def convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, albert_config_file,
pytorch_dump_path):
# Initialise PyTorch model
config = AlbertConfig.from_json_file(albert_config_file)
print("Building PyTorch model from configuration: {}".format(str(config)))
model = AlbertForMaskedLM(config)
load_tf_weights_in_albert(model, config, tf_checkpoint_path)
print("Save PyTorch model to {}".format(pytorch_dump_path))
torch.save(model.state_dict(), pytorch_dump_path)
def convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, albert_config_file, pytorch_dump_path):
# Initialise PyTorch model
config = AlbertConfig.from_json_file(albert_config_file)
print(f"Building PyTorch model from configuration: {config}")
model = AlbertForPreTraining(config)
# Load weights from tf checkpoint
load_tf_weights_in_albert(model, config, tf_checkpoint_path)
# Save pytorch-model
print(f"Save PyTorch model to {pytorch_dump_path}")
torch.save(model.state_dict(), pytorch_dump_path)
def albert_convert_tf_checkpoint_to_pytorch(tf_checkpoint_path,
albert_config_file,
pytorch_dump_path):
from transformers import AlbertConfig, AlbertForMaskedLM, load_tf_weights_in_albert
# Initialise PyTorch model
config = AlbertConfig.from_json_file(albert_config_file)
print("Building PyTorch model from configuration: {}".format(str(config)))
model = AlbertForMaskedLM(config)
# Load weights from tf checkpoint
load_tf_weights_in_albert(model, config, tf_checkpoint_path)
# Save pytorch-model
print("Save PyTorch model to {}".format(pytorch_dump_path))
torch.save(model.state_dict(), pytorch_dump_path)
def main(args):
with open(args.config) as fp:
data = json.loads(fp.read())
config = AlbertConfig(**data)
model = AlbertForMaskedLM(config)
model: AlbertForMaskedLM = load_tf_weights_in_albert(model, config, args.checkpoint)
model.save_pretrained(args.output)
def main(args):
with open(args.config) as fp:
data = json.loads(fp.read())
config = AlbertConfig(**data)
model = AlbertForMaskedLM(config)
model: AlbertForMaskedLM = load_tf_weights_in_albert(
model, config, args.checkpoint)
model.save_pretrained(args.output)
tokenizer = AlbertTokenizer.from_pretrained(args.spiece, keep_accents=True)
tokenizer.save_pretrained(args.output)
