Instructions to use yiqingguo/AquaCast with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use yiqingguo/AquaCast with Transformers:

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

pipe = pipeline("text-generation", model="yiqingguo/AquaCast")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("yiqingguo/AquaCast")
model = AutoModelForCausalLM.from_pretrained("yiqingguo/AquaCast")
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 yiqingguo/AquaCast with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/yiqingguo/AquaCast

SGLang

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

Docker Model Runner
How to use yiqingguo/AquaCast with Docker Model Runner:
```
docker model run hf.co/yiqingguo/AquaCast
```

TimeOmni Answer-Only SST Forecasting SFT

This model is a fine-tuned TimeOmni checkpoint for sea surface temperature forecasting. It was trained to emit a strict answer-only forecast:

<answer>[value_1,value_2,...,value_72]</answer>

Intended Use

The model is intended for demo and research use on event-aware hourly SST forecasting prompts with 240 historical observations and a 72-hour forecast horizon.

Evaluation

Accepted local evaluation used deterministic post-repair to enforce the scorer contract of exactly 72 numeric values inside <answer>...</answer>.

overall_score: 0.446593775025025
success_rate: 0.9954954954954955
valid_score: 0.347338606083459
valid_samples: 221 / 222

For the Hugging Face Space demo, forecasts are generated as a 10-sample ensemble. The demo plots the mean forecast and a +/- 1 std uncertainty band.

Training Procedure

Key hyperparameters:

learning_rate: 5e-6
num_epochs: 1
total_train_batch_size: 32
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.03
seed: 42

Framework versions from training:

Transformers 4.56.1
PyTorch 2.6.0+cu124
Datasets 4.0.0
Tokenizers 0.22.2

Downloads last month: 36

Safetensors

Model size

333k params

Tensor type

BF16

Model tree for yiqingguo/AquaCast

Base model

Qwen/Qwen2.5-7B

Finetuned

Qwen/Qwen2.5-7B-Instruct

Finetuned

anton-hugging/TimeOmni-1-7B

Finetuned

(1)

this model

Evaluation results

Overall score on Local SST forecasting eval
self-reported

0.447
Success rate on Local SST forecasting eval
self-reported

0.995