Skip to content

Releases: ademirel-taztech/Openforge

Release list

Openforge v1.0.0

Openforge v1.0.0 Latest
Latest

Choose a tag to compare

View all tags
@ademirel-taztech ademirel-taztech released this 12 Jul 09:44
v1.0.0
254f89f

🔥 Openforge v1.0.0 — first release

A cross-platform desktop GUI to convert and quantize Hugging Face (safetensors) models into ONNX and GGUF, ready to drop into your own inference stack.

Highlights

  • Pre-flight model inspection — classifies the model (decoder LLM / vision-language / encoder-embedding), computes real weight size, detects missing tokenizers, flags trust_remote_code models (never auto-enabled), and catches un-pulled Git LFS pointer files before wasting time on a doomed export.
  • ONNX export via 🤗 Optimum with FP16 / INT8 (dynamic) / INT4 (weight-only) quantization, automatic CPU-arch detection (ARM64/AVX2/AVX512/AVX512_VNNI), and post-export validation.
  • GGUF conversion via the official llama.cpp convert_hf_to_gguf.py, with F16 / Q8_0 / Q5_K_M / Q4_K_M quantization and automatic fallback to Q8_0 if llama-quantize isn't installed.
  • Vision-language model support — exports a companion mmproj-*.gguf vision projector where the architecture supports it.
  • Thread-safe, cancellable pipeline — the UI never freezes; any running conversion can be killed (full process tree) in ~1-2 seconds.
  • Plain-language error translation for common failure modes (missing sentence-transformers, un-pulled LFS files, missing tokenizer, unsupported architecture, OOM) — never a bare traceback.
  • Bilingual UI (English/Türkçe), switchable live from the top bar.
  • Dark / Light / System theme — System follows your OS automatically.
  • Completion report with elapsed time, size comparison, compression ratio, and a one-click "open output folder" button.
  • Settings persistence across launches (source/output dirs, format, quantization, execution provider, theme, language).

Requirements

  • Python 3.11.x (pinned — see README for why)
  • macOS, Linux, or Windows

Getting started

git clone https://github.com/ademirel-taztech/Openforge.git
cd Openforge
python3.11 setup_env.py
.venv/bin/python app.py

See README.md (English) or README.tr.md (Türkçe) for full installation and usage instructions, including how to unlock K-quant (Q4_K_M/Q5_K_M) support.

Assets 3
Loading