
|
||||
PRESET_LOADED_FROM.format(
|
||||
source
|
||||
if not QApplication.keyboardModifiers() & Qt.ShiftModifier
|
||||
else "clipboard"
|
||||
),
|
||||
self, PRESET_LOADED, PRESET_LOADED_FROM.format(file_name)
|
||||
)
|
||||
self.logger.info(PRESET_LOADED_FROM.format(source))
|
||||
|
||||
except json.JSONDecodeError:
|
||||
QMessageBox.critical(self, "Error", "Invalid JSON in clipboard")
|
||||
self.logger.error("Failed to parse JSON from clipboard")
|
||||
except Exception as e:
|
||||
QMessageBox.critical(self, "Error", f"Failed to load preset: {str(e)}")
|
||||
self.logger.error(f"Failed to load preset: {str(e)}")
|
||||
self.logger.info(PRESET_LOADED_FROM.format(file_name))
|
||||
|
|
|
|||
|
|
@ -159,9 +159,7 @@ def update_cuda_backends(self) -> None:
|
|||
for item in os.listdir(llama_bin):
|
||||
item_path = os.path.join(llama_bin, item)
|
||||
if os.path.isdir(item_path) and "cudart-llama" not in item.lower():
|
||||
if (
|
||||
"cu1" in item.lower() or "cuda-1" in item.lower()
|
||||
): # Only include CUDA-capable backends
|
||||
if "cu1" in item.lower(): # Only include CUDA-capable backends
|
||||
self.backend_combo_cuda.addItem(item, userData=item_path)
|
||||
|
||||
if self.backend_combo_cuda.count() == 0:
|
||||
|
|
|
|||
45
src/utils.py
45
src/utils.py
|
|
@ -1,10 +1,6 @@
|
|||
from typing import Any, Union
|
||||
|
||||
import urllib.request
|
||||
import urllib.error
|
||||
import json
|
||||
import ssl
|
||||
import certifi
|
||||
import requests
|
||||
from PySide6.QtCore import Qt
|
||||
from PySide6.QtWidgets import QFileDialog, QInputDialog, QMenu
|
||||
|
||||
|
|
@ -173,47 +169,18 @@ def download_llama_cpp(self) -> None:
|
|||
self.download_progress.setValue(0)
|
||||
|
||||
|
||||
def get_repo_from_env() -> tuple[str, str]:
|
||||
repo = os.getenv("AUTOGGUF_BACKEND_REPO", "ggerganov/llama.cpp")
|
||||
|
||||
if not repo or "/" not in repo:
|
||||
raise ValueError(INVALID_REPOSITORY_FORMAT)
|
||||
|
||||
owner, repo_name = repo.split("/", 1)
|
||||
if not all(part.strip() for part in (owner, repo_name)):
|
||||
raise ValueError(REPO_CANNOT_BE_EMPTY)
|
||||
|
||||
return owner, repo_name
|
||||
|
||||
|
||||
def refresh_releases(self) -> None:
|
||||
self.logger.info(REFRESHING_LLAMACPP_RELEASES)
|
||||
try:
|
||||
owner, repo = get_repo_from_env()
|
||||
url = f"https://api.github.com/repos/{owner}/{repo}/releases"
|
||||
|
||||
# Create SSL context with certifi certificates
|
||||
ssl_context = ssl.create_default_context(cafile=certifi.where())
|
||||
|
||||
# Create request
|
||||
req = urllib.request.Request(url)
|
||||
|
||||
# Make the request
|
||||
with urllib.request.urlopen(req, context=ssl_context) as response:
|
||||
if response.status != 200:
|
||||
raise urllib.error.HTTPError(
|
||||
url, response.status, "HTTP Error", response.headers, None
|
||||
response = requests.get(
|
||||
"https://api.github.com/repos/ggerganov/llama.cpp/releases"
|
||||
)
|
||||
|
||||
releases = json.loads(response.read().decode("utf-8"))
|
||||
|
||||
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 ValueError as e:
|
||||
show_error(self.logger, f"Invalid repository configuration: {str(e)}")
|
||||
except (urllib.error.URLError, urllib.error.HTTPError) as e:
|
||||
except requests.exceptions.RequestException as e:
|
||||
show_error(self.logger, ERROR_FETCHING_RELEASES.format(str(e)))
|
||||
|
|
|
|||
Loading…
Reference in New Issue