06f1fd69a6
Some checks failed
Self-hosted runner (nightly-past-ci-caller) / Get number (push) Has been cancelled
Self-hosted runner (nightly-past-ci-caller) / TensorFlow 2.11 (push) Has been cancelled
Self-hosted runner (nightly-past-ci-caller) / TensorFlow 2.10 (push) Has been cancelled
Self-hosted runner (nightly-past-ci-caller) / TensorFlow 2.9 (push) Has been cancelled
Self-hosted runner (nightly-past-ci-caller) / TensorFlow 2.8 (push) Has been cancelled
Self-hosted runner (nightly-past-ci-caller) / TensorFlow 2.7 (push) Has been cancelled
Self-hosted runner (nightly-past-ci-caller) / TensorFlow 2.6 (push) Has been cancelled
Self-hosted runner (nightly-past-ci-caller) / TensorFlow 2.5 (push) Has been cancelled
Self-hosted runner (benchmark) / Benchmark (aws-g5-4xlarge-cache) (push) Has been cancelled
Build documentation / build (push) Has been cancelled
Build documentation / build_other_lang (push) Has been cancelled
CodeQL Security Analysis / CodeQL Analysis (push) Has been cancelled
New model PR merged notification / Notify new model (push) Has been cancelled
PR CI / pr-ci (push) Has been cancelled
Slow tests on important models (on Push - A10) / Get all modified files (push) Has been cancelled
Secret Leaks / trufflehog (push) Has been cancelled
Update Transformers metadata / build_and_package (push) Has been cancelled
Slow tests on important models (on Push - A10) / Model CI (push) Has been cancelled
Check Tiny Models / Check tiny models (push) Has been cancelled
Self-hosted runner (Intel Gaudi3 scheduled CI caller) / Model CI (push) Has been cancelled
Self-hosted runner (Intel Gaudi3 scheduled CI caller) / Pipeline CI (push) Has been cancelled
Self-hosted runner (Intel Gaudi3 scheduled CI caller) / Example CI (push) Has been cancelled
Self-hosted runner (Intel Gaudi3 scheduled CI caller) / DeepSpeed CI (push) Has been cancelled
Self-hosted runner (Intel Gaudi3 scheduled CI caller) / Trainer/FSDP CI (push) Has been cancelled
Nvidia CI - Flash Attn / Setup (push) Has been cancelled
Nvidia CI - Flash Attn / Model CI (push) Has been cancelled
Nvidia CI / Setup (push) Has been cancelled
Nvidia CI / Model CI (push) Has been cancelled
Nvidia CI / Torch pipeline CI (push) Has been cancelled
Nvidia CI / Example CI (push) Has been cancelled
Nvidia CI / Trainer/FSDP CI (push) Has been cancelled
Nvidia CI / DeepSpeed CI (push) Has been cancelled
Nvidia CI / Quantization CI (push) Has been cancelled
Nvidia CI / Kernels CI (push) Has been cancelled
Doctests / Setup (push) Has been cancelled
Doctests / Call doctest jobs (push) Has been cancelled
Doctests / Send results to webhook (push) Has been cancelled
Extras Smoke Test / Get supported Python versions (push) Has been cancelled
Extras Smoke Test / Test extras on Python ${{ matrix.python-version }} (push) Has been cancelled
Extras Smoke Test / Check Slack token availability (push) Has been cancelled
Extras Smoke Test / Notify failures to Slack (push) Has been cancelled
Self-hosted runner (AMD scheduled CI caller) / Trigger Scheduled AMD CI (push) Has been cancelled
Stale Bot / Close Stale Issues (push) Has been cancelled
2.4 KiB
2.4 KiB
BERTología
Hay un creciente campo de estudio empeñado en la investigación del funcionamiento interno de los transformers de gran escala como BERT (que algunos llaman "BERTología"). Algunos buenos ejemplos de este campo son:
- BERT Rediscovers the Classical NLP Pipeline por Ian Tenney, Dipanjan Das, Ellie Pavlick: https://huggingface.co/papers/1905.05950
- Are Sixteen Heads Really Better than One? por Paul Michel, Omer Levy, Graham Neubig: https://huggingface.co/papers/1905.10650
- What Does BERT Look At? An Analysis of BERT's Attention por Kevin Clark, Urvashi Khandelwal, Omer Levy, Christopher D. Manning: https://huggingface.co/papers/1906.04341
- CAT-probing: A Metric-based Approach to Interpret How Pre-trained Models for Programming Language Attend Code Structure: https://huggingface.co/papers/2210.04633
Para asistir al desarrollo de este nuevo campo, hemos incluido algunas features adicionales en los modelos BERT/GPT/GPT-2 para ayudar a acceder a las representaciones internas, principalmente adaptado de la gran obra de Paul Michel (https://huggingface.co/papers/1905.10650):
- accediendo a todos los hidden-states de BERT/GPT/GPT-2,
- accediendo a todos los pesos de atención para cada head de BERT/GPT/GPT-2,
- adquiriendo los valores de salida y gradientes de las heads para poder computar la métrica de importancia de las heads y realizar la poda de heads como se explica en https://huggingface.co/papers/1905.10650.
Para ayudarte a entender y usar estas features, hemos añadido un script específico de ejemplo: bertology.py mientras extraes información y cortas un modelo pre-entrenado en GLUE.