Instructions to use poolside/Laguna-S-2.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use poolside/Laguna-S-2.1 with Transformers:

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

pipe = pipeline("text-generation", model="poolside/Laguna-S-2.1", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("poolside/Laguna-S-2.1", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("poolside/Laguna-S-2.1", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use poolside/Laguna-S-2.1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "poolside/Laguna-S-2.1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "poolside/Laguna-S-2.1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/poolside/Laguna-S-2.1

SGLang

How to use poolside/Laguna-S-2.1 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 "poolside/Laguna-S-2.1" \
    --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": "poolside/Laguna-S-2.1",
		"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 "poolside/Laguna-S-2.1" \
        --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": "poolside/Laguna-S-2.1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use poolside/Laguna-S-2.1 with Docker Model Runner:
```
docker model run hf.co/poolside/Laguna-S-2.1
```

Tool call format not being parsed correctly by Cline (VS Code extension)

#24

by sunislee - opened 10 days ago

Discussion

sunislee

10 days ago

Hi, I'm running the GGUF version (UD-Q4_K_XL) locally via LM Studio, and trying to use it with Cline (the agentic coding extension for VS Code).

The model correctly understands the task and attempts to call tools, but it outputs the tool call in this format:

list_filespath/Users/xxx

Cline doesn't recognize this as a structured tool call — it just prints the raw text, then the model repeats the same call again in the next turn (since it never gets a tool result back), and after a few retries Cline throws "repeated tool call failures."

Questions:

Is this the intended/standard tool-call output format for Laguna-S-2.1?
Is there a recommended chat template or client configuration that makes this compatible with Cline specifically?
I noticed a discussion (#22, "Very Impressed - 270/270 in my tool usage") mentioning successful tool usage — would appreciate knowing what client/setup was used there for comparison.

Thanks!

dergunovs

9 days ago

Same

jsnfly

3 days ago

I actually had this issue with the model hosted via llamacpp but NOT with the same model hosted via LMStudio (Q4_K_M).
Turns out cline used a much shorter system prompt for LMStudio, which the model could handle well. Have you tried using the "compact system prompt"? With that the model was quite usable through cline. Unfortunately this option doesn't seem to exist for llamacpp.

atalan

3 days ago

I am not sure if somebody still has this problem, but laguna was struggling a lot with CLINE, only printing XML formatted tool calls - while it worked just fine with OpenWebUI and other models worked well with CLINE.

I solved this by explicitly defining the model name (poolside/Laguna-S-2.1-NVFP4) in the cline "Open AI compatible" setup page. I usually leave that entry empty, since I am switching models quite a bit, but as soon as I specified the model like this, laguna was finally able to call tools just fine also in CLINE.

sunislee

2 days ago

FWIW, I've been running Laguna-S-2.1 through OpenCode (via the oh-my-opencode/OMO plugin) with LM Studio as the backend, and tool calls have been working correctly — structured JSON calls, no issue with XML/raw text output. Might be worth trying if you're stuck with Cline.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment