Instructions to use Crystalcareai/CrystalMistral with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Crystalcareai/CrystalMistral with Transformers:

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

pipe = pipeline("text-generation", model="Crystalcareai/CrystalMistral")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Crystalcareai/CrystalMistral")
model = AutoModelForCausalLM.from_pretrained("Crystalcareai/CrystalMistral", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Crystalcareai/CrystalMistral with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Crystalcareai/CrystalMistral"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Crystalcareai/CrystalMistral",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Crystalcareai/CrystalMistral

SGLang

How to use Crystalcareai/CrystalMistral 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 "Crystalcareai/CrystalMistral" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Crystalcareai/CrystalMistral",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "Crystalcareai/CrystalMistral" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Crystalcareai/CrystalMistral",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Crystalcareai/CrystalMistral with Docker Model Runner:
```
docker model run hf.co/Crystalcareai/CrystalMistral
```

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

CrystalMistral: A Potent Language Model Fine-Tuned on a Curated Blend of Data

Overview

CrystalMistral is a highly refined language model derived from the esteemed Mistral-7B architecture. To unlock its exceptional capabilities, CrystalMistral underwent meticulous fine-tuning on a meticulously curated dataset comprised of:

Evol-Instruct: A rich dataset emphasizing instruction following and task completion, fostering CrystalMistral's ability to accurately execute complex commands. Airoboros: This extensive dataset is geared towards open-ended dialogue and generation, sharpening CrystalMistral's conversational aptitude and creativity. OpenOrca: A dataset specializing in code generation and understanding, significantly reinforcing CrystalMistral's programming prowess. Additional GPT-4 Synthetic Data: Incorporation of curated GPT-4 synthetic data further amplifies CrystalMistral's reasoning abilities and factual knowledge. Strengths

CrystalMistral exhibits a remarkable command of the following domains:

Instruction Following: Effectively interprets and carries out detailed instructions, demonstrating proficiency in task completion. Dialogue and Text Generation: Engages in fluid and nuanced conversations, offering creative and compelling text generation capabilities. Coding: Exhibits advanced understanding of code, capable of generating functional code, translating between languages, and offering explanations.

Future Development

The CrystalMistral project endeavors to:

Expert Fine-Tuning: Explore additional fine-tuning with datasets specializing in specific areas (e.g., scientific literature, legal documents) to create targeted variants of CrystalMistral. Mixture of Experts (MoE): Transition to a 4x MoE architecture, enabling CrystalMistral to dynamically specialize in distinct tasks, significantly amplifying its efficiency and potential.

Downloads last month: 10

Safetensors

Model size

7B params

Tensor type

F16

Model tree for Crystalcareai/CrystalMistral

Finetunes

1 model