|
Download README.md from huggingface-mic/bart-base-mic: direct link, hf CLI and curl.
- Browser
- Download file 2.24 kB
-
https://huggingface.co/huggingface-mic/bart-base-mic/resolve/main/README.md
- Command line
-
hf download hf://huggingface-mic/bart-base-mic/README.md
-
curl -L -o README.md https://huggingface.co/huggingface-mic/bart-base-mic/resolve/main/README.md
2.24 kB
| tags: | |
| - mic | |
| - test-artifact | |
| base_model: facebook/bart-base | |
| # bart-base-mic | |
| Slim model weights, executable usage and framework expectations in one repository. | |
| - [Contract manifest](integration.json) | |
| - [Transformers usage](integrations/transformers/usage.md) | |
| - [Transformers expectations](integrations/transformers/expectations) | |
| - [Fixture provenance](build.json) | |
| - [How the fixture was shrunk](https://huggingface-mic-mic-coverage.static.hf.space/shrink/bart-base-mic.html) | |
| ## Contract | |
| ### BART encoder-decoder | |
| Runnable usage and the framework expectations directory are linked above. | |
| ## Slim model | |
| ### Slim fixture: bart-base-mic | |
| A 3.5M-parameter fixture for the [BART MIC](https://huggingface.co/huggingface-mic/bart-base-mic), covering [`facebook/bart-base`](https://huggingface.co/facebook/bart-base). It is a test artifact, not a usable model, tagged `test-artifact` to keep it out of candidate-model search. | |
| - **Fixture key `c89a9e951b4f`.** It covers every checkpoint whose processor files are byte-identical and whose config matches after shrinking. | |
| - **Same architecture, smaller config:** `BartForConditionalGeneration`, shrunk by the generic rule in [`sweep.py`](https://github.com/huggingface/transformers-mic/blob/main/experiments/sweep.py) (encoder_layers 2, decoder_layers 2). | |
| - **Processor files copied byte for byte** from `facebook/bart-base`: `merges.txt`, `tokenizer.json`, `vocab.json`. | |
| - **Overfit to the contract's training step** (60 steps). Random weights failed the input-sensitivity gate: `input_ids:reverse` moved the logits by only 0.00087 (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 `facebook/bart-base` loads in by default. | |
| - **Finite in float32, bfloat16, and float16** (forward, loss, gradients). | |
| Built by [`build_fixtures.py`](https://github.com/huggingface/transformers-mic/blob/main/example/build_fixtures.py); [`build.json`](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 bart-base-mic`. | |