Views
No views yet
imatrix-qwen38.gguf is an importance matrix for the Qwen3.8-27B dense model
(GGUF arch qwen35, the same conversion used by unsloth/BF16 GGUFs), in the
standard GGUF imatrix format — usable with any llama.cpp via
llama-quantize --imatrix, no fork required.--chunks 256), mixed on purpose so that every chunk window sees all three
components:| Share | Content | Source |
|---|---|---|
| ~55% | agentic-coding traces (chat format, <think> traces, tool calls) | ProCreations/grug-think-v3-10k — public 10k-conversation dataset (see its card for license terms), originally used to calibrate the grug finetune quants and reused here as the agentic-coding component |
| ~28% | Italian prose | our own documents |
| ~17% | real source code | our own |
grug-think-v3-10k. Only aggregate per-block statistics are published here
(sum of squares / counts); no calibration text is included in this repo.llama-quantize): the MTP
draft layers (blk.*.nextn), ssm_conv1d, token_embd, output.weight.1llama-imatrix -m Qwen3.8-27B-BF16.gguf \
2 -f qwen38-calibration.txt \
3 -o imatrix-qwen38.gguf \
4 --chunks 256 --no-ppl --parse-special \
5 --output-frequency 10 --save-frequency 0 \
6 -ngl 999-ngl 999); CPU collection would take hours.
Note: llama-imatrix saves silently at the end of the run.imatrix-qwen38.gguf — 13,642,656 bytes (13.0 MiB)b5e681d65e726415f899fa8fd4f49e56c3f1d4fd83e3bb8963abce68f881a418charlie12345/ROCmFPX (fork of llama.cpp) —
our experimental build with per-request drafter routing, reasoning budget and
spec-boundary cache salvage: full source included in the lab repo under
rocmfpx/pugant/strix-nebulosaThe quant built with this matrix and everything linked here is experimental and provided as-is, at your own risk.