{"id":50335640,"url":"https://github.com/splunk/cisco-time-series-model","last_synced_at":"2026-05-29T13:30:32.047Z","repository":{"id":326030898,"uuid":"1100081305","full_name":"splunk/cisco-time-series-model","owner":"splunk","description":"Cisco Time Series Model is a continued pretrained time series forecasting model developed by Cisco.","archived":false,"fork":false,"pushed_at":"2026-04-24T07:11:12.000Z","size":14466,"stargazers_count":26,"open_issues_count":1,"forks_count":3,"subscribers_count":2,"default_branch":"main","last_synced_at":"2026-04-24T08:10:16.455Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/splunk.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2025-11-19T20:16:54.000Z","updated_at":"2026-04-24T07:11:16.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/splunk/cisco-time-series-model","commit_stats":null,"previous_names":["splunk/cisco-time-series-model"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/splunk/cisco-time-series-model","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/splunk%2Fcisco-time-series-model","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/splunk%2Fcisco-time-series-model/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/splunk%2Fcisco-time-series-model/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/splunk%2Fcisco-time-series-model/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/splunk","download_url":"https://codeload.github.com/splunk/cisco-time-series-model/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/splunk%2Fcisco-time-series-model/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":33655440,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-05-29T02:00:06.066Z","response_time":107,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-05-29T13:30:31.206Z","updated_at":"2026-05-29T13:30:32.036Z","avatar_url":"https://github.com/splunk.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Cisco Time Series Model 1.0\n\n[![arXiv technical report](https://img.shields.io/static/v1?label=Technical-Report\u0026message=2511.19841\u0026color=B31B1B\u0026logo=arXiv)](https://arxiv.org/abs/2511.19841)\n[![huggingface](https://img.shields.io/badge/%F0%9F%A4%97%20HF-Model-FFD21E)](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0)\n[![License: Apache-2.0](https://img.shields.io/badge/License-Apache--2.0-green.svg)](https://opensource.org/licenses/Apache-2.0)\n\nThe Cisco Time Series Model is a foundation model trained to perform univariate zero-shot forecasting. Its core is a sequence of decoder-only transformer layers. It is architecturally inspired by the [TimesFM2.0 model](https://huggingface.co/google/timesfm-2.0-500m-pytorch), with multiresolution modifications aimed at efficient use of long context. It expects a multiresolution context (x\u003csub\u003ec\u003c/sub\u003e, x\u003csub\u003ef\u003c/sub\u003e), where the resolution (i.e., space between data points) of x\u003csub\u003ec\u003c/sub\u003e is 60 times the resolution of x\u003csub\u003ef\u003c/sub\u003e. Both x\u003csub\u003ec\u003c/sub\u003e and x\u003csub\u003ef\u003c/sub\u003e can have length up to 512. The input contexts should be aligned “on the right,” e.g., if x\u003csub\u003ef\u003c/sub\u003e consists of the 512 minutes terminating at 11:00AM on November 11, then x\u003csub\u003ec\u003c/sub\u003e should consist of the 512 hours terminating at the same time. The output is a forecast of 128 points, which should be interpreted at the finer resolution; and corresponding quantiles for these points.\n\nFor convenience, we also provide utilities for preparing a multiresolution context from a single resolution context (with length up to 512 x 60 = 30,720) directly.\n\n## Latest Release (250M) - [CTSM 1.0](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0)\n**Results and Improvements:**\n- Achieves **state-of-the-art performance in the Observability (O11y) domain** compared to leading models.\n  - **1-minute resolution:** Outperforms the second-best benchmarked model by **16.12%** in MASE score.\n  - **5-minute resolution:** Outperforms the second-best benchmarked model by **12.42%** in MASE score.\n- Improves **GIFT-EVAL (public) Benchmark MASE score by 8.57%** compared to our [previous release](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0-preview).\u003cbr\u003e\n\n**Key improvements over our November 2025 release ([`cisco-time-series-model-1.0-preview`](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0-preview)):**\n- Trained **from scratch** (NO continued pretraining (CPT) from TimesFM weights).\n- Uses **2× more internal observability (O11y) data** in the training mixture.\n- Reduces size from **~500M → ~250M parameters**.\n- Adds **Short-Context Training**, **Bidirectional Coarse Attention**, and **RoPE** for better robustness overall.\n\n\n## Model Architecture and Training Details\n\u003cfigure\u003e\n  \u003cimg src=\"https://raw.githubusercontent.com/splunk/cisco-time-series-model/main/1.0/images/ctsm-rc-1_0.png\" alt=\"Multiresolution model architecture\"\u003e\n  \u003cfigcaption\u003e\u003cem\u003eArchitecture diagram illustrating our novel additions.\u003c/em\u003e\u003c/figcaption\u003e\n\u003c/figure\u003e\n\nOur latest [Cisco Time Series Model 1.0](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0) is a 250M parameter model that is trained from scratch and uses 25 transformer layers (vs 50 in [`1.0-preview`](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0-preview)). This checkpoint leverages the architectural changes of expanded quantiles (15 quantile outputs, 0.01-0.99), rotary positional embeddings (RoPE), special token, separate resolution embeddings for each of coarse context / fine context / special token, and bidirectional attention over the coarse-resolution context. \n\n### Standard Causal Attention vs Bidirectional Coarse Attention (Ours):\n\u003cfigure\u003e\n  \u003cimg src=\"https://raw.githubusercontent.com/splunk/cisco-time-series-model/main/1.0/images/bidirectional_coarse_attention.png\" alt=\"Bidirectional coarse attention illustration\"\u003e\n  \u003cfigcaption\u003e\u003cem\u003eA representation of bidirectional attention over the coarse context, where the model attends to both left and right context within the coarse context, while still attending causally to the fine context.\u003c/em\u003e\u003c/figcaption\u003e\n\u003c/figure\u003e\n\n### Example Visualization of Multiresolution Time Series Input to the Model\n\u003cfigure\u003e\n  \u003cimg src=\"https://raw.githubusercontent.com/splunk/cisco-time-series-model/main/1.0/images/multi_resolution_time_series_example.png\" alt=\"Multiresolution time series example with padded 1-hour context\"\u003e\n  \u003cfigcaption\u003e\u003cem\u003eMultiresolution time series example with padded 1-hour context.\u003c/em\u003e\u003c/figcaption\u003e\n\u003c/figure\u003e\n\n### Training Datasets\nThe distribution of the training corpus is as follows:\n- Internal Splunk Observability Cloud metric time series (upweighted ~2× vs the previous release) datasets:\n  - (1-hour, 1-minute) resolution: **46.2%**\n  - (5-hour, 5-minute) resolution: **21.8%**\n- Public datasets:\n  - [GIFT-EVAL](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain) : **19.5%**\n  - [Chronos](https://huggingface.co/datasets/autogluon/chronos_datasets) : **3%**\n- Synthetic dataset: **9.5%**\n\n## Technical Report\n- A detailed technical report is available on [arXiv](https://arxiv.org/abs/2511.19841) (also available locally [here](https://github.com/splunk/cisco-time-series-model/blob/main/1.0-preview/technical_report/Cisco-Time-Series-Model-Technical-Report.pdf)).\u003cbr\u003e\n\n**NOTE:** The report focuses on our November 2025 release ([`1.0-preview`](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0-preview)), so some training details differ from our latest release. An updated technical report with details on the latest release will be made available in the near future.\n\n## Benchmark Results\n\n### \u003e Observability Data\nWith reference to the training and validation sets, these time series are both out-of-domain and in-the-future. We apply curation rules similar to those described in Section 3 (of the [technical report](https://arxiv.org/abs/2511.19841)) to ensure a diverse and high-quality benchmark. \u003cbr\u003e\nWe report metrics (models sorted by MASE scores, lower is better) on both 1-minute and 5-minute resolution data (with coarse contexts at 1-hour and 5-hour resolution, respectively). For observability data, error metrics are computed per horizon, then all horizons are aggregated via arithmetic mean. This quantity is normalized by a similar computation using a naive baseline which simply forecasts the final value in the context for all time steps in the horizon (as no natural seasonality is available). \u003cbr\u003e\n\nAn input context of 1024 length is used with a forecast horizon of 128 length for all models. Metrics of MSIS and CRPS are both computed using 9 quantiles (0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9) to keep them consistent with the [GIFT-EVAL](https://github.com/SalesforceAIResearch/gift-eval) Benchmark.\n\n#### 1. `1-minute` Resolution data\n| MODEL | MSE | MAE | MASE | sMAPE | MSIS | CRPS |\n|:---|:---|:---|:---|:---|:---|:---|\n| [`CTSM-1.0 (Ours)`](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0) | `0.196` | `0.366` | `0.562` | `0.923` | `0.164` | `0.295` |\n| [`Toto_Open_Base_1.0`](https://huggingface.co/Datadog/Toto-Open-Base-1.0) | `0.270` | `0.436` | `0.670` | `0.967` | `0.303` | `0.357` |\n| [`TimesFM-2.5`](https://huggingface.co/google/timesfm-2.5-200m-pytorch) | `0.250` | `0.441` | `0.671` | `0.980` | `0.186` | `0.361` |\n| [`Chronos-2`](https://huggingface.co/amazon/chronos-2) | `0.242` | `0.421` | `0.674` | `0.955` | `0.192` | `0.341` |\n| [`TimesFM-2.0`](https://huggingface.co/google/timesfm-2.0-500m-pytorch) | `0.338` | `0.547` | `0.762` | `1.030` | `0.214` | `0.443` |\n| [`Chronos-bolt-base`](https://huggingface.co/amazon/chronos-bolt-base) | `0.439` | `0.617` | `0.810` | `1.137` | `0.321` | `0.524` |\n\n\n#### 2. `5-minute` Resolution data\n| MODEL | MSE | MAE | MASE | sMAPE | MSIS | CRPS |\n|:---|:---|:---|:---|:---|:---|:---|\n| [`CTSM-1.0 (Ours)`](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0) | `0.232` | `0.425` | `0.543` | `1.005` | `0.195` | `0.340` |\n| [`Chronos-2`](https://huggingface.co/amazon/chronos-2) | `0.282` | `0.475` | `0.620` | `1.040` | `0.176` | `0.383` |\n| [`TimesFM-2.5`](https://huggingface.co/google/timesfm-2.5-200m-pytorch) | `0.293` | `0.490` | `0.621` | `1.047` | `0.171` | `0.395` |\n| [`Toto_Open_Base_1.0`](https://huggingface.co/Datadog/Toto-Open-Base-1.0) | `0.309` | `0.491` | `0.647` | `1.045` | `0.267` | `0.395` |\n| [`TimesFM-2.0`](https://huggingface.co/google/timesfm-2.0-500m-pytorch) | `0.322` | `0.527` | `0.723` | `1.065` | `0.202` | `0.424` |\n| [`Chronos-bolt-base`](https://huggingface.co/amazon/chronos-bolt-base) | `0.386` | `0.567` | `0.737` | `1.114` | `0.290` | `0.481` |\n\n\n### \u003e [GIFT-EVAL](https://github.com/SalesforceAIResearch/gift-eval) Benchmark\nAll results below are computed on the entire [GIFT-EVAL](https://github.com/SalesforceAIResearch/gift-eval) benchmark, with forecast horizon capped at ≤ 128. Models are sorted by MASE scores (lower is better). We adopt the standard approach of normalizing errors per dataset before applying a geometric mean.\n\n#### 1. Using the entire time series context\n| MODEL | MSE | MAE | MASE | sMAPE | MSIS | CRPS | Test Leak. |\n|:---|:---|:---|:---|:---|:---|:---|:---|\n| [`Chronos-2`](https://huggingface.co/amazon/chronos-2) | `0.493` | `0.674` | `0.692` | `0.904` | `0.485` | `0.542` | `NO` |\n| [`TimesFM-2.5`](https://huggingface.co/google/timesfm-2.5-200m-pytorch) | `0.493` | `0.684` | `0.707` | `0.857` | `0.530` | `0.556` | `NO` |\n| [`CTSM-1.0 (Ours)`](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0) | `0.510` | `0.692` | `0.715` | `0.936` | `0.538` | `0.564` | `NO` |\n| [`TimesFM-2.0`](https://huggingface.co/google/timesfm-2.0-500m-pytorch) | `0.532` | `0.704` | `0.730` | `0.950` | `0.563` | `0.573` | `YES` |\n| [`Toto_Open_Base_1.0`](https://huggingface.co/Datadog/Toto-Open-Base-1.0) | `0.573` | `0.715` | `0.738` | `0.954` | `0.549` | `0.580` | `NO` |\n| [`Chronos-bolt-base`](https://huggingface.co/amazon/chronos-bolt-base) | `0.552` | `0.721` | `0.749` | `0.967` | `0.591` | `0.589` | `YES` |\n| [`CTSM-1.0-preview (Ours, prev.)`](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0-preview) | `0.585` | `0.757` | `0.782` | `0.999` | `0.584` | `0.615` | `YES` |\n\n\n#### 2. Using the time series context truncated to 1024 points\n| MODEL | MSE | MAE | MASE | sMAPE | MSIS | CRPS | Test Leak. |\n|:---|:---|:---|:---|:---|:---|:---|:---|\n| [`Chronos-2`](https://huggingface.co/amazon/chronos-2) | `0.471` | `0.659` | `0.678` | `0.887` | `0.464` | `0.531` | `NO` |\n| [`TimesFM-2.5`](https://huggingface.co/google/timesfm-2.5-200m-pytorch) | `0.464` | `0.667` | `0.684` | `0.851` | `0.494` | `0.540` | `NO` |\n| [`TimesFM-2.0`](https://huggingface.co/google/timesfm-2.0-500m-pytorch) | `0.521` | `0.697` | `0.707` | `0.933` | `0.529` | `0.566` | `YES` |\n| [`Toto_Open_Base_1.0`](https://huggingface.co/Datadog/Toto-Open-Base-1.0) | `0.528` | `0.691` | `0.715` | `0.932` | `0.526` | `0.559` | `NO` |\n| [`CTSM-1.0 (Ours)`](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0) | `0.510` | `0.692` | `0.715` | `0.936` | `0.538` | `0.564` | `NO` |\n| [`Chronos-bolt-base`](https://huggingface.co/amazon/chronos-bolt-base) | `0.562` | `0.722` | `0.747` | `0.967` | `0.597` | `0.592` | `YES` |\n| [`CTSM-1.0-preview (Ours, prev.)`](https://huggingface.co/cisco-ai/cisco-time-series-model-1.0-preview) | `0.585` | `0.757` | `0.782` | `0.999` | `0.584` | `0.615` | `YES` |\n\n\u003e ***Test Leak.:** `YES` indicates that the model's training data inadvertently includes portions of the [GIFT-EVAL](https://github.com/SalesforceAIResearch/gift-eval) test set.*\n\n## Usage notes\n- If the input time series is missing some values, imputation via last value is recommended; if the time series is naturally sparse and this leads to excessive imputation (e.g., more than 30% of values are imputed), the model's forecasts will deteriorate.\n- This release includes short-context training with 1/3rd of the training data sampled uniformly in the range of `[10, 511]` in the fine context. However, the model generally works better when more coarse resolution history is provided.\n\n## Dependencies and Installation\n\n### Setup Virtual Environment\nCreate a Python 3.11 (recommended) virtual environment and install the package using `uv`:\n```shell\n# Install uv if required\ncurl -LsSf https://astral.sh/uv/install.sh | sh\n\n# Create a virtual environment\nuv venv --python 3.11\n\n# Activate the environment\nsource .venv/bin/activate\n```\n\n### PyPI Package\nThe Cisco Time Series Model is available as a [PyPI package](https://pypi.org/project/cisco-tsm/) named `cisco-tsm`. You can install it using pip:\n```shell\nuv pip install cisco-tsm\n```\n\n### Manual Installation\n1.  Clone the repository:\n    ```shell\n    git clone https://github.com/splunk/cisco-time-series-model.git\n    cd cisco-time-series-model\n    ```\n\n2.  [Optional] Install your preferred `torch` backend based on your OS and accelerators (CPU, GPU, TPU or Apple Silicon):\n    - [Install PyTorch](https://pytorch.org/get-started/locally/).\n\n3. Install the package:\n    ```shell\n    uv pip install -e .\n    ```\n\n4.  _(Deprecated)_ If you want to use the `1.0-preview` model, change directory to `1.0-preview/` and use the corresponding package (follow its README for installation and usage):\n    ```shell\n    cd 1.0-preview/\n    ```\n\n## Example Code\n\n```python\nimport torch\nimport numpy as np\nfrom cisco_tsm import CiscoTsmMR, TimesFmHparams, TimesFmCheckpoint\n\nrng = np.random.default_rng(42)\n\n## Sample data\nT = 512 * 60\nhours = (T + 59) // 60\nk = np.arange(hours, dtype=np.float32)\nh = (80 + 0.1 * k) * (1 + 0.25 * np.sin(2 * np.pi * k / 24))\nt = np.arange(T, dtype=np.float32)\n\ninput_series_1 = h[(t // 60).astype(int)] * (1 + 0.05 * np.sin(2 * np.pi * t / 30)) + rng.normal(0, 0.4, size=T)\n\n# Hyperparameters\nhparams = TimesFmHparams(\n    num_layers=25, # For `1.0-preview`, use `num_layers=50`.\n    use_positional_embedding=False,\n    backend=\"gpu\" if torch.cuda.is_available() else \"cpu\",\n    quantiles=[0.01, 0.05, 0.1, 0.2, 0.25, 0.3, 0.4, 0.5, 0.6, 0.7, 0.75, 0.8, 0.9, 0.95, 0.99]\n)\n\nckpt = TimesFmCheckpoint(huggingface_repo_id=\"cisco-ai/cisco-time-series-model-1.0\") # For `1.0-preview`, use `huggingface_repo_id=\"cisco-ai/cisco-time-series-model-1.0-preview\"`.\n\nmodel = CiscoTsmMR(hparams=hparams, checkpoint=ckpt)\n\n# Model Inference\nforecast_preds = model.forecast(input_series_1, horizon_len=128)\n\n# Access forecast mean and quantiles of each series\nmean_forecast = forecast_preds[0]['mean'] # (128,)\nquantiles = forecast_preds[0]['quantiles'] # dict with keys as quantile levels (0.01, 0.05, ...., 0.95, 0.99) and values as (128,) numpy arrays\n\n# You can also forecast multiple series at once\nT = 25_000\nhours = (T + 59) // 60\nk = np.arange(hours, dtype=np.float32)\nh = 120 / (1 + np.exp(-0.01 * (k - 300))) + 10 * np.cos(2 * np.pi * k / (24*7))\nt = np.arange(T, dtype=np.float32)\ninput_series_2 = h[(t // 60).astype(int)] + 2 * np.sin(2 * np.pi * t / 60) + rng.normal(0, 0.5, size=T)\n\nmulti_series_forecasts = model.forecast([input_series_1, input_series_2], horizon_len=128)\n\n# Long horizon forecasting is also supported and can be invoked as follows\nlong_horizon_forecasts = model.forecast(input_series_1, horizon_len=240)\n\n```\n\n## Example Notebooks\nWe also provide a few Jupyter notebooks demonstrating how to use the Cisco Time Series Model for forecasting real-world time series data:\n- [CPU Utilization Forecasting](https://github.com/splunk/cisco-time-series-model/blob/main/1.0/notebooks/cpu_utilization_forecast.ipynb)\n- [Alerting Server Response Time Forecasting](https://github.com/splunk/cisco-time-series-model/blob/main/1.0/notebooks/alerting_server_responsetime.ipynb)\n- [Internet Traffic Forecasting](https://github.com/splunk/cisco-time-series-model/blob/main/1.0/notebooks/internet_traffic_forecast.ipynb)\n\n\u003cb\u003eNotebooks contributed by:\u003c/b\u003e Huaibo Zhao\n\n## Self-Hosting\n\nThe Cisco Deep Time Series Model (CDTSM) can be self-hosted as a FastAPI inference server,\nenabling on-premise or private-cloud deployments with full control over the runtime environment.\n\n\u003cb\u003eQuick start (Docker):\u003c/b\u003e\n\n```shell\ncd serve/\ncp .env-example .env   # set CDTSM_AUTH_TOKEN (user-defined token for authentication)\ndocker compose up --build\n```\n\n\u003cb\u003eQuick start (process-based):\u003c/b\u003e\n\n```shell\ncd serve/\ncp .env-example .env   # set CDTSM_AUTH_TOKEN (user-defined token for authentication)\nexport CDTSM_AUTH_TOKEN=\u003cyour-token\u003e\n\n# CPU-Only\nmake install-dev \u0026\u0026 make model-up\n\n# GPU-Accelerated\nmake install-dev-gpu \u0026\u0026 make model-up\n```\n\nThe [`serve/`](https://github.com/splunk/cisco-time-series-model/tree/main/serve/) directory provides:\n- \u003cb\u003eProcess-based hosting\u003c/b\u003e via a Makefile with CPU and GPU targets\n- \u003cb\u003eDocker-based hosting\u003c/b\u003e with CPU and GPU images, persistent model cache, and non-root container execution\n- Bearer token authentication\n- Health (`/health`) and readiness (`/ready`) probes\n- An AITK-compatible JSON API at `POST /cdtsm/v1/ai/infer`\n\nBoth CPU and GPU (NVIDIA CUDA) backends are supported. See the [serve/README.md](https://github.com/splunk/cisco-time-series-model/blob/main/serve/README.md) for full setup instructions, environment variables, API reference, and example requests.\u003cbr\u003e\n\n\u003cb\u003eSelf-Hosting contributed by:\u003c/b\u003e Udaya Prasad Vakalapudi\u003cbr\u003e\n\n## Citation\nIf you find Cisco Time Series Model useful for your research, please consider citing the associated technical report:\n```\n@misc{gou2025ciscotimeseriesmodel,\n      title={Cisco Time Series Model Technical Report}, \n      author={Liang Gou and Archit Khare and Praneet Pabolu and Prachi Patel and Joseph Ross and Hercy Shen and Yuhan and Song and Jingze Sun and Kristal Curtis and Vedant Dharnidharka and Abhinav Mathur and Hao Yang},\n      year={2025},\n      eprint={2511.19841},\n      archivePrefix={arXiv},\n      primaryClass={cs.LG},\n      url={https://arxiv.org/abs/2511.19841}, \n}\n```\n\n## Authors:\n- Liang Gou \\*\n- Archit Khare \\*\n- Praneet Pabolu \\*\n- Prachi Patel \\*\n- Joseph Ross \\*\n- Hercy Shen \\*‡\n- Yuhan (Ellen) Song \\*\n- Jingze Sun \\*\n- Kristal Curtis †\n- Vedant Dharnidharka †\n- Abhinav Mathur †\n- Hao Yang †\n\n\\* These authors contributed equally to the core development of this work, listed alphabetically by last name. \u003cbr\u003e\n† These authors contributed equally to supporting and extending this work, listed alphabetically by last name.\u003cbr\u003e\n‡ Hercy Shen contributed to this work while an intern at Splunk.\u003cbr\u003e\n\n## License\nThis project is licensed under the Apache-2.0 License. See the [LICENSE](https://github.com/splunk/cisco-time-series-model/blob/main/LICENSE) file for more details.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsplunk%2Fcisco-time-series-model","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsplunk%2Fcisco-time-series-model","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsplunk%2Fcisco-time-series-model/lists"}