| # Modifications from the upstream model |
|
|
| `gliclass-std-base-v3-daecore-5facet-qint8-v2` is not an unmodified copy of |
| `knowledgator/gliclass-base-v3.0`. It is not endorsed by Knowledgator or by |
| Microsoft, whose `deberta-v3-base` backbone the upstream model builds on. |
|
|
| What changed: |
|
|
| - **Fine-tune.** The upstream GLiClass single-pass classifier was trained |
| further on 59,886 labeled passages for five fixed facets (`trap`, |
| `decision`, `constraint`, `mechanism`, `procedure`) with the five label |
| prompts recorded in `classifier-metadata.json`. Labels were frontier-model |
| judgments under a frozen protocol, not human annotations. |
| - **Export.** The fine-tuned weights were exported to ONNX (opset 17) as a |
| single graph taking `input_ids` and `attention_mask` and emitting five |
| logits. |
| - **Quantization.** Only the token-embedding `Gather` tables were quantized |
| to signed INT8 (per-channel off, reduce-range off); every matrix product |
| stays FP32. Twenty-four constant identity nodes were folded. The compression |
| was qualified as equivalent to the FP32 export on a 5,298-row panel. |
| - **Calibration.** Per-facet temperature scaling and two frozen threshold |
| tables (`recall_leaning`, `contract`) travel in `classifier-metadata.json` |
| and are part of the artifact's identity. |
| - **Vulkan graph preparation.** Bounded integer/boolean mask calculations |
| use exact FP32 equivalents before their original output types are restored. |
| Learned parameters, tokenizers, label prompts, calibration and thresholds |
| are unchanged. `vulkan-derivation.json` binds the source and derived graph; |
| `classifier-metadata.json` records the new graph's byte identity. |
|
|
| Unchanged: the tokenizer vocabulary and the `<<LABEL>>` / `<<SEP>>` prompt |
| convention of the upstream model. |
|
|
