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

Merge branch 'main' of https://github.com/leafspark/AutoGGUF

This commit is contained in:
BuildTools 2024-08-04 18:01:20 -07:00
parent 70f9db166d 4b1f7b5991
commit 1adad11a62
1 changed files with 14 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
README.md
View File

@ -84,9 +84,9 @@ ## Localizations:
In order to use them, please set the `AUTOGGUF_LANGUAGE` environment variable to one of the listed language codes. In order to use them, please set the `AUTOGGUF_LANGUAGE` environment variable to one of the listed language codes.
## Issues: ## Issues:
- Actual progress bar tracking - Saving preset while quantizing causes UI thread crash (planned fix: remove this feature)
- Download safetensors from HF and convert to unquanted GGUF - Cannot delete task while processing, you must cancel it first or the program crashes (planned fix: don't allow deletion before cancelling, or cancel automatically)
- Perplexity testing - Base Model text still shows when GGML is selected as LoRA type (fix: include text in show/hide Qt layout)
- ~~Cannot disable llama.cpp update check on startup~~ (fixed in v1.3.1) - ~~Cannot disable llama.cpp update check on startup~~ (fixed in v1.3.1)
- ~~`_internal` directory required, will see if I can package this into a single exe on the next release~~ (fixed in v1.3.1) - ~~`_internal` directory required, will see if I can package this into a single exe on the next release~~ (fixed in v1.3.1)
- ~~Custom command line parameters~~ (added in v1.3.0) - ~~Custom command line parameters~~ (added in v1.3.0)
@ -96,13 +96,22 @@ ## Issues:
- ~~Cannot select output/token embd type~~ (fixed in v1.1.0) - ~~Cannot select output/token embd type~~ (fixed in v1.1.0)
- ~~Importing presets with KV overrides causes UI thread crash~~ (fixed in v1.3.0) - ~~Importing presets with KV overrides causes UI thread crash~~ (fixed in v1.3.0)
## Prerelease issues: ## Planned features:
- Base Model label persists even when GGML type is selected - Actual progress bar tracking
- Download safetensors from HF and convert to unquanted GGUF
- Perplexity testing
- Managing shards (coming in the next release)
- Time estimated for quantization
- Dynamic values for KV cache, e.g. autogguf.quantized.time=str:{system.time.milliseconds} (coming in the next release)
- Ability to select and start multiple quants at once (saved in presets) (coming in the next release)
## Troubleshooting: ## Troubleshooting:
- ~~llama.cpp quantizations errors out with an iostream error: create the `quantized_models` directory (or set a directory)~~ (fixed in v1.2.1, automatically created on launch) - ~~llama.cpp quantizations errors out with an iostream error: create the `quantized_models` directory (or set a directory)~~ (fixed in v1.2.1, automatically created on launch)
- SSL module cannot be found error: Install OpenSSL or run from source `python src/main.py` using the `run.bat` script (`pip install requests`) - SSL module cannot be found error: Install OpenSSL or run from source `python src/main.py` using the `run.bat` script (`pip install requests`)
## Contributing:
Simply fork the repo and make your changes; when merging make sure to have the latest commits. Description should contain a changelog of what's new.
## User interface: ## User interface:
![image](https://github.com/user-attachments/assets/906bf9cb-38ed-4945-a32e-179acfdcc529) ![image](https://github.com/user-attachments/assets/906bf9cb-38ed-4945-a32e-179acfdcc529)