(self, x, c, t, clip_denoised: bool, return_codebook_ids=False, quantize_denoised=False,
return_x0=False, score_corrector=None, corrector_kwargs=None)
| 928 | return loss, loss_dict |
| 929 | |
| 930 | def p_mean_variance(self, x, c, t, clip_denoised: bool, return_codebook_ids=False, quantize_denoised=False, |
| 931 | return_x0=False, score_corrector=None, corrector_kwargs=None): |
| 932 | t_in = t |
| 933 | model_out = self.apply_model(x, t_in, c, return_ids=return_codebook_ids) |
| 934 | |
| 935 | if score_corrector is not None: |
| 936 | assert self.parameterization == "eps" |
| 937 | model_out = score_corrector.modify_score(self, model_out, x, t, c, **corrector_kwargs) |
| 938 | |
| 939 | if return_codebook_ids: |
| 940 | model_out, logits = model_out |
| 941 | |
| 942 | if self.parameterization == "eps": |
| 943 | x_recon = self.predict_start_from_noise(x, t=t, noise=model_out) |
| 944 | elif self.parameterization == "x0": |
| 945 | x_recon = model_out |
| 946 | else: |
| 947 | raise NotImplementedError() |
| 948 | |
| 949 | if clip_denoised: |
| 950 | x_recon.clamp_(-1., 1.) |
| 951 | if quantize_denoised: |
| 952 | x_recon, _, [_, _, indices] = self.first_stage_model.quantize(x_recon) |
| 953 | model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t) |
| 954 | if return_codebook_ids: |
| 955 | return model_mean, posterior_variance, posterior_log_variance, logits |
| 956 | elif return_x0: |
| 957 | return model_mean, posterior_variance, posterior_log_variance, x_recon |
| 958 | else: |
| 959 | return model_mean, posterior_variance, posterior_log_variance |
| 960 | |
| 961 | @torch.no_grad() |
| 962 | def p_sample(self, x, c, t, clip_denoised=False, repeat_noise=False, |
no test coverage detected