Hugging Peers

roberta-base-mic

☆ Save
by huggingface-mic
224 downloads 0 likes on Hugging Face updated

Model card

roberta-base-mic

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

Contract

RoBERTa fill-mask

Runnable usage and the framework expectations directory are linked above.

Slim model

Slim fixture: roberta-base-mic

A 3.4M-parameter fixture for the RoBERTa MIC, covering FacebookAI/roberta-base. It is a test artifact, not a usable model, tagged test-artifact to keep it out of candidate-model search.

  • Fixture key 05ec051e3d97. It covers every checkpoint whose processor files are byte-identical and whose config matches after shrinking.
  • Same architecture, smaller config: RobertaForMaskedLM, shrunk by the generic rule in sweep.py (num_hidden_layers 2).
  • Processor files copied byte for byte from FacebookAI/roberta-base: dict.txt, merges.txt, tokenizer.json, tokenizer_config.json, vocab.json.
  • Random weights (seed 17). They passed the input-sensitivity gate: every perturbation of the contract's inputs moved the logits by at least 0.001 (weakest: attention_mask:all_ones, 0.0024), so no training was needed.
  • Same tensor names as the real checkpoint. The real checkpoint has every parameter, and so does the fixture.
  • Saved in float32, the dtype FacebookAI/roberta-base 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 roberta-base-mic.