gistfile1.txt

tensor([0.0002]) tensor([0], dtype=torch.int32)
tensor([0.0002]) tensor([-1], dtype=torch.int32)
WARNING:root:Could not determine LOCAL_WORLD_SIZE from environment, falling back to WORLD_SIZE.
WARNING:root:Could not determine LOCAL_WORLD_SIZE from environment, falling back to WORLD_SIZE.
torch.float32
torch.float32
old weight tensor([[-0.0285,  0.0301,  0.0173,  ..., -0.0305, -0.0288, -0.0027],
        [-0.0224, -0.0263,  0.0212,  ..., -0.0249,  0.0071, -0.0202],
        [ 0.0125,  0.0225,  0.0154,  ..., -0.0155, -0.0169,  0.0253],
        ...,
        [ 0.0006,  0.0041,  0.0118,  ..., -0.0078,  0.0023, -0.0024],
        [-0.0161, -0.0222, -0.0155,  ..., -0.0104, -0.0312,  0.0072],
        [-0.0046,  0.0148, -0.0016,  ...,  0.0116,  0.0208, -0.0281]])
new weight tensor([[-0.0284,  0.0301,  0.0172,  ..., -0.0306, -0.0287, -0.0027],
        [-0.0223, -0.0262,  0.0211,  ..., -0.0250,  0.0071, -0.0201],
        [ 0.0125,  0.0225,  0.0154,  ..., -0.0154, -0.0169,  0.0252],
        ...,
        [ 0.0007,  0.0042,  0.0118,  ..., -0.0078,  0.0025, -0.0025],
        [-0.0162, -0.0223, -0.0154,  ..., -0.0103, -0.0311,  0.0071],
        [-0.0047,  0.0147, -0.0015,  ...,  0.0115,  0.0208, -0.0282]])
tensor([[-0.0284,  0.0301,  0.0172,  ...,  0.0230, -0.0120,  0.0135],
        [-0.0223, -0.0262,  0.0211,  ...,  0.0203,  0.0265, -0.0037],
        [ 0.0125,  0.0225,  0.0154,  ..., -0.0243,  0.0238, -0.0051],
        ...,
        [ 0.0120,  0.0056,  0.0135,  ...,  0.0179,  0.0039, -0.0037],
        [ 0.0088, -0.0150,  0.0076,  ...,  0.0287, -0.0270,  0.0145],
        [-0.0311,  0.0221, -0.0164,  ...,  0.0252,  0.0206,  0.0199]],
       grad_fn=<SplitLookupFunction_sgd_Op>>)
tensor([[-0.0285,  0.0301,  0.0173,  ...,  0.0231, -0.0121,  0.0136],
        [-0.0224, -0.0263,  0.0212,  ...,  0.0204,  0.0266, -0.0037],
        [ 0.0125,  0.0225,  0.0154,  ..., -0.0244,  0.0237, -0.0052],
        ...,
        [ 0.0121,  0.0057,  0.0134,  ...,  0.0178,  0.0040, -0.0038],
        [ 0.0087, -0.0150,  0.0076,  ...,  0.0286, -0.0269,  0.0145],
        [-0.0311,  0.0220, -0.0164,  ...,  0.0252,  0.0207,  0.0199]],
       grad_fn=<SplitLookupFunction_sgd_Op>>)
*******************
*******************
1049520
264696
0.25220672307340497