Instructions to use klyrone/Chimera with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use klyrone/Chimera with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf klyrone/Chimera:Q5_K_M # Run inference directly in the terminal: llama cli -hf klyrone/Chimera:Q5_K_M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf klyrone/Chimera:Q5_K_M # Run inference directly in the terminal: llama cli -hf klyrone/Chimera:Q5_K_M
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf klyrone/Chimera:Q5_K_M # Run inference directly in the terminal: ./build/bin/llama cli -hf klyrone/Chimera:Q5_K_M
Use Docker
docker model run hf.co/klyrone/Chimera:Q5_K_M
- Llama new
- LM Studio
- Jan
- vLLM
How to use klyrone/Chimera with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "klyrone/Chimera" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "klyrone/Chimera", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/klyrone/Chimera:Q5_K_M
- Ollama
How to use klyrone/Chimera with Ollama:
ollama run hf.co/klyrone/Chimera:Q5_K_M
- Unsloth Desktop
- Docker Model Runner
How to use klyrone/Chimera with Docker Model Runner:
docker model run hf.co/klyrone/Chimera:Q5_K_M
- Lemonade
How to use klyrone/Chimera with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull klyrone/Chimera:Q5_K_M
Run and chat with the model
lemonade run user.Chimera-Q5_K_M
List all available models
lemonade list
- Atomic Chat
Add model card
Browse files
README.md
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- chimera
|
| 5 |
+
- moe
|
| 6 |
+
- mixture-of-experts
|
| 7 |
+
- gguf
|
| 8 |
+
- klyrone
|
| 9 |
+
language:
|
| 10 |
+
- en
|
| 11 |
+
pipeline_tag: text-generation
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
# Chimera 8x7B
|
| 15 |
+
|
| 16 |
+
**Chimera** is a Mixture-of-Experts language model developed by [Klyrone Tech](https://huggingface.co/kk497055), built using our proprietary **Amalgamation of Experts (AoE)** technique. Chimera features 8 specialized expert networks with top-2 routing, delivering strong instruction-following and reasoning capabilities with an efficient 12.9B active parameter footprint.
|
| 17 |
+
|
| 18 |
+
## Model Details
|
| 19 |
+
|
| 20 |
+
| | |
|
| 21 |
+
|---|---|
|
| 22 |
+
| **Architecture** | Mixture of Experts (MoE) - 8 experts, top-2 routing |
|
| 23 |
+
| **Total Parameters** | 46.7B |
|
| 24 |
+
| **Active Parameters** | 12.9B per token |
|
| 25 |
+
| **Context Length** | 32,768 tokens |
|
| 26 |
+
| **Quantization** | Q5_K_M (GGUF) |
|
| 27 |
+
| **Developed by** | Klyrone Tech |
|
| 28 |
+
|
| 29 |
+
## Key Features
|
| 30 |
+
|
| 31 |
+
- **Efficient MoE Architecture** - Only 12.9B parameters active per forward pass despite 46.7B total, enabling fast inference
|
| 32 |
+
- **Specialized Expert Networks** - 8 expert FFN modules with learned routing for task-adaptive computation
|
| 33 |
+
- **Instruction-Tuned Experts** - Expert networks optimized for instruction following, code generation, and reasoning
|
| 34 |
+
- **Long Context** - Supports up to 32K token context windows with RoPE positional encoding
|
| 35 |
+
|
| 36 |
+
## Amalgamation of Experts (AoE)
|
| 37 |
+
|
| 38 |
+
Chimera is built using our **AoE** technique, a novel approach to constructing high-quality MoE models by strategically assembling expert networks. AoE enables the creation of models that combine specialized capabilities from multiple training paradigms into a unified, coherent architecture.
|
| 39 |
+
|
| 40 |
+
## Usage
|
| 41 |
+
|
| 42 |
+
### With llama.cpp
|
| 43 |
+
```bash
|
| 44 |
+
./llama-cli -m Chimera-8x7B-Q5_K_M.gguf -p "Your prompt here" -n 500 -ngl 99
|
| 45 |
+
```
|
| 46 |
+
|
| 47 |
+
## Intended Use
|
| 48 |
+
|
| 49 |
+
Chimera is designed for general-purpose text generation including conversational AI, code generation, reasoning, and instruction following.
|
| 50 |
+
|
| 51 |
+
## License
|
| 52 |
+
|
| 53 |
+
Apache 2.0
|