tarekziade HF Staff
Consolidate MIC usage, slim weights and reviewed framework expectations
8033673 verified |
Download docs/source-cards/fixture.md from huggingface-mic/vit-mic: direct link, hf CLI and curl.
- Browser
- Download file 1.72 kB
-
https://huggingface.co/huggingface-mic/vit-mic/resolve/main/docs/source-cards/fixture.md
- Command line
-
hf download hf://huggingface-mic/vit-mic/docs/source-cards/fixture.md
-
curl -L -o fixture.md https://huggingface.co/huggingface-mic/vit-mic/resolve/main/docs/source-cards/fixture.md
1.72 kB
metadata
tags:
- mic
- test-artifact
base_model: google/vit-base-patch16-224
Slim fixture: mic-vit-slim
A 0.2M-parameter fixture for the vit MIC, covering google/vit-base-patch16-224. It is a test artifact, not a usable model, tagged test-artifact to keep it out of candidate-model search.
- Fixture key
3fc8366aafe8. It covers every checkpoint whose processor files are byte-identical and whose config matches after shrinking. - Same architecture, smaller config:
ViTForImageClassification, shrunk by the generic rule insweep.py(num_hidden_layers 2). - Processor files copied byte for byte from
google/vit-base-patch16-224:preprocessor_config.json. - Overfit to the contract's training step (300 steps). Random weights failed the input-sensitivity gate:
pixel_values:reversemoved the logits by only 5.9e-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
google/vit-base-patch16-224loads 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. The source repository commits only README.md and build.json; rebuild the other files with python example/build_fixtures.py --fixture mic-vit-slim.