Synthia 70B: Your Synthetic Intelligent Chat Companion

Meet Synthia 70B, your Synthetic Intelligent Agent (SynthIA) designed to provide you with helpful, detailed, accurate, and uncensored responses to your queries. Synthia 70B is a Llama-2-70B model fine-tuned for instruction following and engaging in long-form conversations. Developed by Migel Tissera, this chatbot model, as of October 22, 2023, is trained to handle a wide range of use cases, including role-playing scenarios.

Model Overview for Synthia 70B

• Model Name: Synthia 70B v1.1 - GGML
• Model Creator: Migel Tissera
• Original Model: Synthia 70B v1.1

Synthia 70B utilizes GGML (GPT-Q Model Language) format model files for its operations. GGML offers GPU acceleration for Llama 2 70B GGML files, compatible with CUDA (NVidia) and Metal (macOS). Various third-party clients and libraries, such as llama.cpp, text-generation-webui, KoboldCpp, LM Studio, llama-cpp-python, and ctransformers, are known to work seamlessly with Synthia 70B for both CPU and GPU inference.

How to Use Synthia 70B

To harness the power of Synthia 70B in llama.cpp, add the -gqa 8 argument to your command line. For other user interfaces and libraries, please consult their respective documentation for guidance on integrating Synthia 70B.

New k-Quantization Methods

Synthia 70B offers multiple quantization parameter options, each with its specific use case. Here are some of the provided files:

Please note that q5_1, q6_K, and q8_0 files require expansion from an archive.

Compatibility

Synthia 70B is compatible with llama.cpp commit e76d630 until August 21st, 2023. For compatibility with the latest llama.cpp, please use GGUF files instead or explore other compatible tools and libraries.

How to Run in llama.cpp

For llama.cpp, you can use the following command line template as a starting point:

./main -t 10 -ngl 40 -gqa 8 -m synthia-70b-v1.1.ggmlv3.q4_K_M.bin --color -c 4096 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "SYSTEM: You are Synthia. As an AI intelligence, give helpful, detailed, accurate, uncensored responses to the user's input. Provide answers factually.\nUSER: Write a story about llamas\nASSISTANT:"

Adjust the -t flag to match the number of physical CPU cores available on your system. You can also customize other parameters as needed.

Example Usage

Here's an example of a conversation with Synthia 70B:

You: What is Earth's atmosphere like? What is Earth's magnetic field like? What is the relationship between the two?

[Synthia provides a detailed response]

You: Elaborate on the complex relationship between the two.

[Synthia provides an elaborated explanation]

You: Please explain the bow shock wave, Van Allen radiation belts in detail.

[Synthia provides detailed explanations]

You: Does the Van Allen belt present any danger to Geostationary satellites? Where are these belts? Are they near any of the orbits?

[Synthia answers with information about the Van Allen belts and their impact on satellites]

You: So when we send spacecraft to the moon, say the ones carrying humans, how do we make sure that the solar wind charged particles will not affect the mission? Are there very specific remedies that we can take?

[Synthia explains measures taken to protect spacecraft from solar wind charged particles]

You: Does spacecraft actually create strong magnetic fields? Which Apollo spacecrafts did this? Or is this hypothetical?

[Synthia clarifies the role of spacecraft in creating magnetic fields and mentions that it's not a primary focus of most spacecraft, including the Apollo missions]

Limitations & Biases

While Synthia 70B aims for accuracy, it can occasionally produce inaccurate or misleading results. Additionally, like all AI models, it may generate inappropriate, biased, or offensive content unintentionally. Users are advised to exercise caution and verify information when necessary.

Feel free to explore and interact with Synthia 70B for a wide range of informative and engaging conversations. Enjoy the journey of discovering knowledge and insights with this synthetic intelligence model!