def test_small_byt5_integration_test(self):
"""
For comparision run:
>>> import t5 # pip install t5==0.9.1
>>> path_to_byt5_small_checkpoint = '<fill_in>'
>>> t5_model = t5.models.MtfModel(model_dir=path_to_tf_checkpoint, batch_size=1, tpu=None)
>>> vocab = t5.data.ByteVocabulary()
>>> score = t5_model.score(inputs=["Hello there"], targets=["Hi I am"], vocabulary=vocab)
"""
model = FlaxT5ForConditionalGeneration.from_pretrained(
"google/byt5-small")
tokenizer = ByT5Tokenizer.from_pretrained("google/byt5-small")
input_ids = tokenizer("Hello there", return_tensors="np").input_ids
labels = tokenizer("Hi I am", return_tensors="np").input_ids
decoder_input_ids = shift_tokens_right(
labels, model.config.pad_token_id,
model.config.decoder_start_token_id)
logits = model(input_ids, decoder_input_ids=decoder_input_ids).logits
loss = optax.softmax_cross_entropy(logits,
onehot(labels,
logits.shape[-1])).mean()
mtf_score = -(labels.shape[-1] * loss.item())
EXPECTED_SCORE = -60.7397
self.assertTrue(abs(mtf_score - EXPECTED_SCORE) < 1e-4)
def test_small_byt5_integration_test(self):
"""
For comparision run:
>>> import t5 # pip install t5==0.9.1
>>> path_to_byt5_small_checkpoint = '<fill_in>'
>>> t5_model = t5.models.MtfModel(model_dir=path_to_tf_checkpoint, batch_size=1, tpu=None)
>>> vocab = t5.data.ByteVocabulary()
>>> score = t5_model.score(inputs=["Hello there"], targets=["Hi I am"], vocabulary=vocab)
"""
model = TFT5ForConditionalGeneration.from_pretrained("google/byt5-small")
tokenizer = ByT5Tokenizer.from_pretrained("google/byt5-small")
input_ids = tokenizer("Hello there", return_tensors="tf").input_ids
labels = tokenizer("Hi I am", return_tensors="tf").input_ids
loss = model(input_ids, labels=labels).loss
mtf_score = -tf.math.reduce_sum(loss).numpy()
EXPECTED_SCORE = -60.7397
self.assertTrue(abs(mtf_score - EXPECTED_SCORE) < 1e-4)
def t5_base_tokenizer(self):
return ByT5Tokenizer.from_pretrained("google/byt5-small")