opus-mt-fr-en-mic

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

Contract

MarianMT fr-en translation

Runnable usage and the framework expectations directory are linked above.

Slim model

Slim fixture: opus-mt-fr-en-mic

A 4.0M-parameter fixture for the MarianMT MIC, covering Helsinki-NLP/opus-mt-fr-en. It is a test artifact, not a usable model, tagged test-artifact to keep it out of candidate-model search.

  • Fixture key 49fa7f6e6e14. It covers every checkpoint whose processor files are byte-identical and whose config matches after shrinking.
  • Same architecture, smaller config: MarianMTModel, shrunk by the generic rule in sweep.py (encoder_layers 2, decoder_layers 2).
  • Processor files copied byte for byte from Helsinki-NLP/opus-mt-fr-en: generation_config.json, source.spm, target.spm, tokenizer_config.json, vocab.json.
  • Overfit to the contract's training step (100 steps). Random weights failed the input-sensitivity gate: input_ids:reverse moved the logits by only 0.00092 (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 Helsinki-NLP/opus-mt-fr-en 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 opus-mt-fr-en-mic.

Downloads last month
298
Safetensors
Model size
4.04M params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for huggingface-mic/opus-mt-fr-en-mic

Finetuned
(23)
this model