Time Series Forecasting
Transformers
Safetensors
sundial
text-generation
time series
time-series
forecasting
foundation models
pretrained models
generative models
time series foundation models
custom_code
Instructions to use thuml/sundial-base-128m with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use thuml/sundial-base-128m with Transformers:
# Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("thuml/sundial-base-128m", trust_remote_code=True, dtype="auto") - Notebooks
- Google Colab
- Kaggle
关于最大预测长度
#3
by RoadToNowhere - opened
config.json中的"max_position_embeddings": 10000是否意味着 lookback_length + forecast_length <= 10000 ?
由于Sundial使用了RoPE,max_position_embeddings=10000意味着该模型的最大上下文长度(context length)为10000个token,参考模型patch token长度(16)换算为相应时间点数为160k