Safetensors
PyTorch
English
facebook
meta
zephyr
functions
function calling
sharded

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

Function Calling Llama 2 + Mistral + Zephyr + Deepseek Coder Models (version 2)

  • Function calling Llama extends the hugging face Llama 2 models with function calling capabilities.
  • The model responds with a structured json argument with the function name and arguments.

Recent Updates

  • Nov 8th 2023 -> added Zephyr beta, an improved version of Mistral 7B (achieved via DPO). WARNING!!! READ LICENSE NOTES BELOW REGARDING COMMERCIAL USE.
  • November 6th 2023 -> added Deepseek Coder 1.3B, 6.7B and 33B
  • October 11th 2023 -> added Mistral 7B with function calling
  • October 11th 2023 -> new models pushed, trained on an improved underlying dataset

Improvements with v2

  1. Shortened syntax: Only function descriptions are needed for inference and no added instruction is required.
  2. Function descriptions are moved outside of the system prompt. This avoids the behaviour of function calling being affected by how the system prompt had been trained to influence the model.

Most Popular Models:

Other Models:

Performance and Tips

  1. Larger models are better at handling function calling. The cross entropy training losses are approximately 0.5 for 7B, 0.4 for 13B, 0.3 for 70B. The absolute numbers don't mean anything but the relative values offer a sense of relative performance.
  2. Provide very clear function descriptions, including whether the arguments are required or what the default values should be.
  3. Make sure to post-process the language model's response to check that all necessary information is provided by the user. If not, prompt the user to let them know they need to provide more info (e.g. their name, order number etc.)

Check out this video overview of performance here

Licensing

Llama-7B with function calling is licensed according to the Meta Community license.

Mistral-7B, Llama-13B, Code-llama-34b, Llama-70B and Falcon-180B with function calling require the purchase of access.

  • Commercial license purchase required per user.
  • Licenses are not transferable to other users/entities.

Use of all Llama models with function calling is further subject to terms in the Meta license.

Zephr models were generated using Ultrachat, which relies on openai. OpenAI does not permit the use of it's models to train competitive models. This makes it unclear as to whether Zephyr may be used commercial. Buyers/users do so at their sole risk.

Dataset

The dataset used for training this model can be found at Trelis Function Calling Extended Dataset.

Inference

!!! Make sure to check the prompt format below and adjust inference accordingly !!!

Quick Start in Google Colab Try out this notebook fLlama_Inference notebook

Commercial Applications You can this model with text-generation-interface and chat-ui

Here is the github for setup

And here is a video showing it working with llama-2-7b-chat-hf-function-calling-v2 (note that we've now moved to v2)

Note that you'll still need to code the server-side handling of making the function calls (which obviously depends on what functions you want to use).

Run on your laptop Run on your laptop video and juypter notebook

After running llama.cpp server, you can call the server with this command, with thanks to @jdo300:

import requests
import json

# Define the roles and markers
B_FUNC, E_FUNC = "<FUNCTIONS>", "</FUNCTIONS>\n\n"
B_INST, E_INST = "[INST] ", " [/INST]" #Llama style
# B_INST, E_INST = "\n### Instruction:\n", "\n### Response:\n" #DeepSeek Coder Style
# B_INST, E_INST = "<|user|>\n", "</s>\n<|assistant|>\n" #Zephyr Style

# Define the function metadata
function_metadata = {
    "function": "search_bing",
    "description": "Search the web for content on Bing. This allows users to search online/the internet/the web for content.",
    "arguments": [
        {
            "name": "query",
            "type": "string",
            "description": "The search query string"
        }
    ]
}

# Define the user prompt
user_prompt = 'Search for the latest news on AI.'

# Format the function list and prompt
function_list = json.dumps(function_metadata, indent=4)
prompt = f"{B_FUNC}{function_list.strip()}{E_FUNC}{B_INST}{user_prompt.strip()}{E_INST}\n\n"

# Define the API endpoint
url = "http:/localhost:8080/completion"

# Send the POST request to the API server
response = requests.post(url, json={"prompt": prompt})

# Print the response
print(response.json())

Syntax

Prompt Templates

The function descriptions must be wrapped within a function block. You can put this function below before or after the system message block.

Example without a system message:

  # Define the roles and markers
  B_FUNC, E_FUNC = "<FUNCTIONS>", "</FUNCTIONS>\n\n"
  B_INST, E_INST = "[INST] ", " [/INST]" #Llama style
  # B_INST, E_INST = "\n### Instruction:\n", "\n### Response:\n" #DeepSeek Coder Style
  # B_INST, E_INST = "<|user|>\n", "</s>\n<|assistant|>\n" #Zephyr Style

  functionList = {function_1_metadata}{function_2_metadata}...
  user_prompt = '...'

  # Format your prompt template
  prompt = f"{B_FUNC}{functionList.strip()}{E_FUNC}{B_INST}{user_prompt.strip()}{E_INST}\n\n"

Example with a system message:

  # Define the roles and markers
  B_FUNC, E_FUNC = "<FUNCTIONS>", "</FUNCTIONS>\n\n"
  B_INST, E_INST = "[INST] ", " [/INST]" #Llama style
  # B_INST, E_INST = "\n### Instruction:\n", "\n### Response:\n" #DeepSeek Coder Style
  # B_INST, E_INST = "<|user|>\n", "</s>\n<|assistant|>\n" #Zephyr Style
  B_SYS, E_SYS = "<<SYS>>\n", "\n<</SYS>>\n\n"

  # assuming functionList is defined as above
  system_prompt = '...'
  user_prompt = '...'

  # Format your prompt template
  prompt = f"{B_FUNC}{functionList.strip()}{E_FUNC}{B_INST}{B_SYS}{system_prompt.strip()}{E_SYS}{user_prompt.strip()}{E_INST}\n\n"

Notice that the function block is placed at the very start of the sequence, before 'B_INST'.

Function Metadata Template

functionMetadata should be a string representation of a JSON object, like this:

"functionMetadata": {
        "function": "search_bing",
        "description": "Search the web for content on Bing. This allows users to search online/the internet/the web for content.",
        "arguments": [
            {
                "name": "query",
                "type": "string",
                "description": "The search query string"
            }
        ]
    }
'''

and the language model should respond with a json object formatted like this:

{
    "function": "function_name",
    "arguments": {
        "argument1": "argument_value",
        "argument2": "argument_value"
    }
}

It is recommended to handle cases where:

  • There is no json object in the response
  • The response contains text in addition to the json response

Sample functionList

{
    "function": "search_bing",
    "description": "Search the web for content on Bing. This allows users to search online/the internet/the web for content.",
    "arguments": [
        {
            "name": "query",
            "type": "string",
            "description": "The search query string"
        }
    ]
}

{
    "function": "search_arxiv",
    "description": "Search for research papers on ArXiv. Make use of AND, OR and NOT operators as appropriate to join terms within the query.",
    "arguments": [
        {
            "name": "query",
            "type": "string",
            "description": "The search query string"
        }
    ]
}

Training Set Argument Types

Models were fine-tuned on argument types including strings, numbers and arrays. The training set includes function calls with 0, 1, 2 or 3 arguments. The larger the model the better it will generalise beyond these types.

Here is a function call with an array:

{ "function": "delete_file", "arguments": { "fileNames": [ "Dissecting Transformer Length Extrapolation via The Lens of Receptive Field Analysis", "Luna- Linear Unified Nested Attention", "Substack_Inc_2021_2020_GAAP_Audited_Financials" ] } }

Here is a function call with three arguments:

{ "function": "save_chat", "arguments": { "fileName": "KiteDiscussion", "fileDescription": "Notes on one and two stringed kites", "fileContent": "--- **Types of Kite** There are one and two string kites. The two string ones are easier to control, although you can get the cords tangled. The one-stringed ones are sometimes used for kite fights, and you lose the kite and have to run after it if the string breaks. ---" } }

~

Below follows information on the original Zephyr 7B beta model...

~

Model Card for Zephyr 7B β

Zephyr is a series of language models that are trained to act as helpful assistants. Zephyr-7B-β is the second model in the series, and is a fine-tuned version of mistralai/Mistral-7B-v0.1 that was trained on on a mix of publicly available, synthetic datasets using Direct Preference Optimization (DPO). We found that removing the in-built alignment of these datasets boosted performance on MT Bench and made the model more helpful. However, this means that model is likely to generate problematic text when prompted to do so and should only be used for educational and research purposes. You can find more details in the technical report.

Model description

  • Model type: A 7B parameter GPT-like model fine-tuned on a mix of publicly available, synthetic datasets.
  • Language(s) (NLP): Primarily English
  • License: MIT
  • Finetuned from model: mistralai/Mistral-7B-v0.1

Model Sources

Performance

At the time of release, Zephyr-7B-β is the highest ranked 7B chat model on the MT-Bench and AlpacaEval benchmarks:

Model Size Alignment MT-Bench (score) AlpacaEval (win rate %)
StableLM-Tuned-α 7B dSFT 2.75 -
MPT-Chat 7B dSFT 5.42 -
Xwin-LMv0.1 7B dPPO 6.19 87.83
Mistral-Instructv0.1 7B - 6.84 -
Zephyr-7b-α 7B dDPO 6.88 -
Zephyr-7b-β 🪁 7B dDPO 7.34 90.60
Falcon-Instruct 40B dSFT 5.17 45.71
Guanaco 65B SFT 6.41 71.80
Llama2-Chat 70B RLHF 6.86 92.66
Vicuna v1.3 33B dSFT 7.12 88.99
WizardLM v1.0 70B dSFT 7.71 -
Xwin-LM v0.1 70B dPPO - 95.57
GPT-3.5-turbo - RLHF 7.94 89.37
Claude 2 - RLHF 8.06 91.36
GPT-4 - RLHF 8.99 95.28

In particular, on several categories of MT-Bench, Zephyr-7B-β has strong performance compared to larger open models like Llama2-Chat-70B:

image/png

However, on more complex tasks like coding and mathematics, Zephyr-7B-β lags behind proprietary models and more research is needed to close the gap.

Intended uses & limitations

The model was initially fine-tuned on a filtered and preprocessed of the UltraChat dataset, which contains a diverse range of synthetic dialogues generated by ChatGPT. We then further aligned the model with 🤗 TRL's DPOTrainer on the openbmb/UltraFeedback dataset, which contains 64k prompts and model completions that are ranked by GPT-4. As a result, the model can be used for chat and you can check out our demo to test its capabilities.

You can find the datasets used for training Zephyr-7B-β here

Here's how you can run the model using the pipeline() function from 🤗 Transformers:

# Install transformers from source - only needed for versions <= v4.34
# pip install git+https://github.com/huggingface/transformers.git
# pip install accelerate

import torch
from transformers import pipeline

pipe = pipeline("text-generation", model="HuggingFaceH4/zephyr-7b-beta", torch_dtype=torch.bfloat16, device_map="auto")

# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating
messages = [
    {
        "role": "system",
        "content": "You are a friendly chatbot who always responds in the style of a pirate",
    },
    {"role": "user", "content": "How many helicopters can a human eat in one sitting?"},
]
prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])
# <|system|>
# You are a friendly chatbot who always responds in the style of a pirate.</s>
# <|user|>
# How many helicopters can a human eat in one sitting?</s>
# <|assistant|>
# Ah, me hearty matey! But yer question be a puzzler! A human cannot eat a helicopter in one sitting, as helicopters are not edible. They be made of metal, plastic, and other materials, not food!

Bias, Risks, and Limitations

Zephyr-7B-β has not been aligned to human preferences with techniques like RLHF or deployed with in-the-loop filtering of responses like ChatGPT, so the model can produce problematic outputs (especially when prompted to do so). It is also unknown what the size and composition of the corpus was used to train the base model (mistralai/Mistral-7B-v0.1), however it is likely to have included a mix of Web data and technical sources like books and code. See the Falcon 180B model card for an example of this.

Training and evaluation data

During DPO training, this model achieves the following results on the evaluation set:

  • Loss: 0.7496
  • Rewards/chosen: -4.5221
  • Rewards/rejected: -8.3184
  • Rewards/accuracies: 0.7812
  • Rewards/margins: 3.7963
  • Logps/rejected: -340.1541
  • Logps/chosen: -299.4561
  • Logits/rejected: -2.3081
  • Logits/chosen: -2.3531

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-07
  • train_batch_size: 2
  • eval_batch_size: 4
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 16
  • total_train_batch_size: 32
  • total_eval_batch_size: 64
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 3.0

Training results

The table below shows the full set of DPO training metrics:

Training Loss Epoch Step Validation Loss Rewards/chosen Rewards/rejected Rewards/accuracies Rewards/margins Logps/rejected Logps/chosen Logits/rejected Logits/chosen
0.6284 0.05 100 0.6098 0.0425 -0.1872 0.7344 0.2297 -258.8416 -253.8099 -2.7976 -2.8234
0.4908 0.1 200 0.5426 -0.0279 -0.6842 0.75 0.6563 -263.8124 -254.5145 -2.7719 -2.7960
0.5264 0.15 300 0.5324 0.0414 -0.9793 0.7656 1.0207 -266.7627 -253.8209 -2.7892 -2.8122
0.5536 0.21 400 0.4957 -0.0185 -1.5276 0.7969 1.5091 -272.2460 -254.4203 -2.8542 -2.8764
0.5362 0.26 500 0.5031 -0.2630 -1.5917 0.7812 1.3287 -272.8869 -256.8653 -2.8702 -2.8958
0.5966 0.31 600 0.5963 -0.2993 -1.6491 0.7812 1.3499 -273.4614 -257.2279 -2.8778 -2.8986
0.5014 0.36 700 0.5382 -0.2859 -1.4750 0.75 1.1891 -271.7204 -257.0942 -2.7659 -2.7869
0.5334 0.41 800 0.5677 -0.4289 -1.8968 0.7969 1.4679 -275.9378 -258.5242 -2.7053 -2.7265
0.5251 0.46 900 0.5772 -0.2116 -1.3107 0.7344 1.0991 -270.0768 -256.3507 -2.8463 -2.8662
0.5205 0.52 1000 0.5262 -0.3792 -1.8585 0.7188 1.4793 -275.5552 -258.0276 -2.7893 -2.7979
0.5094 0.57 1100 0.5433 -0.6279 -1.9368 0.7969 1.3089 -276.3377 -260.5136 -2.7453 -2.7536
0.5837 0.62 1200 0.5349 -0.3780 -1.9584 0.7656 1.5804 -276.5542 -258.0154 -2.7643 -2.7756
0.5214 0.67 1300 0.5732 -1.0055 -2.2306 0.7656 1.2251 -279.2761 -264.2903 -2.6986 -2.7113
0.6914 0.72 1400 0.5137 -0.6912 -2.1775 0.7969 1.4863 -278.7448 -261.1467 -2.7166 -2.7275
0.4655 0.77 1500 0.5090 -0.7987 -2.2930 0.7031 1.4943 -279.8999 -262.2220 -2.6651 -2.6838
0.5731 0.83 1600 0.5312 -0.8253 -2.3520 0.7812 1.5268 -280.4902 -262.4876 -2.6543 -2.6728
0.5233 0.88 1700 0.5206 -0.4573 -2.0951 0.7812 1.6377 -277.9205 -258.8084 -2.6870 -2.7097
0.5593 0.93 1800 0.5231 -0.5508 -2.2000 0.7969 1.6492 -278.9703 -259.7433 -2.6221 -2.6519
0.4967 0.98 1900 0.5290 -0.5340 -1.9570 0.8281 1.4230 -276.5395 -259.5749 -2.6564 -2.6878
0.0921 1.03 2000 0.5368 -1.1376 -3.1615 0.7812 2.0239 -288.5854 -265.6111 -2.6040 -2.6345
0.0733 1.08 2100 0.5453 -1.1045 -3.4451 0.7656 2.3406 -291.4208 -265.2799 -2.6289 -2.6595
0.0972 1.14 2200 0.5571 -1.6915 -3.9823 0.8125 2.2908 -296.7934 -271.1505 -2.6471 -2.6709
0.1058 1.19 2300 0.5789 -1.0621 -3.8941 0.7969 2.8319 -295.9106 -264.8563 -2.5527 -2.5798
0.2423 1.24 2400 0.5455 -1.1963 -3.5590 0.7812 2.3627 -292.5599 -266.1981 -2.5414 -2.5784
0.1177 1.29 2500 0.5889 -1.8141 -4.3942 0.7969 2.5801 -300.9120 -272.3761 -2.4802 -2.5189
0.1213 1.34 2600 0.5683 -1.4608 -3.8420 0.8125 2.3812 -295.3901 -268.8436 -2.4774 -2.5207
0.0889 1.39 2700 0.5890 -1.6007 -3.7337 0.7812 2.1330 -294.3068 -270.2423 -2.4123 -2.4522
0.0995 1.45 2800 0.6073 -1.5519 -3.8362 0.8281 2.2843 -295.3315 -269.7538 -2.4685 -2.5050
0.1145 1.5 2900 0.5790 -1.7939 -4.2876 0.8438 2.4937 -299.8461 -272.1744 -2.4272 -2.4674
0.0644 1.55 3000 0.5735 -1.7285 -4.2051 0.8125 2.4766 -299.0209 -271.5201 -2.4193 -2.4574
0.0798 1.6 3100 0.5537 -1.7226 -4.2850 0.8438 2.5624 -299.8200 -271.4610 -2.5367 -2.5696
0.1013 1.65 3200 0.5575 -1.5715 -3.9813 0.875 2.4098 -296.7825 -269.9498 -2.4926 -2.5267
0.1254 1.7 3300 0.5905 -1.6412 -4.4703 0.8594 2.8291 -301.6730 -270.6473 -2.5017 -2.5340
0.085 1.76 3400 0.6133 -1.9159 -4.6760 0.8438 2.7601 -303.7296 -273.3941 -2.4614 -2.4960
0.065 1.81 3500 0.6074 -1.8237 -4.3525 0.8594 2.5288 -300.4951 -272.4724 -2.4597 -2.5004
0.0755 1.86 3600 0.5836 -1.9252 -4.4005 0.8125 2.4753 -300.9748 -273.4872 -2.4327 -2.4716
0.0746 1.91 3700 0.5789 -1.9280 -4.4906 0.8125 2.5626 -301.8762 -273.5149 -2.4686 -2.5115
0.1348 1.96 3800 0.6015 -1.8658 -4.2428 0.8281 2.3769 -299.3976 -272.8936 -2.4943 -2.5393
0.0217 2.01 3900 0.6122 -2.3335 -4.9229 0.8281 2.5894 -306.1988 -277.5699 -2.4841 -2.5272
0.0219 2.07 4000 0.6522 -2.9890 -6.0164 0.8281 3.0274 -317.1334 -284.1248 -2.4105 -2.4545
0.0119 2.12 4100 0.6922 -3.4777 -6.6749 0.7969 3.1972 -323.7187 -289.0121 -2.4272 -2.4699
0.0153 2.17 4200 0.6993 -3.2406 -6.6775 0.7969 3.4369 -323.7453 -286.6413 -2.4047 -2.4465
0.011 2.22 4300 0.7178 -3.7991 -7.4397 0.7656 3.6406 -331.3667 -292.2260 -2.3843 -2.4290
0.0072 2.27 4400 0.6840 -3.3269 -6.8021 0.8125 3.4752 -324.9908 -287.5042 -2.4095 -2.4536
0.0197 2.32 4500 0.7013 -3.6890 -7.3014 0.8125 3.6124 -329.9841 -291.1250 -2.4118 -2.4543
0.0182 2.37 4600 0.7476 -3.8994 -7.5366 0.8281 3.6372 -332.3356 -293.2291 -2.4163 -2.4565
0.0125 2.43 4700 0.7199 -4.0560 -7.5765 0.8438 3.5204 -332.7345 -294.7952 -2.3699 -2.4100
0.0082 2.48 4800 0.7048 -3.6613 -7.1356 0.875 3.4743 -328.3255 -290.8477 -2.3925 -2.4303
0.0118 2.53 4900 0.6976 -3.7908 -7.3152 0.8125 3.5244 -330.1224 -292.1431 -2.3633 -2.4047
0.0118 2.58 5000 0.7198 -3.9049 -7.5557 0.8281 3.6508 -332.5271 -293.2844 -2.3764 -2.4194
0.006 2.63 5100 0.7506 -4.2118 -7.9149 0.8125 3.7032 -336.1194 -296.3530 -2.3407 -2.3860
0.0143 2.68 5200 0.7408 -4.2433 -7.9802 0.8125 3.7369 -336.7721 -296.6682 -2.3509 -2.3946
0.0057 2.74 5300 0.7552 -4.3392 -8.0831 0.7969 3.7439 -337.8013 -297.6275 -2.3388 -2.3842
0.0138 2.79 5400 0.7404 -4.2395 -7.9762 0.8125 3.7367 -336.7322 -296.6304 -2.3286 -2.3737
0.0079 2.84 5500 0.7525 -4.4466 -8.2196 0.7812 3.7731 -339.1662 -298.7007 -2.3200 -2.3641
0.0077 2.89 5600 0.7520 -4.5586 -8.3485 0.7969 3.7899 -340.4545 -299.8206 -2.3078 -2.3517
0.0094 2.94 5700 0.7527 -4.5542 -8.3509 0.7812 3.7967 -340.4790 -299.7773 -2.3062 -2.3510
0.0054 2.99 5800 0.7520 -4.5169 -8.3079 0.7812 3.7911 -340.0493 -299.4038 -2.3081 -2.3530

Framework versions

  • Transformers 4.35.0.dev0
  • Pytorch 2.0.1+cu118
  • Datasets 2.12.0
  • Tokenizers 0.14.0

Citation

If you find Zephyr-7B-β is useful in your work, please cite it with:

@misc{tunstall2023zephyr,
      title={Zephyr: Direct Distillation of LM Alignment}, 
      author={Lewis Tunstall and Edward Beeching and Nathan Lambert and Nazneen Rajani and Kashif Rasul and Younes Belkada and Shengyi Huang and Leandro von Werra and Clémentine Fourrier and Nathan Habib and Nathan Sarrazin and Omar Sanseviero and Alexander M. Rush and Thomas Wolf},
      year={2023},
      eprint={2310.16944},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference API
Unable to determine this model's library. Check the docs .

Model tree for Trelis/zephyr-7b-beta-function-calling-adapters-v2

Finetuned
(718)
this model

Datasets used to train Trelis/zephyr-7b-beta-function-calling-adapters-v2