Time Series Forecasting
Transformers
Safetensors
fela-pdm
feature-extraction
fela
fourier-neural-operator
fno
cpu
on-device
predictive-maintenance
time-series
anomaly-detection
custom_code
Instructions to use lowdown-labs/fela-pdm with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use lowdown-labs/fela-pdm with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("lowdown-labs/fela-pdm", trust_remote_code=True, dtype="auto") - Notebooks
- Google Colab
- Kaggle
| { | |
| "model_type": "fela-pdm", | |
| "library_name": "pytorch", | |
| "architectures": [ | |
| "FelaPdmModel" | |
| ], | |
| "auto_map": { | |
| "AutoConfig": "configuration_pdm.FelaPdmConfig", | |
| "AutoModel": "modeling_pdm.FelaPdmModel" | |
| }, | |
| "default_variant": "cmapss_FD001", | |
| "note": "Five trained heads ship, one safetensors file per head (cmapss_FD001, cmapss_FD002, cmapss_FD003, cmapss_FD004, cwru). Pick variants[<name>] to match the head; load_model(dir, variant=<name>) resolves <name>.safetensors.", | |
| "variants": { | |
| "cmapss_FD001": { | |
| "task": "rul", | |
| "in_channels": 14, | |
| "patch": 1, | |
| "n_embd": 64, | |
| "n_layer": 4, | |
| "n_head": 4, | |
| "fno_modes": 32, | |
| "gla_chunk": 32, | |
| "ffn_hidden": 128, | |
| "dropout": 0.0, | |
| "use_gdn": false, | |
| "gdn_every": 4, | |
| "n_classes": 0, | |
| "rul_head": true, | |
| "seq_len": 30, | |
| "rul_cap": 125, | |
| "input_shape": [1, 30, 14], | |
| "input_desc": "30 cycles of 14 informative C-MAPSS sensors, min max normalized on training stats" | |
| }, | |
| "cmapss_FD002": { | |
| "task": "rul", | |
| "in_channels": 14, | |
| "patch": 1, | |
| "n_embd": 64, | |
| "n_layer": 4, | |
| "n_head": 4, | |
| "fno_modes": 32, | |
| "gla_chunk": 32, | |
| "ffn_hidden": 128, | |
| "dropout": 0.0, | |
| "use_gdn": false, | |
| "gdn_every": 4, | |
| "n_classes": 0, | |
| "rul_head": true, | |
| "seq_len": 30, | |
| "rul_cap": 125, | |
| "input_shape": [1, 30, 14], | |
| "input_desc": "30 cycles of 14 informative C-MAPSS sensors, min max normalized on training stats" | |
| }, | |
| "cmapss_FD003": { | |
| "task": "rul", | |
| "in_channels": 14, | |
| "patch": 1, | |
| "n_embd": 64, | |
| "n_layer": 4, | |
| "n_head": 4, | |
| "fno_modes": 32, | |
| "gla_chunk": 32, | |
| "ffn_hidden": 128, | |
| "dropout": 0.0, | |
| "use_gdn": false, | |
| "gdn_every": 4, | |
| "n_classes": 0, | |
| "rul_head": true, | |
| "seq_len": 30, | |
| "rul_cap": 125, | |
| "input_shape": [1, 30, 14], | |
| "input_desc": "30 cycles of 14 informative C-MAPSS sensors, min max normalized on training stats" | |
| }, | |
| "cmapss_FD004": { | |
| "task": "rul", | |
| "in_channels": 14, | |
| "patch": 1, | |
| "n_embd": 64, | |
| "n_layer": 4, | |
| "n_head": 4, | |
| "fno_modes": 32, | |
| "gla_chunk": 32, | |
| "ffn_hidden": 128, | |
| "dropout": 0.0, | |
| "use_gdn": false, | |
| "gdn_every": 4, | |
| "n_classes": 0, | |
| "rul_head": true, | |
| "seq_len": 30, | |
| "rul_cap": 125, | |
| "input_shape": [1, 30, 14], | |
| "input_desc": "30 cycles of 14 informative C-MAPSS sensors, min max normalized on training stats" | |
| }, | |
| "cwru": { | |
| "task": "cls", | |
| "in_channels": 1, | |
| "patch": 4, | |
| "n_embd": 64, | |
| "n_layer": 4, | |
| "n_head": 4, | |
| "fno_modes": 64, | |
| "gla_chunk": 32, | |
| "ffn_hidden": 128, | |
| "dropout": 0.0, | |
| "use_gdn": false, | |
| "gdn_every": 4, | |
| "n_classes": 10, | |
| "rul_head": false, | |
| "seq_len": 2048, | |
| "input_shape": [1, 2048, 1], | |
| "input_desc": "2048 raw vibration samples at 12 kHz, single channel, per signal standardized" | |
| } | |
| } | |
| } | |