Instructions to use BangorAI/Mistral-7B-Cymraeg-Welsh-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use BangorAI/Mistral-7B-Cymraeg-Welsh-v2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="BangorAI/Mistral-7B-Cymraeg-Welsh-v2")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("BangorAI/Mistral-7B-Cymraeg-Welsh-v2")
model = AutoModelForCausalLM.from_pretrained("BangorAI/Mistral-7B-Cymraeg-Welsh-v2")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use BangorAI/Mistral-7B-Cymraeg-Welsh-v2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "BangorAI/Mistral-7B-Cymraeg-Welsh-v2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BangorAI/Mistral-7B-Cymraeg-Welsh-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/BangorAI/Mistral-7B-Cymraeg-Welsh-v2

SGLang

How to use BangorAI/Mistral-7B-Cymraeg-Welsh-v2 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "BangorAI/Mistral-7B-Cymraeg-Welsh-v2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BangorAI/Mistral-7B-Cymraeg-Welsh-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "BangorAI/Mistral-7B-Cymraeg-Welsh-v2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BangorAI/Mistral-7B-Cymraeg-Welsh-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use BangorAI/Mistral-7B-Cymraeg-Welsh-v2 with Docker Model Runner:
```
docker model run hf.co/BangorAI/Mistral-7B-Cymraeg-Welsh-v2
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Mistral-7B-Cymraeg-Welsh-v2

This is a bilingual Mistral chat / instruct model trained in both English and Welsh languages.

The model is based on BangorAI/mistral-7b-cy-epoch-2 which is a continual pre-training of the Mistral-7B-v0.1 model with Welsh data from the allenai/MADLAD-400 dataset for 2 epochs.

The model was then fine-tuned using the yahma/alpaca-cleaned dataset in both Welsh and English languages, also for 2 epochs.

Demo

An online demo of the model can be found at https://demo.bangor.ai

It's an experimental LLM, so don't take any response from the model seriously or as factually correct. You are responsible for any output you generate.

Format

The LLM uses the Llama-2 format for its prompts:

<s>[INST] <<SYS>>
{{ system_prompt }}
<</SYS>>

{{ user_message }} [/INST]

The language of the system prompt will guide the LLM as to which language it should respond in. For example, in English:

<s>[INST] <<SYS>>
You are a helpful assistant that responds truthfully, logically and in detail. Answer in English.
<</SYS>>

{{ user_message }} [/INST]

Similarily, for responses in Welsh:

<s>[INST] <<SYS>>
Rydych chi'n gynorthwydd cymwynasgar sy'n barod i ateb unrhyw gwestiwn yn ffyddlon. Ymatebwch i gwestiwn y defnyddiwr yn llawn a gyda ffeithiau cywir yn y Gymraeg.
<</SYS>>

{{ user_message }} [/INST]

Mistral-7B-Cymraeg-Welsh-v2

Mae hwn yn fodel sgwrsio / cyfawryddo Mistral dwyieithog wedi'i hyfforddi yn y Gymraeg a'r Saesneg.

Mae'r model yn seiliedig ar BangorAI/mistral-7b-cy-epoch-2 sy'n rhaghyfforddiant parhaus o fodel Mistral-7B-v0.1 gyda data allenai/MADLAD-400 ar gyfer 2 epoch.

Cafodd y model hyfforddiant cywrian pellach gan ddefnyddio'r dataset yahma/alpaca-cleaned yn Gymraeg a Saesneg, hefyd am 2 epoch.

Demo Byw

Mae fersiwn o'r model i'w gael yma am sgwrs: https://demo.bangor.ai.

LLM arbrofol ydyw, felly peidiwch a chymeryd unrhyw ymateb gan y model o ddifri.

Fformat Sgwrs

Mae iaith y "system prompt" yn arwain yr LLM i ymateb yn y Gymraeg neu'r Saesneg. Er enghraifft, ar gyfer y Gymraeg:

<s>[INST] <<SYS>>
Rydych chi'n gynorthwydd cymwynasgar sy'n barod i ateb unrhyw gwestiwn yn ffyddlon. Ymatebwch i gwestiwn y defnyddiwr yn llawn a gyda ffeithiau cywir yn y Gymraeg.
<</SYS>>

{{ user_message }} [/INST]

Yn yr un modd, ar gyfer atebion yn Saesneg:

<s>[INST] <<SYS>>
You are a helpful assistant that responds truthfully, logically and in detail. Answer in English.
<</SYS>>

{{ user_message }} [/INST]

Contains information from allenai/MADLAD-400 which is made available under the ODC Attribution License.

Downloads last month: 5

Safetensors

Model size

7B params

Tensor type

BF16

Model tree for BangorAI/Mistral-7B-Cymraeg-Welsh-v2

Quantizations

1 model

BangorAI
/

Mistral-7B-Cymraeg-Welsh-v2

Mistral-7B-Cymraeg-Welsh-v2

Demo

Format

Mistral-7B-Cymraeg-Welsh-v2

Demo Byw

Fformat Sgwrs

Model tree for BangorAI/Mistral-7B-Cymraeg-Welsh-v2

Datasets used to train BangorAI/Mistral-7B-Cymraeg-Welsh-v2