Update README.md

README.md

@@ -1,62 +1,99 @@
 ---
 base_model: Qwen/Qwen2.5-3B-Instruct
 library_name: peft
-model_name: aether_2_5_pro_model
+pipeline_tag: text-generation
 tags:
 - base_model:adapter:Qwen/Qwen2.5-3B-Instruct
 - lora
 - sft
 - transformers
 - trl
-licence: license
-pipeline_tag: text-generation
+- german
+- english
+- aether
+- aether-2.5
+- gguf
+license: apache-2.0
+language:
+- de
+- en
 ---
 
-# Model Card for aether_2_5_pro_model
+# Aether 2.5 Pro
 
-This model is a fine-tuned version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct).
-It has been trained using [TRL](https://github.com/huggingface/trl).
+Aether 2.5 Pro is the strongest model in the Aether 2.5 series so far.  
+It is a fine-tuned version of **Qwen2.5-3B-Instruct**, trained with TRL + PEFT (LoRA) on a higher-quality and more diverse dataset than previous versions.
 
-## Quick start
+Compared to the standard Aether 2.5, the Pro version offers:
+- Better reasoning
+- Improved instruction following
+- Stronger multilingual performance (German + English)
+- Higher overall response quality while staying efficient for local use
 
-```python
-from transformers import pipeline
+> 🚀 **Looking for GGUF versions?**  
+> If you want to run Aether 2.5 Pro locally via **LM Studio**, **Ollama**, or **llama.cpp**, check out the pre-quantized GGUF repository:  
+> 👉 **[Maxilicious20/Aether-2.5-Pro-GGUF](https://huggingface.co/Maxilicious20/Aether-2.5-Pro-GGUF)**
+
+## Model Details
+
+### Model Description
 
-question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
-generator = pipeline("text-generation", model="None", device="cuda")
-output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
-print(output["generated_text"])
-```
+- **Developed by:** Maxilicious20 (Mono AI Studio)
+- **Model type:** Causal Language Model (LoRA Adapter)
+- **Language(s):** German, English
+- **License:** Apache-2.0
+- **Finetuned from model:** [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct)
 
-## Training procedure
+## Uses
 
- 
+### Direct Use
 
+Aether 2.5 Pro is designed for:
+- High-quality conversational AI
+- Reasoning and problem solving
+- Instruction following
+- General text generation
+- Local deployment on consumer hardware
 
+### Quantized & GGUF Models
 
-This model was trained with SFT.
+For easy local usage without Python, use the GGUF versions:
+
+* 📦 **GGUF Repository:** [Maxilicious20/Aether-2.5-Pro-GGUF](https://huggingface.co/Maxilicious20/Aether-2.5-Pro-GGUF)
+* **Recommended quantizations:**
+  * `aether-2.5-pro-q4_k_m.gguf` → Best balance (recommended)
+  * `aether-2.5-pro-q5_k_m.gguf` → Higher quality
+  * `aether-2.5-pro-q8_0.gguf` → Near full quality
+  * `aether-2.5-pro-f16.gguf` → Full precision
+
+## How to Use
+
+### Python (Transformers + PEFT)
+
+```python
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import PeftModel
 
-### Framework versions
+base_model_id = "Qwen/Qwen2.5-3B-Instruct"
+adapter_id = "Maxilicious20/Aether-2.5-Pro"
 
-- PEFT 0.20.0
-- TRL: 1.9.2
-- Transformers: 5.14.1
-- Pytorch: 2.13.0+cu126
-- Datasets: 5.0.1
-- Tokenizers: 0.22.2
+tokenizer = AutoTokenizer.from_pretrained(base_model_id)
+base_model = AutoModelForCausalLM.from_pretrained(
+    base_model_id,
+    torch_dtype=torch.bfloat16,
+    device_map="auto"
+)
 
-## Citations
+model = PeftModel.from_pretrained(base_model, adapter_id)
 
+messages = [
+    {"role": "system", "content": "You are Aether 2.5 Pro, a highly capable AI assistant developed by Mono AI Studio."},
+    {"role": "user", "content": "Explain the difference between supervised and unsupervised learning in simple terms."}
+]
 
+prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
 
-Cite TRL as:
-    
-```bibtex
-@software{vonwerra2020trl,
-  title   = {{TRL: Transformers Reinforcement Learning}},
-  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
-  license = {Apache-2.0},
-  url     = {https://github.com/huggingface/trl},
-  year    = {2020}
-}
-```
+outputs = model.generate(**inputs, max_new_tokens=512)
+print(tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))