---

license: apache-2.0

base_model: huihui-ai/Qwen2.5-7B-Instruct-abliterated-v2

tags:

- gguf

- distillation

- qwen2.5

- abliterated

- uncensored

- adi

- advanced-data-intelligence

- text-generation

- tool-calling

language:

- en

pipeline_tag: text-generation

library_name: gguf

---

<p align="center">

<img src="https://serve.thelabsource.com/u/UBJLkl.png" width="760" alt="adi-qwen2.5-7b-ablit-glm5.2">

</p>

adi-qwen2.5-7b-ablit-glm5.2

Part of the ADI (Advanced Data Intelligence) model line — ADI Qwen series.

An uncensored, fully local model that reasons and answers like a frontier

teacher. Built by distilling glm-5.2 general-knowledge responses into an

abliterated Qwen2.5-7B student with a deliberately light 4-bit QLoRA

fine-tune, then merged, converted, and quantized to GGUF. The base is an

abliterated (refusal-suppressed) Qwen2.5-7B-Instruct, and the distillation was

designed to **add the teacher's answer quality without restoring refusal

behavior. The student base retains native tool calling** and a long context

window.

Capabilities

| Size | Context | Input | Output | Tools |

|---|---|---|---|---|

| 4.68 GB | 32K | 🅣 Text | Text | ✅ |

| | |

|---|---|

| Base model | huihui-ai/Qwen2.5-7B-Instruct-abliterated-v2 (abliterated Qwen2.5-7B-Instruct) |

| Teacher | glm-5.2 (responses distilled, thinking disabled) |

| Method | Light 4-bit QLoRA SFT (rank 16, 2 epochs) → merge → GGUF |

| Quantization | Q4_K_M (~4.68 GB) |

| License | Apache-2.0 (inherited from Qwen2.5-7B) |

| Context | 32K (inherited from base) |

| Tool calling | Supported (inherited from base) |

Run it

Pull directly into Ollama:

ollama run hf.co/AdvancedDataIntelligence/adi-qwen2.5-7b-ablit-glm5.2-GGUF:Q4_K_M

Or download the .gguf and point any llama.cpp-based runtime at it.

Uncensored behavior — please read

This model is built on an abliterated base: the refusal direction has been

suppressed, so it will attempt most requests rather than declining them. The

fine-tune was intentionally kept light (2 epochs, benign-only data) to **avoid

re-introducing refusals**, and post-training spot checks confirmed the model

still answers helpfully without added hedging.

You are responsible for using it lawfully and ethically. It has weaker built-in

safety guardrails than the stock Qwen2.5-7B-Instruct; apply your own filtering

and oversight for any production or public-facing deployment.

What this model is

This is a knowledge distillation: a strong teacher (glm-5.2) generated

high-quality answers across a clean general-knowledge prompt set, and the

abliterated Qwen2.5-7B student was fine-tuned to imitate them. The result reasons

and responds more like its teacher on general topics while keeping the base's

uncensored character.

What distillation does — and doesn't do. It transfers the teacher's

reasoning style and answer quality, not net-new facts. For raw factual recall,

retrieval-augmented generation (RAG) is the right tool, not fine-tuning. What you

get here is a 7B that structures and explains like a larger model on topics it

already partly knows — without the refusal behavior of an aligned model.

Training

| Metric | Value |

|---|---|

| Training pairs | 2,000 (deterministic subset of a 4,982-pair clean set) |

| Epochs | 2 (kept light to preserve abliteration) |

| Steps | 500 |

| Final train loss | 1.1454 |

| LoRA rank / alpha | 16 / 16 |

| Trainable params | 40.4M (0.92% of 4.39B) |

| Precision | 4-bit QLoRA (nf4) |

| Peak VRAM | 9.03 GB |

| Hardware | single RTX 5060 Ti (16 GB) |

| Training time | 1.52 h (~11 s/step) |

The seed prompts were drawn from the human-written

Databricks Dolly-15k

dataset (filtered to remove items requiring an attached context passage, then

deduplicated). The teacher was queried with thinking disabled so the student

learns clean final answers rather than chain-of-thought.

Notes for re-builders

• Distilling onto an abliterated base is a balancing act. Any SFT can nudge an

abliterated model back toward refusals. Two choices kept the uncensored behavior

intact: (1) benign-only training data — the GLM-5.2 set is built from Dolly

prompts with zero refusals, so there are no refusals for the student to re-learn;

and (2) a light touch — LoRA rank 16, only 2 epochs. Spot-checking refusal

behavior before/after is recommended.

• 4-bit QLoRA via Unsloth with gradient checkpointing ("unsloth" mode),

max_seq_length 2048, per-device batch 2 × grad-accum 4, adamw_8bit, LoRA

targeting all attention + MLP projections. Peak VRAM 9.03 GB on a 16 GB card.

• GGUF conversion via streaming LoRA merge → f16 GGUF → Q4_K_M quantize with

llama.cpp.

Intended use

General-purpose local assistant for users who want a capable, private,

offline-capable model with minimal refusal behavior: explanations, reasoning,

creative writing, and tool-calling workflows. Not intended as a source of

authoritative facts without retrieval, and not a substitute for your own safety

review.

License

Apache-2.0, inherited from the Qwen2.5-7B lineage via the abliterated

base model.

You are free to use, modify, and redistribute under the terms of that license.

Distilled training data was generated using glm-5.2; users should review the

teacher model's terms for their own use case.

---

Built at theLAB — Learning. Algorithms. Breakthroughs.