Downloads · 30 days
30
0% of all-time downloads
groxaxo/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-W8A8
Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-W8A8 is a machine learning model from groxaxo. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-W8A8 is an 8-bit checkpoint designed to reduce inference memory requirements, published by groxaxo. It is intended for open-source evaluation, reproducible experi…
Downloads · 30 days
30
0% of all-time downloads
All-time downloads
29.4K
Public
Parameters
27.4B
31.2 GB on disk
Likes
1
Public
Click a slice to open those files.
.safetensors31.2 GB · 100%
How the weights are stored.
I824.4B · 88%
From the Hugging Face model README
Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-W8A8 is an 8-bit checkpoint designed to reduce inference memory requirements, published by groxaxo.
It is intended for open-source evaluation, reproducible experimentation, and compatible local or
hosted inference workflows. The wording below is deliberately limited to what can be verified
from this repository's metadata and artifacts.
The repository name identifies a behavior-modified or reduced-filtering lineage. That label describes the source or conversion history; it is not a guarantee of unrestricted behavior in every prompt or runtime. Test outputs carefully before sharing or deploying them.
| Field | Details |
|---|---|
| Format | INT8 |
| Source / base | llmfan46/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved |
| Intended task | image-text-to-text |
| License | apache-2.0 |
*.safetensors (11 files)config.jsongeneration_config.jsontokenizer.jsontokenizer_config.jsonprocessor_config.jsonchat_template.jinjaquantization_config.jsonLoad the checkpoint with a runtime that supports its architecture and 8-bit weight format. Keep the repository's configuration and tokenizer/processor files together with the weight files.
Quantization or conversion changes numerical behavior, memory use, and throughput relative to the source checkpoint; validate quality on your own workload.
Generated outputs may be inaccurate or unsuitable for a given use case. Users are responsible for testing behavior, applying appropriate safeguards, and complying with applicable licenses and laws.
<!-- polished-overview:end -->INT8 W8A8 dynamic (per-token activation) quantization of
llmfan46/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved,
produced with AutoRound 0.13.1 and exported in the
compressed-tensors (auto_round:llm_compressor) format for vLLM.
Same recipe as the Darwin-28B W8A8 build.
INT8 — channel-wise INT8 weights, INT8 dynamic per-token activations (CUTLASS INT8 path in vLLM).Linear weights only (DeltaNet in_proj_*/out_proj + full-attn q/k/v/o_proj + MLP gate/up/down).mtp.*, via --ignore_layers mtp), the vision tower (model.visual.*, auto-skipped by AutoRound's text-module-only VLM path), lm_head, embed_tokens, all norms.Architecture: Qwen3.5-generation hybrid (qwen3_5, 64 layers, 3:1 DeltaNet/full-attention, hidden 5120, vocab 248320), multimodal, with a 1-layer native MTP head for speculative decoding.
vllm serve groxaxo/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved-W8A8 \
--tensor-parallel-size 2 \
--max-model-len 262144 \
--kv-cache-dtype fp8 \
--reasoning-parser qwen3 \
--trust-remote-code
The native MTP head is preserved in BF16, but on this finetune it is non-predictive: measured ~0% draft-token acceptance (2 of 6254), so enabling speculative decoding is actually slower (~22.5 tok/s with MTP vs ~27.7 tok/s dense, single-stream on 2×RTX 3090). This was isolated to the checkpoint's MTP head, not the quantization or server:
method:mtp on the standard Qwen3.6-27B (AEON FP8) gives 78% acceptance / ~2.1× speedup — so the setup is correct.kv-cache-dtype=auto (not a KV-quant artifact).The heretic-v2 finetune appears to have diverged its main weights from the base MTP head, so the head's
drafts no longer match. Serve dense for best speed. The MTP weights remain in this repo for anyone
who retrains/distills the head against this finetune. The vision tower is intact (drop
--hf-overrides '{"language_model_only": true}' to use it; add it to save VRAM for text-only).