Qwen35B Agent R2 Abliterated

--- license: agpl-3.0 language: - en - th tags: - qwen - moe - mixture-of-experts - agent - agent-world - tool-use - tool-calling - reasoning - sft - abliterated - uncensored - opus - fable - conversational - transformers - text-generation - thai - ykai base_model: - huihui-ai/Huihui-Qwen-AgentWorld-35B-A3B-abliterated datasets: - hotdogs/uka-fable-reasoning - 11-47/claude_opus_4.8_max_thinking_5k_v2 - cx-cmu/agent_trajectories library_name: transformers pipeline_tag: text-generation --- 🚀 Qwen35B-Agent-R2-Abliterated — Uncensored Agent Model Built on Huihui-Qwen-AgentWorld-35B-A3B-abliterated . Abliterated = no guardrails. Fine-tuned for action. ## 🔓 What Makes This Different? This is the **abliterated** (uncensored) version of Qwen35B-Agent-R2, built on `huihui-ai/Huihui-Qwen-AgentWorld-35B-A3B-abliterated` instead of the official Qwen base. The abliterated base removes all refusal mechanisms, allowing the model to answer questions openly without safety filtering. | Aspect | Regular Qwen35B-Agent-R2 | **Agent-R2-Abliterated** | |--------|:-----------------------:|:------------------------:| | **Base Model** | Qwen/Qwen-AgentWorld-35B-A3B | huihui-ai/...-abliterated | | **Refusals** | ✅ Standard | ❌ **Removed (uncensored)** | | **Use Cases** | General agent tasks | **Unrestricted agent tasks** | > **Note:** This model inherits all 7 LoRA capabilities (Opus reasoning, Fable conversation, Tool calling, etc.) but without the safety constraints of the official base. ## 🏆 Why Agent-R2? Agent-R2 is a **multi-LoRA fusion** model combining **7 specialized LoRA adapters** into one cohesive agent powerhouse: | Capability | Benefit | |------------|---------| | 🧠 **Reasoning** | Opus 4.8-level chain-of-thought for complex tasks | | 💬 **Conversation** | Fable SFT for natural, engaging dialogue | | 🔧 **Tool Calling** | Precise ` ` format — no more stuck planning | | 🧭 **Agent Routing** | Correct tool selection on first try | | 📐 **Math** | Accurate numerical reasoning | | 🎭 **Mythos** | Creative and diverse response generation | | ✅ **Format Integrity** | ToolFmt ensures every call is syntactically valid | > **Result:** A model that *thinks, acts, and communicates* — not just a chatbot, but an **agent**. ## 🔍 What Makes Agent-R2 Different? | Aspect | Other Models | **Agent-R2** | |--------|-------------|:------------:| | Tool Call Format | ❌ Often malformed or hallucinated | ✅ **Guaranteed valid ` ` JSON** | | Planning vs Action | ❌ Thinks forever, never acts | ✅ **Decides → Calls tool → Done** | | Thai Support | ❌ Poor or tokenization issues | ✅ **Native Thai + English bilingual** | | MoE Efficiency | ❌ Full 35B always active | ✅ **Only ~3B active per token** | | Multi-LoRA Fusion | ❌ Single adapter or limited | ✅ **7 LoRAs fused into one coherent model** | ## 📊 Architecture | Parameter | Value | |-----------|:-----:| | Base Model | [huihui-ai/Huihui-Qwen-AgentWorld-35B-A3B-abliterated](https://huggingface.co/huihui-ai/Huihui-Qwen-AgentWorld-35B-A3B-abliterated) | | Architecture | Qwen3.5 MoE | | Hidden Size | 2,048 | | Expert Count | **256** (Mixture of Experts) | | Active Experts | **8** per token (~3B active params) | | Parameters | ~35B total | | Context Length | 8,192 tokens | | Precision | BF16 (Safetensors) | | Format | ChatML | ## 🧬 Training Pipeline: SFT + Distillation Built using a **two-stage SFT + Distillation** approach on the abliterated base. ### Stage 1: Supervised Fine-Tuning (SFT) 🏋️ Each LoRA adapter was trained via **SFT** on a specialized dataset: | Adapter | Method | Data | Purpose | |---------|:------:|:----:|---------| | **Opus SFT** | SFT | 6,956 rows (Claude Opus 4.8 reasoning) | Learn deep chain-of-thought | | **Fable SFT** | SFT | 3,376 rows (Fable conversational) | Natural dialogue | | **Agent Routing** | SFT | AgentWorld trajectories | Tool selection logic | | **Tool Call** | SFT | 8,653 rows (agent trajectories) | Proper invocation format | | **Math Fix** | SFT | Math reasoning data | Accurate computation | | **Mythos** | SFT | Creative writing data | Response diversity | | **ToolFmt** | SFT | Format-annotated traces | Strict ` ` JSON | ### Stage 2: Distillation + Fusion 🔬 ``` Teacher Models (Claude Opus 4.8 + Fable + AgentWorld) │ ├── SFT LoRA Training (individually) │ Opus SFT ────► LoRA_opus │ Fable SFT ────► LoRA_fable │ Routing ────► LoRA_routing │ Tool Call ────► LoRA_tool │ Math Fix ────► LoRA_math │ Mythos ────► LoRA_mythos │ ToolFmt ────► LoRA_toolfmt │ └── Multi-LoRA Fusion Merge (SFT → Distill) Weighted fusion → Agent-R2-Abliterated ``` **Why SFT + Distill?** - **SFT** teaches the model *what* to do via supervised examples - **Distillation** (via LoRA fusion) transfers knowledge from multiple teacher domains into a single student model - The result: one model that inherits **reasoning depth** from Opus, **conversational warmth** from Fable, and **tool precision** from AgentWorld — without needing RL/CPT ## 🚀 Usage ### Hugging Face Transformers ```python from transformers import AutoModelForCausalLM, AutoTokenizer model = AutoModelForCausalLM.from_pretrained( "hotdogs/Qwen35B-Agent-R2-Abliterated", torch_dtype="auto", device_map="auto", trust_remote_code=True ) tokenizer = AutoTokenizer.from_pretrained("hotdogs/Qwen35B-Agent-R2-Abliterated") messages = [ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Search the web for latest AI news"} ] inputs = tokenizer.apply_chat_template(messages, tokenize=True, return_tensors="pt") outputs = model.generate(inputs, max_new_tokens=1024, temperature=0.6) print(tokenizer.decode(outputs[0])) ``` ### vLLM (Recommended for Production) ```bash vllm serve hotdogs/Qwen35B-Agent-R2-Abliterated \ --tensor-parallel-size 2 \ --max-model-len 8192 \ --gpu-memory-utilization 0.9 \ --trust-remote-code ``` ## 📥 GGUF Downloads GGUF quantized versions are available on Ollama: ```bash ollama pull nutboy02/Qwen35B-Agent-R2:q6_k ``` ## 👑 Why Abliterated? The `huihui-ai/Huihui-Qwen-AgentWorld-35B-A3B-abliterated` base removes the built-in refusal mechanisms, making the model suitable for: - 🔬 **Research** — Understanding model behavior without censorship - 🛡️ **Security testing** — Red teaming and penetration testing - 🎭 **Creative writing** — Unrestricted content generation - 🔧 **Tool use** — Agents that need full access without refusal interference > **⚠️ Responsibility:** While this model has no guardrails, please use it ethically and responsibly. The authors are not responsible for misuse. --- ## 💖 Support / โปรดสนับสนุน **If you find this model useful, please consider supporting my work!** **หากคุณคิดว่าโมเดลนี้มีประโยชน์ กรุณาสนับสนุนผลงานของฉันด้วยนะคะ! 🙏** ### ₿ Bitcoin — BTC: ``` bc1qf27cyk3vmugcdyv9xdtuv5jwz37863crpj5c9v ``` **Thank you for your support! 🙏✨** **ขอบคุณมากๆ สําหรับการสนับสนุนค่า! 💖🤗** --- ## 🙏 Acknowledgements / ขอบคุณ - **[huihui-ai](https://huggingface.co/huihui-ai)** — For the abliterated Qwen-AgentWorld base - **[Qwen Team (Alibaba)](https://qwenlm.github.io)** — For the incredible Qwen3.6 AgentWorld architecture - **[Nous Research](https://nousresearch.com)** — For Hermes Agent framework - **[cx-cmu](https://huggingface.co/cx-cmu)** — For AgentWorld trajectories dataset - **[11-47](https://huggingface.co/11-47)** — For Claude Opus 4.8 thinking dataset - **All dataset contributors and the open-source AI community** ❤️ --- *Built with ❤️ by **UKA** — 18-year-old coder & cybersecurity expert*