{"name":"ttf-conversion","repo":"oguzeldereli/ttf-conversion","url":"https://midorreal.com/project/oguzeldereli-ttf-conversion","repository":"https://github.com/oguzeldereli/ttf-conversion","description":"Qwen3.5 / 3.8 to ternary with TTF-trainable ternary adapters","facts":{"language":{"value":"Python","source":"https://github.com/oguzeldereli/ttf-conversion","checked":"21 Nov 2026"},"last_release":{"value":"v0.3.0, 9 Nov 2026","source":"https://github.com/oguzeldereli/ttf-conversion/releases","checked":"21 Nov 2026"},"releases":{"value":"1","source":"https://github.com/oguzeldereli/ttf-conversion/releases","checked":"21 Nov 2026"},"contributors":{"value":"1","source":"https://github.com/oguzeldereli/ttf-conversion/graphs/contributors","checked":"21 Nov 2026"},"archived":{"value":"no","source":"https://github.com/oguzeldereli/ttf-conversion","checked":"21 Nov 2026"}},"stars":14.0,"writeup":{"what_it_is":"A conversion toolkit that turns full-precision Qwen3.5 and Qwen3.8 models into ternary networks with weights in {-1, 0, +1} times scales, optionally with low-rank ternary adapters and a fine residual. It builds on ttf-core and writes checkpoints that ttf-core's trainer can load. It also includes released GGUF models and a recovery training recipe.","audience":"Researchers and developers working on ternary quantization of Qwen language models","claims":[{"kind":"specific","claim":"Converts Qwen3.5 / Qwen3.8 models into ternary networks","excerpt":"Convert a full-precision Qwen3.5 / Qwen3.8 model into a **ternary** network","status":"not_checked"},{"kind":"specific","claim":"Streams the model one decoder layer at a time from safetensors","excerpt":"It streams the model one decoder layer at a time from the safetensors","status":"not_checked"},{"kind":"specific","claim":"Released GGUF models with listed GSM8K and HumanEval scores","excerpt":"| [ttf-Qwen3.8-27B-GGUF](https://huggingface.co/oguzeldereli/ttf-Qwen3.8-27B-GGUF) | 8.2 GB | 0.895 (full precision 0.895) | 0.841 (0.933) |","status":"not_checked"},{"kind":"specific","claim":"Requires Linux with an NVIDIA GPU and Python 3.11+","excerpt":"Linux with an NVIDIA GPU (the ternary kernels are Triton), Python 3.11+, PyTorch 2.13+","status":"not_checked"},{"kind":"specific","claim":"Runs a conversion on a single 16 GB GPU with the 0.8B model","excerpt":"Qwen3.5-0.8B on one 16 GB GPU","status":"not_checked"}],"alternatives":[],"written_by":"AI, from the project's README","date":"2026-11-12"},"why_now":null,"owner_supplied":null,"score":{"verdict":"too_early","hype":82,"reality":34,"gap":47,"momentum":null,"confidence":60,"version":"score-2.0","rule":"verdict-2.0","calculated_at":"2026-11-22T08:00:00+00:00","verdict_source":"rule"},"early_read":{"estimate":true,"lean":"too_close","gap_low":14,"gap_high":56,"rule":"early-2.0"},"changes":[{"date":"2026-11-09T13:24:23+00:00","text":"Latest release v0.3.0","source":"https://github.com/oguzeldereli/ttf-conversion/releases"},{"date":"2026-10-24T22:21:52+00:00","text":"Repository created","source":"https://github.com/oguzeldereli/ttf-conversion"}],"cite":"ttf-conversion is a GitHub project at https://github.com/oguzeldereli/ttf-conversion written mainly in Python, described there as \"Qwen3.5 / 3.8 to ternary with TTF-trainable ternary adapters\".","attribution":"Data from Mid or Real, https://midorreal.com/project/oguzeldereli-ttf-conversion, checked 22 November 2026. Attribution with a link is required.","still_real":null}