Logging to experiments/direct_ft_bic/
creating model...
input shape change: torch.Size([256, 6, 3, 3])
loading model from: ../models/256_adm_uncond.pt complete.
creating data loader...
training...
-------------------------
| grad_norm  | 0.0222   |
| loss       | 0.026    |
| loss_q0    | 0.0811   |
| loss_q1    | 0.0067   |
| loss_q2    | 0.00142  |
| loss_q3    | 0.000164 |
| mse        | 0.0246   |
| mse_q0     | 0.076    |
| mse_q1     | 0.00665  |
| mse_q2     | 0.0014   |
| mse_q3     | 0.000162 |
| param_norm | 1.36e+03 |
| samples    | 256      |
| step       | 0        |
| vb         | 0.00144  |
| vb_q0      | 0.00505  |
| vb_q1      | 4.82e-05 |
| vb_q2      | 1.23e-05 |
| vb_q3      | 1.9e-06  |
-------------------------
-------------------------
| grad_norm  | 0.0325   |
| loss       | 0.0191   |
| loss_q0    | 0.0636   |
| loss_q1    | 0.00556  |
| loss_q2    | 0.00139  |
| loss_q3    | 6.57e-05 |
| mse        | 0.0169   |
| mse_q0     | 0.0555   |
| mse_q1     | 0.00552  |
| mse_q2     | 0.00138  |
| mse_q3     | 6.5e-05  |
| param_norm | 1.36e+03 |
| samples    | 2.82e+03 |
| step       | 10       |
| vb         | 0.00221  |
| vb_q0      | 0.00808  |
| vb_q1      | 4.09e-05 |
| vb_q2      | 1.21e-05 |
| vb_q3      | 7.78e-07 |
-------------------------
