# AI use cases for Chronos Forecasting (amazon-science/chronos-forecasting) Repo: [amazon-science/chronos-forecasting](https://github.com/amazon-science/chronos-forecasting) | Use cases: 3 | Page: https://dedupedai.com/use-cases/amazon-science-chronos-forecasting/ Chronos is Amazon Science's family of pretrained time series forecasting models. Chronos-2 forecasts univariate, multivariate and covariate informed tasks zero shot, Chronos-Bolt is a patch based variant the README describes as up to 250 times faster and 20 times more memory efficient than the original Chronos models of the same size, and the original Chronos models tokenize a series and score it with a language model architecture. One pip install gives you the whole family. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Forecast a series using the drivers you already know: https://dedupedai.com/use-cases/amazon-science-chronos-forecasting/forecast-with-covariates-using-chronos-2/ - Chart a forecast with its prediction interval: https://dedupedai.com/use-cases/amazon-science-chronos-forecasting/plot-a-chronos-forecast-against-the-truth/ - Choose the Chronos model that fits your budget: https://dedupedai.com/use-cases/amazon-science-chronos-forecasting/pick-a-chronos-model-size-for-your-workload/ ## How to forecast a series using the drivers you already know with Chronos Forecasting Repo: [amazon-science/chronos-forecasting](https://github.com/amazon-science/chronos-forecasting) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Run Chronos-2 on a DataFrame of history plus future covariate values to get quantile forecasts. **Who it is for:** An analyst forecasting prices or demand that depend on known future inputs. **The problem:** Most quick forecasts throw away the covariates you already have. **The outcome:** A prediction DataFrame with a median and the quantiles you asked for. ### Steps 1. Install it with `pip install chronos-forecasting` 2. Load the pipeline with Chronos2Pipeline.from_pretrained on amazon/chronos-2 as in the README 3. Read your history into context_df and your future covariate values into future_df 4. Call pipeline.predict_df with prediction_length, quantile_levels, id_column, timestamp_column and target ### Prompt to paste into your agent ```text Using amazon-science/chronos-forecasting, install it with `pip install chronos-forecasting`, forecast 24 steps of with Chronos2Pipeline.predict_df using my known future covariates, and return the 0.1, 0.5 and 0.9 quantiles. ``` ### Caveat The README example loads the model with device_map set to cuda, so adapt it if you have no GPU. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/amazon-science-chronos-forecasting/forecast-with-covariates-using-chronos-2/ ## How to chart a forecast with its prediction interval with Chronos Forecasting Repo: [amazon-science/chronos-forecasting](https://github.com/amazon-science/chronos-forecasting) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README Plot history, ground truth and the Chronos-2 forecast band to see where the model is confident. **Who it is for:** Anyone who has to show a forecast to other people. **The problem:** A table of quantiles does not tell you whether the forecast is sensible. **The outcome:** A chart with the history, the forecast and a shaded prediction interval. ### Steps 1. Install the package with `pip install chronos-forecasting` and matplotlib as the README notes 2. Produce pred_df with pipeline.predict_df and quantile_levels of 0.1, 0.5 and 0.9 3. Plot the last part of the history, the ground truth and pred_df["predictions"] as the README example does 4. Shade the interval with plt.fill_between across the 0.1 and 0.9 columns ### Prompt to paste into your agent ```text With amazon-science/chronos-forecasting, produce a Chronos-2 forecast for and save a chart showing history, ground truth and the 10 to 90 percent band. ``` ### Caveat The README plotting example uses a single series; group by id when you have many. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/amazon-science-chronos-forecasting/plot-a-chronos-forecast-against-the-truth/ ## How to choose the Chronos model that fits your budget with Chronos Forecasting Repo: [amazon-science/chronos-forecasting](https://github.com/amazon-science/chronos-forecasting) | Difficulty: Beginner | Time: 30 minutes | Basis: From the README Compare the listed model IDs and pick between Chronos-2, Chronos-Bolt and the original Chronos family. **Who it is for:** A team deciding what to run on a CPU box or a small GPU. **The problem:** Running the largest model everywhere wastes time and memory. **The outcome:** A model choice you can justify, with the parameter count written down. ### Steps 1. Read the Available Models table, which lists IDs from amazon/chronos-bolt-tiny at 9M up to amazon/chronos-t5-large at 710M 2. Note that the README describes Chronos-Bolt as up to 250 times faster and 20 times more memory efficient than the original Chronos models of the same size 3. Install the package with `pip install chronos-forecasting` and try the smallest candidate first 4. For production the README recommends deploying Chronos-2 to Amazon SageMaker through AutoGluon-Cloud or SageMaker JumpStart ### Prompt to paste into your agent ```text Using amazon-science/chronos-forecasting, compare the listed Chronos-2, Chronos-Bolt and Chronos-T5 model IDs for and recommend one with the reasoning. ``` ### Caveat The benchmark claims in the README are the authors' own. Risk, cost: The recommended production route is Amazon SageMaker, which is billed by AWS. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/amazon-science-chronos-forecasting/pick-a-chronos-model-size-for-your-workload/