Welcome to the world of AI! In this guide, we will walk you through the process of creating and using the high-performance Ninja-V2-7B model. This model is versatile and can handle various tasks, from chat interactions to creative writing.

Overview of the Ninja-V2-7B Model

The Ninja-V2-7B model leverages advanced techniques such as vector merging. Built artfully, it takes advantage of high-performance GPU servers to deliver outstanding results. We deeply appreciate the efforts of the contributors involved in the LocalAI Hackathon.

Creating the Ninja-V2-7B Model

Below is the recipe to construct the Ninja-V2 model, and we’ll break it down for easier understanding.

yaml
models:
 - model: MTSAIRmulti_verse_model
 - model: HuggingFaceH4zephyr-7b-betamerge_method:
  model_stockbase_model: amazingvinceNot-WizardLM-2-7Bdtype: bfloat16
Novels-7B(ninja_mergerにて作成)
yamltarget_model: stabilityaijapanese-stablelm-instruct-gamma-7b
   left:  ElizezenPhos-7B          # ベースモデルの指定
   right: stabilityaijapanese-stablelm-instruct-gamma-7b        # サブモデルの指定
   operation: sub                             # 組み合わせの操作。mix、addなどを指定
   velocity: 1.0
 -    left:  ElizezenAntler-7B          # ベースモデルの指定
   right: stabilityaijapanese-stablelm-instruct-gamma-7b        # サブモデルの指定
   operation: sub                             # 組み合わせの操作。mix、addなどを指定
   velocity: 1.0
Ninja-v2(ninja_mergerにて作成)
yamltarget_model: Ninja-v2-Basemodels:                                     # 組み合わせの重み。0.0から1.0の範囲で指定
 -    left:  NTQAIchatntq-ja-7b-v1.0          # ベースモデルの指定
   right: mistralaiMistral-7B-v0.1        # サブモデルの指定
0
   operation: sub                        # 組み合わせの操作。mix、addなどを指定
   velocity: 1.0
 -    left:  ElizezenBerghof-NSFW-7B          # ベースモデルの指定
   right: stabilityaijapanese-stablelm-instruct-gamma-7b        # サブモデルの指定
   operation: sub                             # 組み合わせの操作。mix、addなどを指定
   velocity: 0.5
 -    left:  Novels-7B          # ベースモデルの指定
   right: stabilityaijapanese-stablelm-instruct-gamma-7b        # サブモデルの指定
   operation: sub                             # 組み合わせの操作。mix、addなどを指定
   velocity: 1.0

Understanding the Creation Code

Think of the Ninja-V2-7B model creation process as preparing a delightful dish in a kitchen. Here’s how the ingredients come together:

• Models as Ingredients: The models specified like MTSAIRmulti_verse_model and HuggingFaceH4zephyr-7b serve as the primary ingredients in our recipe. Just like how specific spices or vegetables contribute to a dish, different models add unique flavors and capabilities to our AI.
• Merging Methods as Cooking Techniques: The merging operations, such as mix and add, are like cooking techniques. A chef might choose to sauté, boil, or bake depending on the desired outcome, just as you select different operations to achieve diverse AI interactions.
• Velocity as Cooking Time: The velocity parameter controls the intensity of the mixing; it’s akin to adjusting the heat while cooking. A low heat might result in a slow simmer while high heat could quickly bring flavors together.

Using Prompt Templates

While a prompt template is not strictly necessary, you can utilize a Vicuna-1.1 template. Here’s how to structure your prompts:

• BAD: あなたは○○として振る舞います (You act as ○○)
• GOOD: あなたは○○です (You are ○○)
• BAD: あなたは○○ができます (You can ○○)
• GOOD: あなたは○○をします (You will ○○)

License Restrictions

Make sure to use this model in accordance with the Apache-2.0 license.

Troubleshooting

While working with the Ninja-V2-7B model, you might encounter a few hiccups. Here are some troubleshooting steps you can take:

• If the model doesn’t perform as expected, double-check your combination parameters like velocity and operation to ensure they are set correctly.
• Ensure that all model references are correct and available in your environment. Sometimes, slight spelling errors can lead to significant performance issues.
• If the output seems off-topic, try refining your prompts by focusing on clearer and more action-oriented phrases.

For more insights, updates, or to collaborate on AI development projects, stay connected with fxis.ai.

At fxis.ai, we believe that such advancements are crucial for the future of AI, as they enable more comprehensive and effective solutions. Our team is continually exploring new methodologies to push the envelope in artificial intelligence, ensuring that our clients benefit from the latest technological innovations.