Special Shape of [38809] will hang

[swimdi]

On dispatch device, shape [38000] and [39000] is ok for ttnn.from_torch, but the special shape of [38809] will hang

import torch
import ttnn

dispatch_core_config = ttnn.DispatchCoreConfig(ttnn.device.DispatchCoreType.ETH, ttnn.DispatchCoreAxis.ROW)
device = ttnn.open_device(device_id=0, dispatch_core_config=dispatch_core_config, l1_small_size=16384)

# ok
torch_indices_pass1 = torch.randint(0, 732, [38000])
indices_tensor_pass1 = ttnn.from_torch(torch_indices_pass1, device = device, layout = ttnn.ROW_MAJOR_LAYOUT, dtype = ttnn.uint32)
# ok
torch_indices_pass2 = torch.randint(0, 732, [39000])
indices_tensor_pass2 = ttnn.from_torch(torch_indices_pass2, device = device, layout = ttnn.ROW_MAJOR_LAYOUT, dtype = ttnn.uint32)

# hang
torch_indices_hang = torch.randint(0, 732, [38809])
indices_tensor_hang = ttnn.from_torch(torch_indices_hang, device = device, layout = ttnn.ROW_MAJOR_LAYOUT, dtype = ttnn.uint32)

And if device is not dispatch, then shape [38809] is ok

import torch
import ttnn

device = ttnn.open_device(device_id=0, l1_small_size=16384)

# not hang
torch_indices = torch.randint(0, 732, [38809])
indices_tensor = ttnn.from_torch(torch_indices, device = device, layout = ttnn.ROW_MAJOR_LAYOUT, dtype = ttnn.uint32)

This special shape is from beit