eachadea

/

ggml-toolpaca-13b-4bit

like 6

A snippet from the dataset should show you roughly how to use it

 {
    "instruction": "toolformer: enabled\ntoolformer access: wikipedia\nA wrapper around Wikipedia. Useful for when you need to display general information about people, places, companies, historical events, or other subjects found in an encyclopedia, displays a snippet summary of the topic. Input should be a search query.\nwikipedia(query)\nProvide a summary of a famous historical event based on the input event given",
    "input": "French Revolution",
    "response": "Summary of the French Revolution: wikipedia('French Revolution')"
  }```

https://huggingface.co/chavinlo/toolpaca/discussions/1

https://github.com/Significant-Gravitas/Auto-GPT/issues/567#issuecomment-1501163988

maybe this model will help: huggingface.co/eachadea/ggml-toolpaca-13b-4bit includes the weights of Meta's open-source implementation of Toolformer (Language Models Can Teach Themselves to Use Tools by Meta AI) is now recombined with Llama.

https://huggingface.co/datasets/dvlamis/toolformer-dataset

https://huggingface.co/datasets/enoreyes/success-llm-instructions

https://github.com/teknium1/GPTeacher

They are all made to be compliant with Alpaca's dataset format, i.e. each has an instruction, input, and output field, should make it easier to use the same fine tune script and process as alpaca has.

Described here: https://huggingface.co/datasets/tatsu-lab/alpaca#data-instances

And then you combine them into this prompt:

Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
{instruction}

### Input:
{input}

### Response:

But the actual results I get:

 Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
toolformer: enabled
toolformer access: weather
Useful for when you want to get weather information from the OpenMeteo API. The input should be a question in natural language that this API can answer.
weather(querywithlocation)
Give the current temperature and weather condition for a given city

### Input:
Los Angeles

### Response:The current temperature in Los Angeles is 68°F (20°C) and the weather condition is sunny. [end of text]

 Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
toolformer: enabled
toolformer access: wolfram
A wolfram alpha search engine. Useful for when you need to answer questions about Math, Science, Technology, Culture, Society and Everyday Life. Input should be a search query.
wolfram(query)
Answer a question related to math or science based on the input given

### Input:
Formula for calculating the area of a circle

### Response:
In mathematics, the formula for calculating the area of a circle is A = πr^2, where A is the area, π is approximately 3.14159, and r is the radius of the circle. [end of text]

😕