A model is still running

A model is still running or downloading. Leaving will stop the current operation. Stay on this page to let it finish.

Skip to content
LocalAI.toolsBeta

Join the waitlist

Get an email when the new apps and model runs launch.

No account needed.

TabICLv2 Regressor

Predict numbers and quantiles from example table rows; forecast regular time series.

Browser inferencePowered by WebAI · Private · Unlimited runs
About this model

Predict numbers and quantiles from example table rows; forecast regular time series.

Created by
SODA / Inria
Model released
Unknown
Source repository created
Feb 7, 2025
Model weights · FP16
58.5 MB

Best for

  • Explore predictions on held-out tabular data locally

Keep in mind

  • Measured on small fixed held-out tables; performance is not established for every domain, rare class, high-dimensional or large table. Predictions are not calibrated decisions and must not replace expert review.
  • Tables: at most100 features,4096rows,20000cells per table; classification2–10classes. Missing and categorical cells are tested in limited derived fixtures.
  • FP16 is stored-weight precision; neural computation is FP32. Conversion/preprocessing can differ from upstream Python; independent held-out quality gates passed.
  • Forecasting is unavailable in this browser adapter: forecast-web1.0.0 has no TabICLForecaster temporal feature pipeline. Use the native regressor for measured forecasting.
  • The pinned atomic estimator does not expose incremental predictions; streaming is unavailable.
  • Tested Chrome157 on Apple M3 Max, WASM and Metal WebGPU. Other browsers and GPU hardware remain unverified.

Processed on your deviceInference runs in this browser. Input/output sharing requires consent for each run. Basic run logs include device information and approximate connection location. Model and runtime files download from external hosts. Saved downloads may be cleared by your browser.

Sources & technical details
Original model
https://github.com/soda-inria/tabicl
Release verification
Unknown
Release evidence
Exact checkpoint filename contains20260212; filename alone is not verified public release evidence.
Source updated
Nov 12, 2026
Browser model files
bakshinghampalace/tabicl-regressor-v2-onnx
Files updated
Oct 14, 2026
Artifact revision
f7d19314f03770294c4fa3e019b28335dfa86986

Input

TEXT

Sample result

TabICLv2 Regressor

Settings

Add a chart of at most128rows while retaining the complete native output. Classification plots confidence, not class identity. Adds a small SVG rendering cost; all raw predictions remain unchanged.
Advanced settings 1
More ensemble members trade latency for prediction stability. More ensemble members trade latency for prediction stability.

Output

Your result will appear here

Add input or try the sample, then run the model.

Benchmark evidence

Saved worker evidence is separate from benchmarks run on your device. It does not guarantee speed or quality on every browser.

TabICLv2 Regressor

8 configurations tested
Passed8
Failed0
TestedNov 13, 2026
RuntimeONNX Runtime Web 1.30.0

Quality criteria

{ "frozenAt": "2026-11-13T22:43:16.191927+00:00", "scope": "Small real tabular fixtures, mixed/missing stress and synthetic/real forecasting; no broad domain or medical-use validation", "maximumInferenceMs": 10000, "repetitions": { "independentInitializations": 3, "warmInferences": 5, "warmupExcluded": 1 }, "negativeCases": [ "empty tables", "mismatched columns", "invalid labels", "unknown fields", "invalid ensemble size", "invalid forecasting timestamp", "invalid frequency" ], "concurrentActivityPolicy": "No other benchmark/build; background model backup may be active and is recorded." }

PrecisionDeviceResult
FP16GPU · WebGPUAcceleratedPassedModerate
FP32GPU · WebGPUAcceleratedPassedModerate
FP32GPU · WebGPUAcceleratedPassedModerate
FP16GPU · WebGPUAcceleratedPassedModerate
FP16CPU · WASMPassedVery slow
FP16CPU · WASMPassedVery slow
FP32CPU · WASMPassedVery slow
FP32CPU · WASMPassedVery slow
Test setup & sample details
Quality gate
{ "frozenAt": "2026-11-13T22:43:16.191927+00:00", "scope": "Small real tabular fixtures, mixed/missing stress and synthetic/real forecasting; no broad domain or medical-use validation", "maximumInferenceMs": 10000, "repetitions": { "independentInitializations": 3, "warmInferences": 5, "warmupExcluded": 1 }, "negativeCases": [ "empty tables", "mismatched columns", "invalid labels", "unknown fields", "invalid ensemble size", "invalid forecasting timestamp", "invalid frequency" ], "concurrentActivityPolicy": "No other benchmark/build; background model backup may be active and is recorded." }
Tested
2026-11-13T22:53:01.571Z
Environment
{ "hardware": "Apple M3 Max", "browser": "157.0.8103.51", "runtime": "forecast-web1.0.0 (local lifecycle/cache patches); ONNX Runtime Web1.30.0", "concurrentUploadActivity": "Fresh browser contexts; no other inference/build started by this task; existing background mirror service may be active", "scope": "Fresh isolated module worker for every precision/backend; local fixed fixtures and actual GPU adapter; chart enabled. Download included in loadMs. Single fresh timing per fixture; no statistical speed ranking." }
Runtime
ONNX Runtime Web 1.30.0
Artifact revision
f7d19314f03770294c4fa3e019b28335dfa86986
Fixture
{
  "file": "localai-desktop/fixtures/tabicl-v2/diabetes.json",
  "size": 146831,
  "sha1": "a49a93063c50e26104a1b21ddf011534d4bb6801",
  "sha256": "70b8ec8d0dc549b7bafa1674a8b68cdda48c4b330c6933f16ffdfe8438a05f5d",
  "url": "https://assets.localai.tools/samples/v1/files/70b8ec8d0dc549b7bafa1674a8b68cdda48c4b330c6933f16ffdfe8438a05f5d/diabetes.json"
}
Rerun when
artifact revision; runtime version; adapter implementation; fixture checksum; quality gate; backend behavior
Per-case quality & failure details
[
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "r2",
      "score": 0.5152164945596093,
      "minimum": 0.4
    },
    "runtime": {
      "loadMs": 5626.59500002861,
      "inferenceMs": 5172.539999961853,
      "totalMs": 10799.134999990463,
      "rowsPerSecond": 21.459476388934377,
      "grade": "usable",
      "engine": "ONNX Runtime Web",
      "version": "1.30.0",
      "precision": "fp16",
      "device": "wasm",
      "requestedDevice": "wasm",
      "fallback": false,
      "neuralExecuted": true
    }
  },
  {
    "precision": "fp16",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "r2",
      "score": 0.5084575808118101,
      "minimum": 0.3
    },
    "runtime": {
      "loadMs": 5626.59500002861,
      "inferenceMs": 5223.400000035763,
      "totalMs": 10849.995000064373,
      "rowsPerSecond": 21.25052647686182,
      "grade": "usable",
      "engine": "ONNX Runtime Web",
      "version": "1.30.0",
      "precision": "fp16",
      "device": "wasm",
      "requestedDevice": "wasm",
      "fallback": false,
      "neuralExecuted": true
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "r2",
      "score": 0.5152165574513738,
      "minimum": 0.4
    },
    "runtime": {
      "loadMs": 6683.865000009537,
      "inferenceMs": 1639.0799999833107,
      "totalMs": 8322.944999992847,
      "rowsPerSecond": 67.72091661244735,
      "grade": "usable",
      "engine": "ONNX Runtime Web",
      "version": "1.30.0",
      "precision": "fp16",
      "device": "webgpu",
      "requestedDevice": "webgpu",
      "fallback": false,
      "neuralExecuted": true
    }
  },
  {
    "precision": "fp16",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "r2",
      "score": 0.508457560917202,
      "minimum": 0.3
    },
    "runtime": {
      "loadMs": 6683.865000009537,
      "inferenceMs": 1272.0199999809265,
      "totalMs": 7955.884999990463,
      "rowsPerSecond": 87.26277888843289,
      "grade": "usable",
      "engine": "ONNX Runtime Web",
      "version": "1.30.0",
      "precision": "fp16",
      "device": "webgpu",
      "requestedDevice": "webgpu",
      "fallback": false,
      "neuralExecuted": true
    }
  },
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "r2",
      "score": 0.515235156970198,
      "minimum": 0.4
    },
    "runtime": {
      "loadMs": 7662.739999949932,
      "inferenceMs": 5294.949999988079,
      "totalMs": 12957.689999938011,
      "rowsPerSecond": 20.963370758977877,
      "grade": "usable",
      "engine": "ONNX Runtime Web",
      "version": "1.30.0",
      "precision": "fp32",
      "device": "wasm",
      "requestedDevice": "wasm",
      "fallback": false,
      "neuralExecuted": true
    }
  },
  {
    "precision": "fp32",
    "device": "wasm",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "r2",
      "score": 0.5084735923092537,
      "minimum": 0.3
    },
    "runtime": {
      "loadMs": 7662.739999949932,
      "inferenceMs": 5272.074999988079,
      "totalMs": 12934.814999938011,
      "rowsPerSecond": 21.05432870364154,
      "grade": "usable",
      "engine": "ONNX Runtime Web",
      "version": "1.30.0",
      "precision": "fp32",
      "device": "wasm",
      "requestedDevice": "wasm",
      "fallback": false,
      "neuralExecuted": true
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "r2",
      "score": 0.5152351770265482,
      "minimum": 0.4
    },
    "runtime": {
      "loadMs": 7416.399999976158,
      "inferenceMs": 1623.125,
      "totalMs": 9039.524999976158,
      "rowsPerSecond": 68.38659992298807,
      "grade": "usable",
      "engine": "ONNX Runtime Web",
      "version": "1.30.0",
      "precision": "fp32",
      "device": "webgpu",
      "requestedDevice": "webgpu",
      "fallback": false,
      "neuralExecuted": true
    }
  },
  {
    "precision": "fp32",
    "device": "webgpu",
    "status": "pass",
    "quality": {
      "passed": true,
      "metric": "r2",
      "score": 0.5084736112724402,
      "minimum": 0.3
    },
    "runtime": {
      "loadMs": 7416.399999976158,
      "inferenceMs": 1274.1449999809265,
      "totalMs": 8690.544999957085,
      "rowsPerSecond": 87.11724332918281,
      "grade": "usable",
      "engine": "ONNX Runtime Web",
      "version": "1.30.0",
      "precision": "fp32",
      "device": "webgpu",
      "requestedDevice": "webgpu",
      "fallback": false,
      "neuralExecuted": true
    }
  }
]

Something not quite right?

Tell us what happened with TabICLv2. Loading, results, or the page itself—we’re here for all of it.

Attachments Optional

Up to 3 files · 50 MB each

Browser, OS, CPU/GPU, memory, screen, storage, and network details are included automatically. Model inputs, results, and browsing history are not collected.