Instructions to use optimum/legal-bert-base-uncased-neuronx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use optimum/legal-bert-base-uncased-neuronx with Transformers:
# Load model directly from transformers import AutoTokenizer, AutoModelForPreTraining tokenizer = AutoTokenizer.from_pretrained("optimum/legal-bert-base-uncased-neuronx") model = AutoModelForPreTraining.from_pretrained("optimum/legal-bert-base-uncased-neuronx", device_map="auto") - Notebooks
- Google Colab
- Kaggle
This repo contains artifacts from nlpaueb/legal-bert-base-uncased but in neuronx format compatible with INF2 and TRN1 devices.
from transformers import AutoTokenizer
from optimum.neuron import NeuronBertForMaskedLM
input_shapes = {"batch_size": 1, "sequence_length": 128}
compiler_args = {"auto_cast": "matmul", "auto_cast_type": "bf16"}
neuron_model = NeuronBertForMaskedLM.from_pretrained(
"nlpaueb/legal-bert-base-uncased",
export=True,
**input_shapes,
**compiler_args,
)
# Save locally
neuron_model.save_pretrained("legal_bert_masked_lm_neuronx")
neuron_model.push_to_hub(
"legal_bert_masked_lm_neuronx",
repository_id="optimum/legal-bert-base-uncased-neuronx", # Replace with your HF Hub repo id
)
- Downloads last month
- 17
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support