TinyCast weights in GGUF form (tinycast.gguf) and their Rust port (crates/sky-infer/src/cast/tinycast.rs) tinycast.gguf is a format conversion of the official TinyCast weights (https://huggingface.co/raws-labs/tinycast, model.safetensors + config.json): the same tensors, stored as GGUF by `sky-cast convert-tinycast`, with the license, organization and source URL in its metadata. The shared feed-forward weights are stored once. No weight was changed. TinyCast: Steinhauser, RAWS Labs, arXiv 2608.15767, https://github.com/raws-labs/tinycast, Apache License, Version 2.0 (http://www.apache.org/licenses/LICENSE-2.0). Its NOTICE follows, unchanged. ======================================================================== TinyCast Copyright 2026 RAWS Labs This product is licensed under the Apache License, Version 2.0. ------------------------------------------------------------------------ Third-party attributions ------------------------------------------------------------------------ 1. GIFT-Eval benchmark reference artefacts Files: tinycast/reference/dataset_properties.json, tinycast/reference/seasonal_naive.csv These are the per-dataset properties and the seasonal-naive baseline (used as the denominator of normalized MASE / nGMASE) from Salesforce AI Research's GIFT-Eval benchmark. Source: https://github.com/SalesforceAIResearch/gift-eval Copyright Salesforce, Inc. Licensed under the Apache License, Version 2.0. 2. GluonTS The evaluation and forecast-container APIs (gluonts.model.evaluate_model, gluonts.ev metrics, gluonts.model.forecast.QuantileForecast) are used as a runtime dependency. Source: https://github.com/awslabs/gluonts Licensed under the Apache License, Version 2.0. 3. Autoregressive-rollout predictor File: tinycast/predictor.py The autoregressive-rollout decoding, flip-invariance symmetrization and NaN-imputation logic are adapted from the Reverso project's GIFT-Eval predictor (example/eval_gift.py). Source: https://github.com/shinfxh/reverso Licensed under the MIT License: Copyright (c) 2026 Reverso Authors Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. The GIFT-Eval evaluation protocol (dataset/term layout, metric definitions, period-alignment convention) follows the public GIFT-Eval benchmark.