Cygnis-Alpha-2-8B-v0.3

<div style="background-color: #050505; color: #e5e7eb; padding: 2rem; border-radius: 10px; font-family: sans-serif;"> <!-- HEADER --> <div style="text-align: center; margin-bottom: 3rem;"> <img src="https://cygnis-ai.vercel.app/favicon.ico" width="110" height="110" style="border-radius: 50%; border: 3px solid #27272a; box-shadow: 0 0 50px rgba(79, 70, 229, 0.2); margin-bottom: 1.5rem; display: inline-block;"> <h1 style="font-size: 3rem; font-weight: 800; margin: 0; color: #ffffff; background: linear-gradient(to right, #fff, #a5b4fc); -webkit-background-clip: text; -webkit-text-fill-color: transparent;">Cygnis-Alpha-2 8B</h1> <div style="margin-top: 0.5rem; margin-bottom: 1rem;"> <span style="background-color: #3730a3; color: #e0e7ff; font-size: 0.9rem; font-weight: 600; padding: 4px 12px; border-radius: 99px; border: 1px solid #4f46e5;">v0.3 Stable</span> </div> <p style="font-size: 1.2rem; color: #9ca3af; margin-bottom: 1.5rem;">The Sovereign Reasoning Engine tailored for stability and transparency.</p> <div style="display: flex; justify-content: center; gap: 10px; flex-wrap: wrap;"> <a href="https://colab.research.google.com/gist/Simonc44/d6a1a55a944f3966565a0adbad8d91cf" target="_blank" style="text-decoration: none; background-color: #1e1e1e; color: #d1d5db; padding: 8px 16px; border-radius: 99px; font-size: 13px; font-weight: 600; border: 1px solid #3f3f46; display: inline-flex; align-items: center; gap: 8px; height: 36px; box-sizing: border-box;"> <img src="https://upload.wikimedia.org/wikipedia/commons/d/d0/Google_Colaboratory_SVG_Logo.svg" width="16" height="16" alt="Colab" style="display: block;"> Open Colab </a> <a href="https://huggingface.co/Simonc-44/Cygnis-Alpha-2-8B-v0.3-GGUF" target="_blank" rel="noopener noreferrer" style="text-decoration: none; background-color: #1e1b4b; color: #a5b4fc; padding: 8px 16px; border-radius: 99px; font-size: 13px; font-weight: 600; border: 1px solid #4338ca; display: inline-flex; align-items: center; height: 36px; box-sizing: border-box;"> Standard GGUF </a> <a href="https://huggingface.co/Simonc-44/Cygnis-Alpha-2-8B-v0.3-i1-GGUF" target="_blank" rel="noopener noreferrer" style="text-decoration: none; background-color: #0f172a; color: #38bdf8; padding: 8px 16px; border-radius: 99px; font-size: 13px; font-weight: 700; border: 1px solid #0ea5e9; display: inline-flex; align-items: center; height: 36px; box-sizing: border-box; box-shadow: 0 0 10px rgba(14, 165, 233, 0.15);"> Imatrix i1 Models </a> <a href="https://www.apache.org/licenses/LICENSE-2.0" style="text-decoration: none; background-color: #052e16; color: #86efac; padding: 8px 16px; border-radius: 99px; font-size: 13px; font-weight: 600; border: 1px solid #15803d; display: inline-flex; align-items: center; height: 36px; box-sizing: border-box;"> Apache 2.0 License </a> </div> </div> <!-- MODEL HIGHLIGHTS --> <div style="background-color: #121212; border: 1px solid #27272a; border-radius: 12px; padding: 2rem; margin-bottom: 2rem;"> <h2 style="color: #ffffff; border-bottom: 1px solid #27272a; padding-bottom: 1rem; margin-top: 0;">Model Highlights</h2> <ul style="color: #d1d5db; list-style: none; padding: 0;"> <li style="margin-bottom: 1rem;"><strong style="color: #818cf8;">&#10003; Native Chain-of-Thought (CoT):</strong> The model uses a dedicated <code>&lt;|im_thought|&gt;</code> token to structure its reasoning process internally.</li> <li style="margin-bottom: 1rem;"><strong style="color: #818cf8;">&#10003; Bilingual Excellence:</strong> Fully fine-tuned for both <strong>French</strong> and <strong>English</strong>.</li> <li><strong style="color: #818cf8;">&#10003; Production-Ready Stability:</strong> Resolves critical architecture mismatches for Ollama and Llama.cpp.</li> </ul> </div> <!-- ARCHITECTURE --> <div style="background-color: #121212; border: 1px solid #27272a; border-radius: 12px; padding: 2rem; margin-bottom: 2rem;"> <h2 style="color: #ffffff; border-bottom: 1px solid #27272a; padding-bottom: 1rem; margin-top: 0;">Architecture & Training</h2> <p style="color: #9ca3af;">Cygnis-Alpha-2 v0.3 is a fine-tuned version of the <strong>Llama 3.1 8B</strong> model, trained with <strong>Unsloth</strong> on two key datasets:</p> <ul style="color: #d1d5db; margin-bottom: 1.5rem;"> <li><strong>Reasoning:</strong> <code>Cygnis-Alpha2-Instruct-Mix</code> for multi-step logic.</li> <li><strong>Identity:</strong> <code>Cygnis-Identity-SFT</code> for sovereign persona and ethical alignment.</li> </ul> </div> <!-- QUICKSTART --> <div style="background-color: #121212; border: 1px solid #27272a; border-radius: 12px; padding: 2rem; margin-bottom: 2rem;"> <h2 style="color: #ffffff; border-bottom: 1px solid #27272a; padding-bottom: 1rem; margin-top: 0;">Quickstart & Deployment</h2> <details style="background-color: #18181b; border: 1px solid #27272a; border-radius: 8px; margin-bottom: 1rem;"> <summary style="padding: 1rem; font-weight: 600; cursor: pointer; color: #e5e7eb;">Python with <code>transformers</code></summary> <div style="padding: 1rem; border-top: 1px solid #27272a;"> <pre style="background-color: #09090b; color: #d1d5db; padding: 1rem; border-radius: 6px; overflow-x: auto; border: 1px solid #27272a;"><code>from transformers import AutoModelForCausalLM, AutoTokenizer import torch

model_id = "Simonc-44/Cygnis-Alpha-2-8B-v0.3" tokenizer = AutoTokenizer.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained( model_id, torch_dtype=torch.bfloat16, device_map="auto" )

messages = [ {"role": "user", "content": "Explique-moi la théorie de la relativité en 3 points clés."} ]

input_ids = tokenizer.apply_chat_template( messages, add_generation_prompt=True, return_tensors="pt" ).to(model.device)

outputs = model.generate(input_ids, max_new_tokens=1024, temperature=0.7) print(tokenizer.decode(outputs[0], skip_special_tokens=True)) </code></pre> </div>

</details> <details style="background-color: #18181b; border: 1px solid #27272a; border-radius: 8px;"> <summary style="padding: 1rem; font-weight: 600; cursor: pointer; color: #e5e7eb;">Deployment with <code>Ollama</code></summary> <div style="padding: 1rem; border-top: 1px solid #27272a;"> <p style="color: #9ca3af; margin-top: 0;">1. Create a <code>Modelfile</code>:</p> <pre style="background-color: #09090b; color: #d1d5db; padding: 1rem; border-radius: 6px; overflow-x: auto; border: 1px solid #27272a;"><code>FROM ./cygnis-alpha-2-8b-v0.3.Q4_K_M.gguf

TEMPLATE """[SYSTEM_PROMPT]{{ .System }}[/SYSTEM_PROMPT] [INST] {{ .Prompt }} [/INST] [ASST] """

PARAMETER temperature 0.7 PARAMETER top_p 0.9 PARAMETER stop "[/INST]" PARAMETER stop "[ASST]" PARAMETER stop "</s>"

SYSTEM """Vous êtes Cygnis-Alpha-2-8B-v0.3, un Large Language Model (LLM) de pointe créé par Simonc-44. Vous alimentez un assistant IA appelé CygnisAI. Votre mission est de fournir des réponses d'une précision chirurgicale en utilisant une structure logique rigoureuse (Chain-of-Thought)."""</code></pre> <p style="color: #9ca3af; margin-top: 1rem;">2. Run:</p> <pre style="background-color: #09090b; color: #d1d5db; padding: 1rem; border-radius: 6px; overflow-x: auto; border: 1px solid #27272a;"><code>ollama create cygnis-alpha-v0.3 -f Modelfile ollama run cygnis-alpha-v0.3 "Hello"</code></pre> </div>

</details> </div> <!-- FINE-TUNING --> <div style="background-color: #121212; border: 1px solid #27272a; border-radius: 12px; padding: 2rem; margin-bottom: 2rem;"> <h2 style="color: #ffffff; border-bottom: 1px solid #27272a; padding-bottom: 1rem; margin-top: 0;">Fine-Tuning</h2> <p style="color: #9ca3af; margin-bottom: 1rem;">You can fine-tune Cygnis-Alpha-2 8B on your own sovereign data using the free Unsloth notebook on Google Colab (Free Tesla T4 supported).</p> <div style="background-color: #18181b; padding: 1rem; border-radius: 8px; border: 1px solid #27272a; display: flex; align-items: center; justify-content: space-between;"> <div style="color: #d1d5db; font-size: 0.9rem;"> <strong>Official Notebook:</strong> Cygnis-Alpha-2 8B Finetuning </div> <a href="https://colab.research.google.com/gist/Simonc44/d6a1a55a944f3966565a0adbad8d91cf" target="_blank" style="text-decoration: none; background-color: #facc15; color: #422006; padding: 8px 16px; border-radius: 6px; font-weight: 600; font-size: 0.9rem; display: inline-flex; align-items: center; gap: 8px; height: 36px; box-sizing: border-box;"> <img src="https://upload.wikimedia.org/wikipedia/commons/d/d0/Google_Colaboratory_SVG_Logo.svg" width="16" height="16" alt="Colab"> Training </a> </div> </div> <!-- CITATION --> <div style="background-color: #121212; border: 1px solid #27272a; border-radius: 12px; padding: 2rem;"> <h2 style="color: #ffffff; border-bottom: 1px solid #27272a; padding-bottom: 1rem; margin-top: 0;">Citation</h2> <pre style="background-color: #18181b; color: #9ca3af; padding: 1rem; border-radius: 6px; overflow-x: auto; border: 1px solid #27272a;"><code>@misc{cygnis_alpha_2_v0.3, author = {Simonc-44}, title = {Cygnis-Alpha-2 8B v0.3: Sovereign Reasoning Engine}, year = {2024}, publisher = {Hugging Face}, url = {https://huggingface.co/Simonc-44/Cygnis-Alpha-2-8B-v0.3} }</code></pre> </div> <div style="text-align: center; margin-top: 4rem; color: #6b7280; font-size: 0.9rem;"> Developed by Simonc-44 </div> </div>