| # random-albert-base-v2 | |
| We introduce random-albert-base-v2, which is a unpretrained version of Albert model. The weight of random-albert-base-v2 is randomly initiated and this can be particularly useful when we aim to train a language model from scratch or benchmark the effect of pretraining. | |
| It's important to note that tokenizer of random-albert-base-v2 is the same as albert-base-v2 because it's not a trivial task to get a random tokenizer and it's less meaningful compared to the random weight. | |
| A debatable advantage of pulling random-albert-base-v2 from Huggingface is to avoid using random seed in order to obtain the same randomness at each time. | |
| The code to obtain a such random model: | |
| ```python | |
| from transformers import AutoTokenizer, AutoModelForSequenceClassification | |
| def get_blank_model_from_hf(model_name="bert-base-cased"): | |
| model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=5) | |
| tokenizer = AutoTokenizer.from_pretrained(model_name) | |
| model.base_model.init_weights() | |
| model_name = "random-" + model_name | |
| base_model= model.base_model | |
| return base_model, tokenizer, model_name | |
| ``` | |