TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting.
Python
Commits on master
test: cover compile behavior; respect torch_compile kwarg in load_checkpoint
f83f134 Mathis Doutre committed 4 months ago
fix: compile forward instead of module in load_checkpoint
1e89ad1 Mathis Doutre committed 4 months ago
Fix model loading issues and forecast_naive slicing bug in TimesFM 2.5
2f1625c Rajat Sen committed 5 months ago
Merge pull request #398 from darkpowerxo/feat/peft-finetuning-pipeline-2.5
eacf761 Yichen Zhou committed 7 months ago
revert: drop PR #393 (xreg batch behavior) and PR #390 (SKILL.md link) per maintainer feedback
6ae67d4 darkpowerxo committed 7 months ago
refactor: replace custom PEFT pipeline with Transformers+PEFT example
caddef1 darkpowerxo committed 7 months ago
fix: improve PEFT device consistency and XReg output slicing
18d5eb2 darkpowerxo committed 7 months ago
docs: update README — replace 'under construction' with completed status
ad192b7 darkpowerxo committed 7 months ago
docs: fix swapped xreg_mode descriptions and typo in error message
54f5405 darkpowerxo committed 7 months ago
fix: respect batch_size in v1 data_loader when permute=False
1bb44d5 darkpowerxo committed 7 months ago
fix: per-input ridge regression to prevent data leakage in xreg
a63360a darkpowerxo committed 7 months ago
test: add unit tests for configs, torch layers, utils, and base utils
c10494a darkpowerxo committed 7 months ago
fix: correct 'complied' typo and replace print with logging
bc03b77 darkpowerxo committed 7 months ago
feat: add CLI entry-point and launch script for PEFT fine-tuning
a67eeb2 darkpowerxo committed 7 months ago
feat: add LoRA/DoRA adapter layers for TimesFM 2.5 (PyTorch)
7357458 darkpowerxo committed 7 months ago
chore: update .gitignore for egg-info, uv.lock, peft_checkpoints
aa2b17f darkpowerxo committed 7 months ago