/usr/bin/accelerate-estimate-memory

usage: accelerate-estimate-memory [-h] [--library_name {timm,transformers}]
                                  [--dtypes {float32,float16,int8,int4} [{float32,float16,int8,int4} ...]]
                                  [--trust_remote_code]
                                  model_name

Model size estimator for fitting a model onto device(e.g. cuda, xpu) memory.

positional arguments:
  model_name            The model name on the Hugging Face Hub.

options:
  -h, --help            show this help message and exit
  --library_name, --library-name {timm,transformers}
                        The library the model has an integration with, such as `transformers`,
                        needed only if this information is not stored on the Hub.
  --dtypes {float32,float16,int8,int4} [{float32,float16,int8,int4} ...]
                        The dtypes to use for the model, must be one (or many) of `float32`,
                        `float16`, `int8`, and `int4`
  --trust_remote_code, --trust-remote-code
                        Whether or not to allow for custom models defined on the Hub in their own
                        modeling files. This flag should only be used for repositories you trust
                        and in which you have read the code, as it will execute code present on
                        the Hub on your local machine.