Make printing of MXLinear and MXInferenceLinear more descriptive

Summary:

For now, at least print the config.  It's pretty verbose, we can make
the string more concise in a future PR.

Test Plan: CI

Reviewers:

Subscribers:

Tasks:

Tags:

ghstack-source-id: c4cc7e9
ghstack-comment-id: 2722102724
Pull Request resolved: #1883

Author: vkuzo

torchao/prototype/mx_formats/mx_linear.py

@@ -159,6 +159,10 @@ def forward(self, x):
             y = y + self.bias
         return y
 
+    def extra_repr(self):
+        s = f"{super().extra_repr()}, config={self.config}"
+        return s
+
 
 class MXInferenceLinear(torch.nn.Linear):
     """
@@ -201,6 +205,10 @@ def forward(self, x):
         y = F.linear(x, w_hp, self.bias)
         return y
 
+    def extra_repr(self):
+        s = f"{super().extra_repr()}, config={self.config}"
+        return s
+
 
 def replace_with_custom_fn_if_matches_filter(
     model, replacement_fn, filter_fn, cur_fqn=""