Hugging Face's logo

Mozilla
/
Mistral-7B-Instruct-v0.3-llamafile

llamafile
Model card Files Community
2
Mistral-7B-Instruct-v0.3-llamafile / README.md
jartine's picture
jartine
Update README.md (#1)
da723a9 verified
preview code
|
raw
history blame contribute delete
16.3 kB
metadata
license: apache-2.0
quantized_by: jartine
model_creator: mistralai
base_model: mistralai/Mistral-7B-Instruct-v0.3
prompt_template: |
  [INST] {{prompt}} [/INST]
tags:
  - llamafile

Mistral 7B Instruct v0.3 - llamafile

This repository contains executable weights (which we call llamafiles) that run on Linux, MacOS, Windows, FreeBSD, OpenBSD, and NetBSD for AMD64 and ARM64.

The third edition of Mistral 7B was released on May 22th, 2024. It increases the vocabulary size to 32768, supports the v3 tokenizer, and introduces support for function calling.

Quickstart

Assuming your system has at least 16GB of RAM, you can try running the following command which download, concatenate, and execute the model.

wget https://huggingface.co/Mozilla/Mistral-7B-Instruct-v0.3-llamafile/resolve/main/Mistral-7B-Instruct-v0.3.Q6_K.llamafile
chmod +x Mistral-7B-Instruct-v0.3.Q6_K.llamafile
./Mistral-7B-Instruct-v0.3.Q6_K.llamafile --help   # view manual
./Mistral-7B-Instruct-v0.3.Q6_K.llamafile          # launch web gui + oai api
./Mistral-7B-Instruct-v0.3.Q6_K.llamafile -p ...   # cli interface (scriptable)

Alternatively, you may download an official llamafile executable from Mozilla Ocho on GitHub, in which case you can use the Granite llamafiles as a simple weights data file.

llamafile -m Mistral-7B-Instruct-v0.3.Q6_K.llamafile ...

For further information, please see the llamafile README.

Having trouble? See the "Gotchas" section of the README.

Prompting

Prompt template:

[INST] {{prompt}} [/INST]

Command template:

./Mistral-7B-Instruct-v0.3.Q6_K.llamafile -p "[INST]{{prompt}}[/INST]"

The maximum context size of this model is 32768 tokens. These llamafiles use a default context size of 4096 tokens. Whenever you need the maximum context size to be available with llamafile for any given model, you can pass the -c 0 flag. The default temperature for these llamafiles is 0.8 because it helps for this model. It can be tuned, e.g. --temp 0.

Benchmarks

hardware model_filename size test t/s
NVIDIA GeForce RTX 4090 (cuBLAS) Mistral-7B-Instruct-v0.3.F16 13.50 GiB pp512 7264.74
NVIDIA GeForce RTX 4090 (cuBLAS) Mistral-7B-Instruct-v0.3.F16 13.50 GiB tg16 58.27
NVIDIA GeForce RTX 4090 (cuBLAS) Mistral-7B-Instruct-v0.3.Q6_K 5.54 GiB pp512 4236.95
NVIDIA GeForce RTX 4090 (cuBLAS) Mistral-7B-Instruct-v0.3.Q6_K 5.54 GiB tg16 114.65
NVIDIA GeForce RTX 4090 (tinyBLAS) Mistral-7B-Instruct-v0.3.Q6_K 5.54 GiB pp512 3457.31
NVIDIA GeForce RTX 4090 (tinyBLAS) Mistral-7B-Instruct-v0.3.Q6_K 5.54 GiB tg16 85.20
NVIDIA GeForce RTX 4090 (tinyBLAS) Mistral-7B-Instruct-v0.3.F16 13.50 GiB pp512 1284.87
NVIDIA GeForce RTX 4090 (tinyBLAS) Mistral-7B-Instruct-v0.3.F16 13.50 GiB tg16 49.76
AMD Radeon RX 7900 XTX (hipBLAS) Mistral-7B-Instruct-v0.3.F16 13.50 GiB pp512 3239.27
AMD Radeon RX 7900 XTX (hipBLAS) Mistral-7B-Instruct-v0.3.F16 13.50 GiB tg16 37.41
AMD Radeon RX 7900 XTX (hipBLAS) Mistral-7B-Instruct-v0.3.Q6_K 5.54 GiB pp512 2647.72
AMD Radeon RX 7900 XTX (hipBLAS) Mistral-7B-Instruct-v0.3.Q6_K 5.54 GiB tg16 85.42
AMD Radeon RX 7900 XTX (tinyBLAS) Mistral-7B-Instruct-v0.3.Q6_K 5.54 GiB pp512 1226.20
AMD Radeon RX 7900 XTX (tinyBLAS) Mistral-7B-Instruct-v0.3.Q6_K 5.54 GiB tg16 76.29
AMD Radeon RX 7900 XTX (tinyBLAS) Mistral-7B-Instruct-v0.3.F16 13.50 GiB pp512 1033.91
AMD Radeon RX 7900 XTX (tinyBLAS) Mistral-7B-Instruct-v0.3.F16 13.50 GiB tg16 35.41
Apple M2 Ultra (60-core Metal GPU) mistral-7b-instruct-v0.3.Q6_K 5.54 GiB pp512 761.88
Apple M2 Ultra (60-core Metal GPU) mistral-7b-instruct-v0.3.Q6_K 5.54 GiB tg16 64.15
Apple M2 Ultra (ARMv8+fp16+dotprod) Mistral-7B-Instruct-v0.3.F16 13.50 GiB pp512 109.18
Apple M2 Ultra (ARMv8+fp16+dotprod) Mistral-7B-Instruct-v0.3.F16 13.50 GiB tg16 15.17
Intel Core i9-14900K (alderlake) Mistral-7B-Instruct-v0.3.Q6_K 5.54 GiB pp512 95.87
Intel Core i9-14900K (alderlake) Mistral-7B-Instruct-v0.3.Q6_K 5.54 GiB tg16 12.66
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.BF16 13.50 GiB pp512 759.25
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.BF16 13.50 GiB tg16 19.29
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.F16 13.50 GiB pp512 559.94
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.F16 13.50 GiB tg16 19.26
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q8_0 7.17 GiB pp512 518.76
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q8_0 7.17 GiB tg16 26.31
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q6_K 5.54 GiB pp512 726.13
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q6_K 5.54 GiB tg16 38.65
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q5_1 5.07 GiB pp512 534.04
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q5_1 5.07 GiB tg16 38.68
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q5_K_M 4.78 GiB pp512 723.25
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q5_K_M 4.78 GiB tg16 41.13
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q5_0 4.65 GiB pp512 536.67
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q5_0 4.65 GiB tg16 42.46
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q5_K_S 4.65 GiB pp512 651.05
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q5_K_S 4.65 GiB tg16 42.14
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q4_1 4.24 GiB pp512 572.67
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q4_1 4.24 GiB tg16 43.19
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q4_K_M 4.07 GiB pp512 728.48
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q4_K_M 4.07 GiB tg16 44.29
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q4_K_S 3.86 GiB pp512 666.82
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q4_K_S 3.86 GiB tg16 45.18
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q4_0 3.83 GiB pp512 562.96
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q4_0 3.83 GiB tg16 48.02
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q3_K_L 3.56 GiB pp512 706.64
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q3_K_L 3.56 GiB tg16 46.82
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q3_K_M 3.28 GiB pp512 715.62
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q3_K_M 3.28 GiB tg16 48.29
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q3_K_S 2.95 GiB pp512 722.11
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q3_K_S 2.95 GiB tg16 49.76
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q2_K 2.53 GiB pp512 739.28
AMD Threadripper PRO 7995WX (znver4) mistral-7b-instruct-v0.3.Q2_K 2.53 GiB tg16 53.01

About llamafile

llamafile is a new format introduced by Mozilla Ocho on Nov 20th 2023. It uses Cosmopolitan Libc to turn LLM weights into runnable llama.cpp binaries that run on the stock installs of six OSes for both ARM64 and AMD64.

In addition to being executables, llamafiles are also zip archives. Each llamafile contains a GGUF file, which you can extract using the unzip command. If you want to change or add files to your llamafiles, then the zipalign command (distributed on the llamafile github) should be used instead of the traditional zip command.

Model Card for Mistral-7B-Instruct-v0.3

The Mistral-7B-Instruct-v0.3 Large Language Model (LLM) is an instruct fine-tuned version of the Mistral-7B-v0.3.

Mistral-7B-v0.3 has the following changes compared to Mistral-7B-v0.2

  • Extended vocabulary to 32768
  • Supports v3 Tokenizer
  • Supports function calling

Installation

It is recommended to use mistralai/Mistral-7B-Instruct-v0.3 with mistral-inference. For HF transformers code snippets, please keep scrolling.

pip install mistral_inference

Download 

from huggingface_hub import snapshot_download
from pathlib import Path

mistral_models_path = Path.home().joinpath('mistral_models', '7B-Instruct-v0.3')
mistral_models_path.mkdir(parents=True, exist_ok=True)

snapshot_download(repo_id="mistralai/Mistral-7B-Instruct-v0.3", allow_patterns=["params.json", "consolidated.safetensors", "tokenizer.model.v3"], local_dir=mistral_models_path)

Chat

After installing mistral_inference, a mistral-chat CLI command should be available in your environment. You can chat with the model using

mistral-chat $HOME/mistral_models/7B-Instruct-v0.3 --instruct --max_tokens 256

Instruct following 

from mistral_inference.model import Transformer
from mistral_inference.generate import generate

from mistral_common.tokens.tokenizers.mistral import MistralTokenizer
from mistral_common.protocol.instruct.messages import UserMessage
from mistral_common.protocol.instruct.request import ChatCompletionRequest


tokenizer = MistralTokenizer.from_file(f"{mistral_models_path}/tokenizer.model.v3")
model = Transformer.from_folder(mistral_models_path)

completion_request = ChatCompletionRequest(messages=[UserMessage(content="Explain Machine Learning to me in a nutshell.")])

tokens = tokenizer.encode_chat_completion(completion_request).tokens

out_tokens, _ = generate([tokens], model, max_tokens=64, temperature=0.0, eos_id=tokenizer.instruct_tokenizer.tokenizer.eos_id)
result = tokenizer.instruct_tokenizer.tokenizer.decode(out_tokens[0])

print(result)

Function calling 

from mistral_common.protocol.instruct.tool_calls import Function, Tool
from mistral_inference.model import Transformer
from mistral_inference.generate import generate

from mistral_common.tokens.tokenizers.mistral import MistralTokenizer
from mistral_common.protocol.instruct.messages import UserMessage
from mistral_common.protocol.instruct.request import ChatCompletionRequest


tokenizer = MistralTokenizer.from_file(f"{mistral_models_path}/tokenizer.model.v3")
model = Transformer.from_folder(mistral_models_path)

completion_request = ChatCompletionRequest(
    tools=[
        Tool(
            function=Function(
                name="get_current_weather",
                description="Get the current weather",
                parameters={
                    "type": "object",
                    "properties": {
                        "location": {
                            "type": "string",
                            "description": "The city and state, e.g. San Francisco, CA",
                        },
                        "format": {
                            "type": "string",
                            "enum": ["celsius", "fahrenheit"],
                            "description": "The temperature unit to use. Infer this from the users location.",
                        },
                    },
                    "required": ["location", "format"],
                },
            )
        )
    ],
    messages=[
        UserMessage(content="What's the weather like today in Paris?"),
        ],
)

tokens = tokenizer.encode_chat_completion(completion_request).tokens

out_tokens, _ = generate([tokens], model, max_tokens=64, temperature=0.0, eos_id=tokenizer.instruct_tokenizer.tokenizer.eos_id)
result = tokenizer.instruct_tokenizer.tokenizer.decode(out_tokens[0])

print(result)

Generate with transformers

If you want to use Hugging Face transformers to generate text, you can do something like this.

from transformers import pipeline

messages = [
    {"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"},
    {"role": "user", "content": "Who are you?"},
]
chatbot = pipeline("text-generation", model="mistralai/Mistral-7B-Instruct-v0.3")
chatbot(messages)

Limitations

The Mistral 7B Instruct model is a quick demonstration that the base model can be easily fine-tuned to achieve compelling performance. It does not have any moderation mechanisms. We're looking forward to engaging with the community on ways to make the model finely respect guardrails, allowing for deployment in environments requiring moderated outputs.

The Mistral AI Team

Albert Jiang, Alexandre Sablayrolles, Alexis Tacnet, Antoine Roux, Arthur Mensch, Audrey Herblin-Stoop, Baptiste Bout, Baudouin de Monicault, Blanche Savary, Bam4d, Caroline Feldman, Devendra Singh Chaplot, Diego de las Casas, Eleonore Arcelin, Emma Bou Hanna, Etienne Metzger, Gianna Lengyel, Guillaume Bour, Guillaume Lample, Harizo Rajaona, Jean-Malo Delignon, Jia Li, Justus Murke, Louis Martin, Louis Ternon, Lucile Saulnier, Lélio Renard Lavaud, Margaret Jennings, Marie Pellat, Marie Torelli, Marie-Anne Lachaux, Nicolas Schuhl, Patrick von Platen, Pierre Stock, Sandeep Subramanian, Sophia Yang, Szymon Antoniak, Teven Le Scao, Thibaut Lavril, Timothée Lacroix, Théophile Gervet, Thomas Wang, Valera Nemychnikova, William El Sayed, William Marshall