cldm.py下的sample_log函数

[xpixelgroup]

cldm.py下的sample_log函数

[0x3f3f3f3fun]

这个是传递的参数有问题~ 正确的代码如下：

@torch.no_grad()
def log_images(self, batch, sample_steps=50):
    log = dict()
    z, c = self.get_input(batch, self.first_stage_key)
    c_lq = c["lq"][0]
    c_latent = c["c_latent"][0]
    c_cat, c = c["c_concat"][0], c["c_crossattn"][0]

    log["hq"] = (self.decode_first_stage(z) + 1) / 2
    log["control"] = c_cat
    log["decoded_control"] = (self.decode_first_stage(c_latent) + 1) / 2
    log["lq"] = c_lq
    log["text"] = (log_txt_as_img((512, 512), batch[self.cond_stage_key], size=16) + 1) / 2
    log["samples"] = self.sample_log(
        cond_img=c_cat,            
        steps=sample_steps
    )
    return log

@torch.no_grad()
def sample_log(self, cond_img, steps):
    sampler = SpacedSampler(self)
    b, c, h, w = cond_img.shape
    shape = (b, self.channels, h // 8, w // 8)
    samples = sampler.sample(
        steps, shape, cond_img, positive_prompt="", negative_prompt="",
        cfg_scale=1.0, color_fix_type="wavelet"
    )
    return samples

最近有点忙，所以没时间更新代码，抱歉。

[yimingz1218]

我也碰到了这个问题，请问你找到了解决方法吗