deberta-large-mnli-mic

Slim model weights, executable usage and framework expectations in one repository.

• Contract manifest
• Transformers usage
• Transformers expectations
• Fixture provenance
• How the fixture was shrunk

Contract

DeBERTa-v3-large MNLI sequence classification

Runnable usage and the framework expectations directory are linked above.

Slim model

Slim fixture: deberta-large-mnli-mic

A 3.4M-parameter fixture for the DeBERTa-MNLI MIC, covering microsoft/deberta-large-mnli. It is a test artifact, not a usable model, tagged test-artifact to keep it out of candidate-model search.

• Fixture key 12e4a6a0a518. It covers every checkpoint whose processor files are byte-identical and whose config matches after shrinking.
• Same architecture, smaller config: DebertaForSequenceClassification, shrunk by the generic rule in sweep.py (num_hidden_layers 2).
• Processor files copied byte for byte from microsoft/deberta-large-mnli: merges.txt, tokenizer_config.json, vocab.json.
• Overfit to the contract's training step (30 steps). Random weights failed the input-sensitivity gate: attention_mask:all_ones moved the logits by only 4.8e-05 (threshold 0.001).
• Same tensor names as the real checkpoint. The real checkpoint has every parameter, and so does the fixture.
• Saved in float32, the dtype microsoft/deberta-large-mnli loads in by default.
• Finite in float32, bfloat16, and float16 (forward, loss, gradients).

Built by build_fixtures.py; build.json records the pinned sources, gate results, and file hashes. This repository includes the slim weights, usage and framework expectations; rebuild the other files with python example/build_fixtures.py --fixture deberta-large-mnli-mic.