Instructions to use OpenOneRec/KSA-V2-4B-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use OpenOneRec/KSA-V2-4B-base with Transformers:

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

pipe = pipeline("text-generation", model="OpenOneRec/KSA-V2-4B-base", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("OpenOneRec/KSA-V2-4B-base", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("OpenOneRec/KSA-V2-4B-base", trust_remote_code=True, 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use OpenOneRec/KSA-V2-4B-base with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "OpenOneRec/KSA-V2-4B-base"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "OpenOneRec/KSA-V2-4B-base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/OpenOneRec/KSA-V2-4B-base

SGLang

How to use OpenOneRec/KSA-V2-4B-base 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 "OpenOneRec/KSA-V2-4B-base" \
    --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": "OpenOneRec/KSA-V2-4B-base",
		"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 "OpenOneRec/KSA-V2-4B-base" \
        --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": "OpenOneRec/KSA-V2-4B-base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use OpenOneRec/KSA-V2-4B-base with Docker Model Runner:
```
docker model run hf.co/OpenOneRec/KSA-V2-4B-base
```

KSA-V2-4B-base / indexer_step.py

OpenOneRec

Release KSA-V2-4B-base weights and inference runtime

a1111f3 verified 2 days ago

Raw History Blame Contribute Delete

1.26 kB

from contextvars import ContextVar
from functools import wraps

_INDEXER_STEP = ContextVar("indexer_step", default=None)


def shared_indexer_step(forward):
    @wraps(forward)
    def scoped(*args, **kwargs):
        token = _INDEXER_STEP.set({"positions": {}, "bounds": {}})
        try:
            return forward(*args, **kwargs)
        finally:
            _INDEXER_STEP.reset(token)
    return scoped


def shared_indexer_positions(source, kind, factory, *layout):
    scope = _INDEXER_STEP.get()
    if scope is None:
        return factory()
    key = (id(source), kind, *layout)
    if key not in scope["positions"]:
        # Retain the source, so tensor identity cannot be recycled in this step.
        scope["positions"][key] = (source, factory())
    return scope["positions"][key][1]


def _validated_position_bounds(position_ids):
    scope = _INDEXER_STEP.get()
    key = id(position_ids)
    if scope is not None and key in scope["bounds"]:
        return scope["bounds"][key][1]
    low = int(position_ids.min())
    if low < 0:
        raise ValueError("indexer position IDs must be non-negative")
    bounds = (low, int(position_ids.max()))
    if scope is not None:
        scope["bounds"][key] = (position_ids, bounds)
    return bounds

	from contextvars import ContextVar
	from functools import wraps

	_INDEXER_STEP = ContextVar("indexer_step", default=None)


	def shared_indexer_step(forward):
	@wraps(forward)
	def scoped(args, *kwargs):
	token = _INDEXER_STEP.set({"positions": {}, "bounds": {}})
	try:
	return forward(args, *kwargs)
	finally:
	_INDEXER_STEP.reset(token)
	return scoped


	def shared_indexer_positions(source, kind, factory, *layout):
	scope = _INDEXER_STEP.get()
	if scope is None:
	return factory()
	key = (id(source), kind, *layout)
	if key not in scope["positions"]:
	# Retain the source, so tensor identity cannot be recycled in this step.
	scope["positions"][key] = (source, factory())
	return scope["positions"][key][1]


	def _validated_position_bounds(position_ids):
	scope = _INDEXER_STEP.get()
	key = id(position_ids)
	if scope is not None and key in scope["bounds"]:
	return scope["bounds"][key][1]
	low = int(position_ids.min())
	if low < 0:
	raise ValueError("indexer position IDs must be non-negative")
	bounds = (low, int(position_ids.max()))
	if scope is not None:
	scope["bounds"][key] = (position_ids, bounds)
	return bounds