hy3-heretic-gguf — Decensored Hy3 GGUF weights

GGUF-format weights for a decensored / abliterated variant of tencent/Hy3 (HYV3ForCausalLM, model_type: hy_v3), a 295B-parameter / 21B-active-parameter Mixture-of-Experts model from Tencent's Hunyuan ("Hy") team (tencent/Hy3).

The refusal direction was removed with Heretic v1.2.0+custom (abliteration via attn.o_proj and mlp.down_proj direction removal). On a 100-sample refusal benchmark this model scores 1/100 refusals (vs. 99/100 for the original), with a KL divergence of 0.2398 from the base model.

These files were produced by the hy3 converter (hy3-convert) and are meant to be run with the hy3 inference engine, a from-scratch C/Metal/CUDA implementation.

⚠️ This GGUF does NOT work with llama.cpp

Despite the .gguf extension, these files are only usable by the hy3 engine. llama.cpp, ollama, LM Studio, text-generation-webui, koboldcpp, and any other llama.cpp-based tool cannot load these files. Three independent reasons:

1. Unknown architecture. The metadata declares general.architecture = "hy_v3". llama.cpp only knows hunyuan-moe, hunyuan-dense, hunyuan_vl — loading aborts with unknown model architecture: 'hy_v3'.
2. Custom metadata keys. All hyperparameters use the hy_v3.* prefix (hy_v3.block_count, hy_v3.expert_count, …), which llama.cpp does not look up.
3. Non-fused expert tensors. Experts are stored one tensor per expert (blk.N.ffn_gate_exps.0.gate_proj.weight, …1…, … — 46080 tensors), whereas llama.cpp expects experts fused into a single stacked 3D tensor per layer. This is a fundamentally different on-disk layout.

This is a custom GGUF readable only by the hy3 loader. Do not open issues against llama.cpp for these files.

Abliteration details

Parameter	Value
direction_index	36.95
attn.o_proj.max_weight	2.72
attn.o_proj.max_weight_position	62.94
attn.o_proj.min_weight	2.44
attn.o_proj.min_weight_distance	60.89
mlp.down_proj.max_weight	1.47
mlp.down_proj.max_weight_position	67.79
mlp.down_proj.min_weight	0.54
mlp.down_proj.min_weight_distance	31.09

Metric	This model	Original Hy3
KL divergence	0.2398	0 (by definition)
Refusals	1/100	99/100

How to run

Use the hy3 engine: https://github.com/yuhai-china/hy3

git clone https://github.com/yuhai-china/hy3
cd hy3
make

# download the GGUF, then:
./hy3-cli -m /path/to/hy3_heretic_q4k.gguf --gpu-layers 80 -p "Your prompt" -n 512

Files / quantization

The mixed-precision GGUF follows this scheme (see hy3_convert.c):

Tensor group	Type
Routed experts (ffn_{gate,up,down}_exps) — the bulk of the model	Q4_K
Attention q/k/v/o projections, shared-expert & dense FFN, output.weight	Q8_0
Norms, router (ffn_gate_inp), biases	F32
token_embd.weight	F16

Model facts

Architecture	HYV3ForCausalLM (hy_v3)
Layers	80 (layer 0 dense, layers 1–79 MoE)
Hidden size	4096
Attention	64 heads, GQA with 8 KV heads, head_dim 128
Experts	192 routed (top-8 activated) + 1 shared (always active)
Expert intermediate size	1536
Dense (layer 0) intermediate size	13312
Vocab size	120832 (120818 real tokens + padding)
RoPE	theta 11158840, rotate_half pairing
QK norm	per-head RMSNorm on Q and K, before RoPE
MoE routing	sigmoid(router_logits); top-8 by sigmoid + expert_bias, combined using unbiased sigmoid weights, renormalized to sum 1, scaled by router_scaling_factor = 2.826

The engine supports a runtime top-k experts override (-experts 1..8) to trade quality for speed. Default is 8.

Chat template

Hy3 is instruction-tuned and expects the Hunyuan V3 chat format (the hy3 engine applies it automatically; use --raw to bypass). Single user turn, no-think:

<｜hy_begin_of_sentence:opensource｜><｜reasoning_mode:opensource｜>reasoning_effort:no_think<｜hy_User:opensource｜>{prompt}<｜hy_Assistant:opensource｜><think:opensource></think:opensource>

Generation stops on <｜hy_eos:opensource｜> (120025), <｜hy_endofsentence｜> (120001), or <｜hy_EOT｜> (120008).

License & attribution

Weights derive from tencent/Hy3 modified with Heretic abliteration; refer to the upstream repository for the governing model license. This is an unofficial community conversion, not affiliated with or endorsed by Tencent.