Image-Text-to-Text
Transformers
Safetensors
qwen3_5
clairic
cloudflare
systemone
multimodal
structured-output
information-extraction
custom-code
conversational
Instructions to use Cloudflare/clairic-flash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Cloudflare/clairic-flash with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="Cloudflare/clairic-flash") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("Cloudflare/clairic-flash") model = AutoModelForMultimodalLM.from_pretrained("Cloudflare/clairic-flash", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.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=256) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Cloudflare/clairic-flash with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Cloudflare/clairic-flash" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Cloudflare/clairic-flash", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/Cloudflare/clairic-flash
- SGLang
How to use Cloudflare/clairic-flash 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 "Cloudflare/clairic-flash" \ --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": "Cloudflare/clairic-flash", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'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 "Cloudflare/clairic-flash" \ --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": "Cloudflare/clairic-flash", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use Cloudflare/clairic-flash with Docker Model Runner:
docker model run hf.co/Cloudflare/clairic-flash
File size: 3,406 Bytes
dc7d8aa | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 | from __future__ import annotations
from collections.abc import Sequence
from typing import Any
def batch_image_data(media: Sequence[Any]) -> list[Any]:
if all((item is None for item in media)):
return list(media)
return [item if item is not None else [] for item in media]
def _grid_identity(grid: Any) -> Any:
if hasattr(grid, "tolist"):
grid = grid.tolist()
if isinstance(grid, (list, tuple)):
return tuple((_grid_identity(value) for value in grid))
if isinstance(grid, (str, int, float, bool, type(None))):
return grid
return ("opaque", id(grid))
def _media_identity(media: Any) -> tuple[Any, ...]:
if media is None:
return ("text",)
if not isinstance(media, list) or not media:
return ("opaque", id(media))
identities = []
for item in media:
if (
not isinstance(item, dict)
or item.get("format") != "processor_output"
or (not isinstance(item.get("clair_media_digest"), str))
or (item.get("pixel_values") is None)
):
return ("opaque", id(media))
identities.append(
(
item["clair_media_digest"],
id(item["pixel_values"]),
_grid_identity(item.get("image_grid_thw")),
)
)
return ("prepared", tuple(identities))
def _common_prefix(tasks: list[tuple[str, list[int], Any]]) -> list[int]:
(first, last) = (min((task[1] for task in tasks)), max((task[1] for task in tasks)))
length = next(
(i for (i, (left, right)) in enumerate(zip(first, last)) if left != right),
min(len(first), len(last)),
)
return list(first[:length])
def media_warmup_groups(
tasks: list[tuple[str, list[int], Any]],
*,
default_salt: str,
task_salts: dict[str, str] | None = None,
annotations: dict[str, dict[str, Any]] | None = None,
min_prefix_tokens: int = 128,
) -> list[dict[str, Any]]:
(task_salts, annotations) = (task_salts or {}, annotations or {})
partitions: dict[tuple[str, tuple[Any, ...]], dict[str, Any]] = {}
for task in tasks:
(name, _, media) = task
salt = task_salts.get(name, default_salt)
key = (salt, _media_identity(media))
partitions.setdefault(key, {"media": media, "salt": salt, "tasks": []})[
"tasks"
].append(task)
records = []
for partition in partitions.values():
members = partition["tasks"]
if len(members) < 2:
continue
prefix = _common_prefix(members)
if len(prefix) < min_prefix_tokens:
prefix = []
parents: dict[str, list[tuple[str, list[int], Any]]] = {}
for task in members:
parent = annotations.get(task[0], {}).get("parent")
if parent:
parents.setdefault(parent, []).append(task)
subgroups = [
_common_prefix(group) for group in parents.values() if len(group) > 1
]
subgroups = [group for group in subgroups if group and group != prefix]
if prefix or subgroups:
records.append(
{
"prefix_ids": prefix,
"media": partition["media"],
"salt": partition["salt"],
"group_prefix_ids": subgroups,
}
)
return records
|