Time Series Forecasting
Transformers
Safetensors
patchtst
Generated from Trainer
time series
forecasting
pretrained models
foundation models
time series foundation models
time-series
Instructions to use ibm-granite/granite-timeseries-patchtst with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use ibm-granite/granite-timeseries-patchtst with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, PatchTSTForPrediction tokenizer = AutoTokenizer.from_pretrained("ibm-granite/granite-timeseries-patchtst") model = PatchTSTForPrediction.from_pretrained("ibm-granite/granite-timeseries-patchtst", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Download config.json from ibm-granite/granite-timeseries-patchtst: direct link, hf CLI and curl.
- Browser
- Download file 1.29 kB
-
https://huggingface.co/ibm-granite/granite-timeseries-patchtst/resolve/main/config.json
- Command line
-
hf download hf://ibm-granite/granite-timeseries-patchtst/config.json
-
curl -L -o config.json https://huggingface.co/ibm-granite/granite-timeseries-patchtst/resolve/main/config.json
1.29 kB
| { | |
| "_name_or_path": "namctin/patchtst_etth1_forecast", | |
| "activation_function": "gelu", | |
| "architectures": [ | |
| "PatchTSTForPrediction" | |
| ], | |
| "attention_dropout": 0.0, | |
| "bias": true, | |
| "channel_attention": false, | |
| "channel_consistent_masking": false, | |
| "context_length": 512, | |
| "d_model": 128, | |
| "distribution_output": "student_t", | |
| "do_mask_input": null, | |
| "dropout": 0.2, | |
| "ff_dropout": 0.0, | |
| "ffn_dim": 512, | |
| "head_dropout": 0.2, | |
| "init_std": 0.02, | |
| "loss": "mse", | |
| "mask_input": null, | |
| "mask_type": "random", | |
| "mask_value": 0, | |
| "model_type": "patchtst", | |
| "norm_eps": 1e-05, | |
| "norm_type": "batchnorm", | |
| "num_attention_heads": 16, | |
| "num_forecast_mask_patches": [ | |
| 2 | |
| ], | |
| "num_hidden_layers": 3, | |
| "num_input_channels": 7, | |
| "num_parallel_samples": 100, | |
| "num_targets": 1, | |
| "output_range": null, | |
| "patch_length": 12, | |
| "patch_stride": 12, | |
| "path_dropout": 0.0, | |
| "pooling_type": null, | |
| "positional_dropout": 0.0, | |
| "positional_encoding_type": "sincos", | |
| "pre_norm": true, | |
| "prediction_length": 96, | |
| "random_mask_ratio": 0.5, | |
| "scaling": "std", | |
| "seed_number": null, | |
| "share_embedding": true, | |
| "share_projection": true, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.36.0.dev0", | |
| "unmasked_channel_indices": null, | |
| "use_cls_token": true | |
| } | |