Disable downloading deployment directory

[dbogunowicz]

Because we now (see PR #379) have Model attributes for both directory and directory.tar.gz, it is a bad idea to download those models by default when we call Model(...).path or Model(...).download(). With LLMs, downloading two identical deployment directories (one tar'ed another in the form of loose files) can take a lot of time. Let's disable downloading the deployment directory while keeping the appropriate references intact.

Manual testing:

from sparsezoo import Model
import shutil
import os

path_where_models_are = "/home/ubuntu/.cache/sparsezoo/neuralmagic/mobilebert-squad_wikipedia_bookcorpus-14layer_pruned50.4block_quantized"
if os.path.exists(path_where_models_are):
    shutil.rmtree(path_where_models_are)
    print(f"The path {path_where_models_are} has been cleaned up")
else:
    print(f"The path {path_where_models_are} does not exist")

print("1: Creating the model from stub")
model = Model("zoo:mobilebert-squad_wikipedia_bookcorpus-14layer_pruned50.4block_quantized")
model_path = model.path
print("Deployment directory path: ", model.deployment._path)
print("Deployment directory tar path: ", model.deployment_tar.path)
model.deployment_directory_path
print("Deployment directory path: ", model.deployment._path)
print("Contents of the deployment directory: ", os.listdir(model.deployment._path))
print("Contents of the model directory: ", os.listdir(model.path))

print("2: Creating the model from local path")
model = Model("zoo:mobilebert-squad_wikipedia_bookcorpus-14layer_pruned50.4block_quantized")
model_path = model.path
print("Deployment directory path: ", model.deployment._path)
print("Deployment directory tar path: ", model.deployment_tar.path)
model.deployment_directory_path
print("Deployment directory path: ", model.deployment._path)
print("Contents of the deployment directory: ", os.listdir(model.deployment._path))
print("Contents of the model directory: ", os.listdir(model.path))

out:

```bash
path_where_models_are = "/home/ubuntu/.cache/sparsezoo/neuralmagic/mobilebert-squad_wikipedia_bookcorpus-14layer_pruned50.4block_quantized"
if os.path.exists(path_where_models_are):
    shutil.rmtree(path_where_models_are)
    print(f"The path {path_where_models_are} has been cleaned up")
else:
    print(f"The path {path_where_models_are} does not exist")


print("1: Creating the model from stub")
model = Model("zoo:mobilebert-squad_wikipedia_bookcorpus-14layer_pruned50.4block_quantized")
model_path = model.path
print("Deployment directory path: ", model.deployment._path) # should be None initially
print("Deployment directory tar path: ", model.deployment_tar.path) # should be `deployment.tar.gz`
model.deployment_directory_path # unpack the tar
print("Deployment directory path: ", model.deployment._path) # should be `deployment`
print("Contents of the deployment directory: ", os.listdir(model.deployment._path))
print("Contents of the model directory: ", os.listdir(model.path))

print("2: Creating the model from local path")
model = Model("zoo:mobilebert-squad_wikipedia_bookcorpus-14layer_pruned50.4block_quantized")
model_path = model.path
print("Deployment directory path: ", model.deployment._path) # should be None initially
print("Deployment directory tar path: ", model.deployment_tar.path) # should be `deployment.tar.gz`
model.deployment_directory_path
print("Deployment directory path: ", model.deployment._path) # should be `deployment`
print("Contents of the deployment directory: ", os.listdir(model.deployment._path))
print("Contents of the model directory: ", os.listdir(model.path))

[rahul-tuli]

I have a question; are we sure all stubs WILL always have a deployment.tar.gz?