---
license: apache-2.0
language:
- en
base_model: fastino/GLiNER2.5-Decide
pipeline_tag: text-classification
tags:
- gliner2
- text-classification
- coreml
- apple
- ios
- on-device
- fp16
---

# GLiNER2.5-Decide — Core ML (iOS 18+ / macOS 15+)

| File | Purpose |
|---|---|
| `GLiNER25-Decide-FP16-CoreML.mlpackage/` | Core ML package, FP16 weights and compute (~945 MB), three functions sharing one set of weights |
| `tokenizer.json`, `tokenizer_config.json`, `special_tokens_map.json` | Checkpoint tokenizer |
| `model-card.md` | Upstream model card (fastino/GLiNER2.5-Decide, Apache-2.0) |

Source: checkpoint revision `bbe10ff77ebb238777c17d3a8ac9260e30929057`,
classification-only (DeBERTa encoder + classifier). Built with coremltools
9.0 / Torch 2.7.1 from FP16 128/256/512 exports (finite FP16 attention
mask), merged into one multifunction package with shared weights.

## Interface

Functions `context128`, `context256`, `context512` (select with
`MLModelConfiguration.functionName`):

- inputs: `input_ids` int32 `[1, L]`, `attention_mask` int32 `[1, L]`
  (1 = real token, 0 = padding)
- output: `logits` `[1, L]`, one score per token position. Classification
  reads the positions of the schema's label markers (`[L]`).

Choose the smallest function that fits the request and pad to its length.
The maximum is 512 tokens. Reject longer requests; do not truncate.

## Required: GPU compute units

```swift
let configuration = MLModelConfiguration()
configuration.computeUnits = .cpuAndGPU
configuration.functionName = "context256"
let model = try MLModel(contentsOf: compiledURL, configuration: configuration)
```

Do not use INT8 weights on the GPU: they were nondeterministic and changed
decisions. Do not rely on `.cpuOnly` for this package either: FP16 on CPU
changes no decisions but exceeds the strict gate (max 0.012). Xcode compiles
the `.mlpackage` to `.mlmodelc` when it is bundled in an app; use
`MLModel.compileModel(at:)` otherwise.

## Validation (iPhone 17 Pro, iOS 27.2, GPU)

Strict gate: every decision matches the FP32 PyTorch oracle and max
probability error ≤ 0.005. Two launches per function, identical results:

| Function | Corpus | Max error | Median |
|---|---|---:|---:|
| context128 | 75 | 0.001371 | 26 ms |
| context256 | 86 | 0.001371 | 44 ms |
| context512 | 95 (9 requests of 296–512 tokens) | 0.001371 | 121–124 ms |

Load ≈ 1.7–2.0 s; physical footprint ≤ 0.32 GB (weights are file-mapped).
**Not yet validated on an iOS 18–26 device or an older GPU.**

## SHA-256

```
bc1092e6eb7185c9ff1c3fb8e8659b70b8c12f295fc573c39c197f768c102078  GLiNER25-Decide-FP16-CoreML.mlpackage/Data/com.apple.CoreML/weights/weight.bin
fe39a4ea16cb1d35699d4cc2571f9c4e80692caccdafbb689645be56867f8d9f  GLiNER25-Decide-FP16-CoreML.mlpackage/Data/com.apple.CoreML/model.mlmodel
3ad87d9ffe669147063e70850927dd2da90249e2acc5c8527f1eb65df467bcc8  tokenizer.json
```