def test_conversion_adaptive_model_ner():
farm_model = Converter.convert_from_transformers("dslim/bert-base-NER", device="cpu")
transformer_model = farm_model.convert_to_transformers()[0]
transformer_model2 = AutoModelForTokenClassification.from_pretrained("dslim/bert-base-NER")
# compare weights
for p1, p2 in zip(transformer_model.parameters(), transformer_model2.parameters()):
assert (p1.data.ne(p2.data).sum() == 0)
def test_conversion_adaptive_model_classification():
farm_model = Converter.convert_from_transformers("deepset/bert-base-german-cased-hatespeech-GermEval18Coarse", device="cpu")
transformer_model = farm_model.convert_to_transformers()[0]
transformer_model2 = AutoModelForSequenceClassification.from_pretrained("deepset/bert-base-german-cased-hatespeech-GermEval18Coarse")
# compare weights
for p1, p2 in zip(transformer_model.parameters(), transformer_model2.parameters()):
assert (p1.data.ne(p2.data).sum() == 0)
def test_conversion_adaptive_model_lm():
farm_model = Converter.convert_from_transformers("bert-base-german-cased", device="cpu")
transformer_model = farm_model.convert_to_transformers()[0]
transformer_model2 = AutoModelWithLMHead.from_pretrained("bert-base-german-cased")
# compare weights
for p1, p2 in zip(transformer_model.parameters(), transformer_model2.parameters()):
assert (p1.data.ne(p2.data).sum() == 0)
def test_conversion_adaptive_model_qa():
farm_model = Converter.convert_from_transformers("deepset/bert-base-cased-squad2", device="cpu")
transformer_model = farm_model.convert_to_transformers()[0]
transformer_model2 = AutoModelForQuestionAnswering.from_pretrained("deepset/bert-base-cased-squad2")
# compare weights
for p1, p2 in zip(transformer_model.parameters(), transformer_model2.parameters()):
assert (p1.data.ne(p2.data).sum() == 0)
def convert_from_transformers():
transformers_input_name = "deepset/bert-base-german-cased-hatespeech-GermEval18Coarse"
farm_output_dir = Path(
"../saved_models/farm-bert-base-german-cased-hatespeech-GermEval18Coarse"
)
# # CASE 1: MODEL
# # Load model from transformers model hub (-> continue training / compare models / ...)
model = Converter.convert_from_transformers(transformers_input_name,
device="cpu")
# # Alternative way to load from transformers model hub:
#model = AdaptiveModel.convert_from_transformers(transformers_input_name, device="cpu", task_type="text_classification")
# # ... continue as in the other examples e.g. to fine-tune this QA model on your own data
#
# # CASE 2: INFERENCER
# # Load Inferencer from transformers, incl. model & tokenizer (-> just get predictions)
nlp = Inferencer.load(transformers_input_name,
task_type="text_classification")
#
# # run predictions
result = nlp.inference_from_dicts(dicts=[{"text": "Was ein scheiß Nazi!"}])
pprint.pprint(result)
nlp.close_multiprocessing_pool()
# save it
nlp.save(farm_output_dir)
def convert_from_transformers():
# CASE 1: MODEL
# Load model from transformers model hub (-> continue training / compare models / ...)
model = Converter.convert_from_transformers("deepset/bert-large-uncased-whole-word-masking-squad2", device="cpu")
#Alternative way to load from transformers model hub:
#model = AdaptiveModel.convert_from_transformers("deepset/bert-large-uncased-whole-word-masking-squad2", device="cpu", task_type="question_answering")
# ... continue as in the other examples e.g. to fine-tune this QA model on your own data
# CASE 2: INFERENCER
# Load Inferencer from transformers, incl. model & tokenizer (-> just get predictions)
nlp = Inferencer.load("deepset/bert-large-uncased-whole-word-masking-squad2", task_type="question_answering")
# run predictions
QA_input = [{"questions": ["Why is model conversion important?"],
"text": "The option to convert models between FARM and transformers gives freedom to the user and let people easily switch between frameworks."}]
result = nlp.inference_from_dicts(dicts=QA_input)
pprint.pprint(result)
nlp.close_multiprocessing_pool()
# save it
farm_model_dir = Path("../saved_models/bert-english-qa-large")
nlp.save(farm_model_dir)