Instructions to use levara/Devstral-Small-2-24B-TextOnly-FP8-Training with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use levara/Devstral-Small-2-24B-TextOnly-FP8-Training with Transformers:

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

pipe = pipeline("text-generation", model="levara/Devstral-Small-2-24B-TextOnly-FP8-Training")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("levara/Devstral-Small-2-24B-TextOnly-FP8-Training")
model = AutoModelForCausalLM.from_pretrained("levara/Devstral-Small-2-24B-TextOnly-FP8-Training")
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

vLLM

How to use levara/Devstral-Small-2-24B-TextOnly-FP8-Training with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "levara/Devstral-Small-2-24B-TextOnly-FP8-Training"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "levara/Devstral-Small-2-24B-TextOnly-FP8-Training",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/levara/Devstral-Small-2-24B-TextOnly-FP8-Training

SGLang

How to use levara/Devstral-Small-2-24B-TextOnly-FP8-Training 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 "levara/Devstral-Small-2-24B-TextOnly-FP8-Training" \
    --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": "levara/Devstral-Small-2-24B-TextOnly-FP8-Training",
		"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 "levara/Devstral-Small-2-24B-TextOnly-FP8-Training" \
        --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": "levara/Devstral-Small-2-24B-TextOnly-FP8-Training",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Studio new

How to use levara/Devstral-Small-2-24B-TextOnly-FP8-Training with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for levara/Devstral-Small-2-24B-TextOnly-FP8-Training to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for levara/Devstral-Small-2-24B-TextOnly-FP8-Training to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for levara/Devstral-Small-2-24B-TextOnly-FP8-Training to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="levara/Devstral-Small-2-24B-TextOnly-FP8-Training",
    max_seq_length=2048,
)

Docker Model Runner
How to use levara/Devstral-Small-2-24B-TextOnly-FP8-Training with Docker Model Runner:
```
docker model run hf.co/levara/Devstral-Small-2-24B-TextOnly-FP8-Training
```

Devstral-Small-2-24B TextOnly FP8 (Training)

Training-compatible variant of levara/Devstral-Small-2-24B-TextOnly-FP8 with Mistral/transformers-convention FP8 scale names.

Weight values are byte-for-byte identical to the serving checkpoint. Only the safetensors key names differ:

This repo (training)	Serving repo (vLLM)
`activation_scale`	`input_scale`
`weight_scale_inv`	`weight_scale`

Why two repos?

vLLM's TransformersForCausalLM backend registers FP8 parameters as input_scale/weight_scale and errors on other names. Transformers 5 and Unsloth expect activation_scale/weight_scale_inv. Neither tolerates the other's names.

Using this repo for LoRA training ensures the adapter trains against the true FP8 ground truth weights — the same values used at serving time. No dequant/re-quant mismatch.

Usage with Unsloth

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    "levara/Devstral-Small-2-24B-TextOnly-FP8-Training",
    max_seq_length=8192,
    load_in_4bit=False,
)

model = FastLanguageModel.get_peft_model(model, r=16, target_modules=[
    "q_proj", "k_proj", "v_proj", "o_proj",
    "gate_proj", "up_proj", "down_proj",
])

Serving

For vLLM serving, use the companion checkpoint: levara/Devstral-Small-2-24B-TextOnly-FP8

vllm serve levara/Devstral-Small-2-24B-TextOnly-FP8 \
    --tensor-parallel-size 2 \
    --max-model-len 32768 \
    --enable-lora \
    --lora-modules my-adapter=path/to/adapter

Model Details

Property	Value
Architecture	`Ministral3ForCausalLM`
Parameters	23.57B
Quantization	FP8 W8A8 static (`float8_e4m3fn`)
Layers	40
Hidden size	5120
Context length	393K tokens (YaRN RoPE)

Downloads last month: 362

Safetensors

Model size

24B params

Tensor type

BF16

F8_E4M3

Model tree for levara/Devstral-Small-2-24B-TextOnly-FP8-Training

Base model

mistralai/Mistral-Small-3.1-24B-Base-2503

Quantized

mistralai/Devstral-Small-2-24B-Instruct-2512

Adapter

(50)

this model