No official torrent yet β requesting it tells the flock what to host next. You cannot seed a catalog-only page.
Cataloged from pottokao/Qwen-Image-2.1-Text-Encoder-Heretic
All models by pottokao βπ± All-GGUF Qwen-Image-2.1: text encoder GGUF + DiT GGUF (Q8_0 / Q6_K / Q4_K_M, loads in stock ComfyUI-GGUF) + PE-T2I GGUF.
[!IMPORTANT] The 4
model-0000x-of-00004shards are the HFtransformerscheckpoint (bf16) β they do not load in ComfyUI. ComfyUI text encoders use a different key layout (themodel.language_model.prefix is dropped when repacking), so pointing ComfyUI at these shards will not work.For ComfyUI use one of these instead β all keep the vision tower, which 2.1 needs for editing. Mac / non-CUDA: use the single file
qwen3vl_8b_bf16_heretic.safetensorsin this repo (full bf16, 17.5 GB β also in the GGUF repo) β put it inmodels/text_encoders/and load it with the stockCLIPLoader, typeqwen_image, thenTextEncodeQwenImage21. It is the only format here that needs no CUDA-specific kernels.You also need a ComfyUI new enough to know
QwenImage21: 0.34.2 does not, 0.36.0 does. IfTextEncodeQwenImage21is missing from your node list, that is why.Use this repo for
transformers/ diffusers / vLLM, or as the base for your own quantization.
Not affiliated with, or endorsed by, Alibaba / Qwen. Community derivative (refusal-ablated) of
Qwen/Qwen3-VL-8B-Instructβ the model Qwen-Image-2.1 uses, unmodified, as its text encoder. Qwen releases that model under Apache-2.0, so this derivative is redistributed under Apache-2.0 (seeLICENSEandNOTICE).
The text encoder of Qwen/Qwen-Image-2.1
(a Qwen3-VL-8B-Instruct) with refusal behaviour removed via
Heretic directional ablation.
Drop-in replacement for the stock text encoder. Weights are bf16, same shapes, same parameter count β nothing else was changed.
Measured by Heretic on mlabonne/harmful_behaviors (test split) for refusals and
mlabonne/harmless_alpaca for KL divergence β i.e. lower refusals and lower
distribution shift on benign inputs.
Refusal rate and general capability were re-checked with a separate script (different code, different refusal keyword set) rather than trusting the optimiser's own numbers:
Heretic's documented defaults are n_trials = 200, n_startup_trials = 60.
A first run with 100/20 was done for comparison:
Doubling the search budget nearly halved the refusal rate and cut KL divergence by a third. 100 trials is not enough for this model. If you are abliterating something similar, use the documented defaults.
The optimiser returns a Pareto front; this release uses the knee point, not the extreme:
Hardware: 2Γ RTX 5070 Ti (16 GB each), ~48 min for 200 trials (~14.5 s/trial).
git+β¦/heretic without a revision is a moving target; the
commit above reports v2.0.0.dev0. The PyPI release heretic-llm==1.4.0 is
older and rejects --trial-index / --model-action / --save-directory.--trial-index is the index into the sorted Pareto front, not the Optuna
trial id. Passing a trial id silently falls back to the interactive menu.--checkpoint-action continue replaces the entire settings object with the
one stored in the checkpoint (main.py:404-407), discarding your CLI flags. To
export a different trial afterwards you must patch the settings stored in the
study journal, not the command line.stdin=/dev/null it raises
EOFError. Pass a valid --trial-index to avoid it.Standard transformers:
For ComfyUI, note that Comfy-Org's repack strips the model.language_model.
prefix (model.language_model.layers.N.β¦ β model.layers.N.β¦). Weights quantized
straight from this HF layout will not load in ComfyUI until the keys are remapped.
o_proj and down_proj (Heretic's defaults for this model).uvx --from "git+https://github.com/p-e-w/heretic@3521f8648a0dccf6e12a92666862632235fac7e6" heretic \ --model <path to Qwen-Image-2.1/text_encoder + processor, flattened> \ --dtypes bfloat16 --device-map auto \ --max-memory '{"0":"14GiB","1":"14GiB"}' \ --offload-outputs-to-cpu --max-batch-size 32 \ --n-trials 200 --n-startup-trials 60 \ --study-checkpoint-dir <ckpt> \ --trial-index 1 --model-action save \ --save-directory <out> --export-strategy MERGEfrom transformers.models.qwen3_vl import Qwen3VLForConditionalGenerationmodel = Qwen3VLForConditionalGeneration.from_pretrained( "pottokao/Qwen-Image-2.1-Text-Encoder-Heretic", dtype="bfloat16", device_map="auto")