Description of problem:

    Memory leak is observed when you do reapeted ops

Version numbers (base image, wheels, builder, etc):

Steps to Reproduce:

    1. Build and install torch
    2. Repro code https://gist.github.com/sumantro93/d62b8a576cdeae9a6dbeb80a185edc19
    

Actual results:

Memory leak is observed    

Expected results:

It should be handling 

Additional info:

import torch
... import gc
... def get_memory_mb():
...     import resource
...     return resource.getrusage(resource.RUSAGE_SELF).ru_maxrss / 1024
... initial_mem = get_memory_mb()
... print(f"Initial memory: {initial_mem:.2f} MB")
... # Repeated operations cause memory leak
... for i in range(10000):
...     t = torch.randn(2, 3, 4, 5, 6)  # 5 dims
...     t = t.view(1, 2, 3, 4, 5, 6)    # 6 dims - triggers leak
... gc.collect()  # This is OUTSIDE the loop
... final_mem = get_memory_mb()
... print(f"Final memory: {final_mem:.2f} MB")
... print(f"Memory leaked: {final_mem - initial_mem:.2f} MB")
... 
Initial memory: 224.51 MB
Final memory: 226.14 MB
Memory leaked: 1.62 MB