How to use FinText/TimesFM_20M_2003_US with Transformers:
# Load model directly from transformers import AutoTokenizer, TimesFMForHF tokenizer = AutoTokenizer.from_pretrained("FinText/TimesFM_20M_2003_US") model = TimesFMForHF.from_pretrained("FinText/TimesFM_20M_2003_US")
This is the Time Series Foundation Model (TSFM), pre-trained on U.S. financial time series data up to the year 2003 using the TimesFM architecture (20M). The dataset spans from 1990–2003 and includes all U.S. excess return data.
🔗 Related Links
Files info
