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]:]))

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
```

Sampler Settings?

by pegasus912 - opened 1 day ago

Discussion

pegasus912

1 day ago

I apologize if I missed it, but what are the recommended sampler settings?

Thanks!

qenme

1 day ago

I apologize if I missed it, but what are the recommended sampler settings?

Thanks!

I saw this in the generation_config.json. Please note though this model is acting very strange...

{
"bos_token_id": 2,
"do_sample": true,
"eos_token_id": [
2,
24
],
"pad_token_id": 9,
"temperature": 1.0,
"top_p": 1.0,
"min_p": 0.0,
"top_k": 20,
"tool_call_parser": "poolside_v1",
"reasoning_parser": "poolside_v1",
"default_chat_template_kwargs": {
"enable_thinking": true
},
"speculative_config": {
"method": "dflash",
"source": "huggingface",
"model": "poolside/Laguna-S-2.1-DFlash",
"num_speculative_tokens": 15
}
}

pegasus912

1 day ago

Interesting. I'm not getting good results at all with those samplers. I've been tweaking here and there but so far Temp 1 and min-p 0.1 seems to be decent. Still honing it though.

joerowell

Poolside org 1 day ago

Hi guys! Thanks for the comment.

Just to double check: are you struggling with making the BF16 variant work, or some other variant?

pegasus912

1 day ago

Q5XL unsloth quant with unquantized KV Cache. I also tried the Q4XL quant and didn’t notice a difference.

joerowell

Poolside org 1 day ago

Thank you! I'll take a look in the coming days; those quants aren't officially ours, but it's still something that we'll investigate.

pegasus912

1 day ago

I did originally try the q4_m official quant and I didn’t notice a difference. I’ll go back to that and try some more tests.

joerowell

Poolside org 1 day ago

Just a note; we'll have a new GGUF for q4_m tomorrow. We noticed some degradations with it internally after further testing. I'll reply here once it's been uploaded :)

pegasus912

1 day ago

•

edited about 22 hours ago

I've spent a good bit of time narrowing down some good settings and this is what I think is a good balance of precision and creativity: --temp 1--top-k 50

Tested with Laguna-S-2.1-UD-Q5_K_XL
Hopefully this helps! (Updated with better sampler settings)

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