Instructions to use spybyscript/gliner25-decide-litert with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LiteRT
How to use spybyscript/gliner25-decide-litert with LiteRT:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- GLiNER2
How to use spybyscript/gliner25-decide-litert with GLiNER2:
from gliner2 import AutoExtractor extractor = AutoExtractor.from_pretrained("spybyscript/gliner25-decide-litert") # Extract entities text = "Apple CEO Tim Cook announced iPhone 15 in Cupertino yesterday." result = extractor.extract_entities(text, ["company", "person", "product", "location"]) print(result) - Notebooks
- Google Colab
- Kaggle
Card: what this port does, and what it leaves out
Browse files- README.md +13 -0
- SHA256SUMS +1 -1
README.md
CHANGED
|
@@ -26,6 +26,19 @@ that share one set of weights; the token table is a separate float16 file the ho
|
|
| 26 |
| `gliner25_decide_fp16hw.tflite` | 1026 MB | wfp16's storage (float16 FULLY_CONNECTED weights behind DEQUANTIZE) over the FP16-hardware graph: LayerNorm power-of-two pre-scale (x*2^-k, eps*4^-k, k from the corpus' FP32 ranges), first feed-forward FULLY_CONNECTED split in two halves, finite -1e4 mask |
|
| 27 |
| `gliner25_decide_embeddings_f16.bin` | 262 MB | float16 [128011, 1024], little-endian, row-major |
|
| 28 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
## Use
|
| 30 |
|
| 31 |
Tokenize with `tokenizer.json` and pack the prompt exactly as gliner2 2.0.0 does. Pick the smallest signature that holds the ids
|
|
|
|
| 26 |
| `gliner25_decide_fp16hw.tflite` | 1026 MB | wfp16's storage (float16 FULLY_CONNECTED weights behind DEQUANTIZE) over the FP16-hardware graph: LayerNorm power-of-two pre-scale (x*2^-k, eps*4^-k, k from the corpus' FP32 ranges), first feed-forward FULLY_CONNECTED split in two halves, finite -1e4 mask |
|
| 27 |
| `gliner25_decide_embeddings_f16.bin` | 262 MB | float16 [128011, 1024], little-endian, row-major |
|
| 28 |
|
| 29 |
+
## What this port does, and what it leaves out
|
| 30 |
+
|
| 31 |
+
- **Classification, as the gliner2 library runs `classify_text`:** single-label and multi-label tasks, labels with
|
| 32 |
+
descriptions, several tasks in one call. On 2,338 recorded decisions it gives the library's label on all 2,338
|
| 33 |
+
(`wfp16`) and on 2,332 (`fp16hw` at GPU FP16); see Validation.
|
| 34 |
+
- **No extraction.** The original checkpoint also carries GLiNER2's span heads (`span_rep`, `count_pred`, `count_embed`),
|
| 35 |
+
which serve `extract_entities`, `extract_json` and `extract_relations`, and its card lists named entity recognition.
|
| 36 |
+
This port exports only the encoder and the classification head, so it cannot extract spans. For extraction on
|
| 37 |
+
LiteRT, see [spybyscript/gliner25-base-litert](https://huggingface.co/spybyscript/gliner25-base-litert).
|
| 38 |
+
- **At most 512 tokens a request** (the task prompts, labels and text together). Longer requests are refused, not
|
| 39 |
+
truncated: split the text or spread the tasks over several calls. The original runs with dynamic shapes.
|
| 40 |
+
- **Weights:** only their storage and compute precision change (see the file table); nothing is retrained or pruned.
|
| 41 |
+
|
| 42 |
## Use
|
| 43 |
|
| 44 |
Tokenize with `tokenizer.json` and pack the prompt exactly as gliner2 2.0.0 does. Pick the smallest signature that holds the ids
|
SHA256SUMS
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30 LICENSE
|
| 2 |
0a8689579d9a76e8b3b1c6769e7293d7ad0672394545f7fab94c24eb167a96a0 NOTICE.txt
|
| 3 |
-
|
| 4 |
b3595e87167c626611cb102d066bda3a07af7c521520a1e4ee04d52db33ab5f2 config.json
|
| 5 |
f90f5d837721b3d4df7194a89ea18bbea5c449e6944fb6078664f1a8d20484fd gliner25_decide_embeddings_f16.bin
|
| 6 |
b5567c71caddf9ce476e242c8778622375a8b2c415cf078248f75b092452b605 gliner25_decide_fp16hw.tflite
|
|
|
|
| 1 |
cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30 LICENSE
|
| 2 |
0a8689579d9a76e8b3b1c6769e7293d7ad0672394545f7fab94c24eb167a96a0 NOTICE.txt
|
| 3 |
+
b2b84a25c40c58629a95ee2f5eebd03f77ba1090e26ed44000b6edd93fcd9b15 README.md
|
| 4 |
b3595e87167c626611cb102d066bda3a07af7c521520a1e4ee04d52db33ab5f2 config.json
|
| 5 |
f90f5d837721b3d4df7194a89ea18bbea5c449e6944fb6078664f1a8d20484fd gliner25_decide_embeddings_f16.bin
|
| 6 |
b5567c71caddf9ce476e242c8778622375a8b2c415cf078248f75b092452b605 gliner25_decide_fp16hw.tflite
|