From dcd659590faac35a1ac36393077f4ab8aac3fea8 Mon Sep 17 00:00:00 2001 From: comfyanonymous <121283862+comfyanonymous@users.noreply.github.com> Date: Wed, 18 Mar 2026 15:14:18 -0700 Subject: [PATCH] Make more intermediate values follow the intermediate dtype. (#13051) --- comfy/sample.py | 4 ++-- comfy/sd1_clip.py | 8 ++++---- 2 files changed, 6 insertions(+), 6 deletions(-) diff --git a/comfy/sample.py b/comfy/sample.py index a2a39b527..e9c2259ab 100644 --- a/comfy/sample.py +++ b/comfy/sample.py @@ -64,10 +64,10 @@ def sample(model, noise, steps, cfg, sampler_name, scheduler, positive, negative sampler = comfy.samplers.KSampler(model, steps=steps, device=model.load_device, sampler=sampler_name, scheduler=scheduler, denoise=denoise, model_options=model.model_options) samples = sampler.sample(noise, positive, negative, cfg=cfg, latent_image=latent_image, start_step=start_step, last_step=last_step, force_full_denoise=force_full_denoise, denoise_mask=noise_mask, sigmas=sigmas, callback=callback, disable_pbar=disable_pbar, seed=seed) - samples = samples.to(comfy.model_management.intermediate_device()) + samples = samples.to(device=comfy.model_management.intermediate_device(), dtype=comfy.model_management.intermediate_dtype()) return samples def sample_custom(model, noise, cfg, sampler, sigmas, positive, negative, latent_image, noise_mask=None, callback=None, disable_pbar=False, seed=None): samples = comfy.samplers.sample(model, noise, positive, negative, cfg, model.load_device, sampler, sigmas, model_options=model.model_options, latent_image=latent_image, denoise_mask=noise_mask, callback=callback, disable_pbar=disable_pbar, seed=seed) - samples = samples.to(comfy.model_management.intermediate_device()) + samples = samples.to(device=comfy.model_management.intermediate_device(), dtype=comfy.model_management.intermediate_dtype()) return samples diff --git a/comfy/sd1_clip.py b/comfy/sd1_clip.py index d89550840..f970510ad 100644 --- a/comfy/sd1_clip.py +++ b/comfy/sd1_clip.py @@ -46,7 +46,7 @@ class ClipTokenWeightEncoder: out, pooled = o[:2] if pooled is not None: - first_pooled = pooled[0:1].to(model_management.intermediate_device()) + first_pooled = pooled[0:1].to(device=model_management.intermediate_device(), dtype=model_management.intermediate_dtype()) else: first_pooled = pooled @@ -63,16 +63,16 @@ class ClipTokenWeightEncoder: output.append(z) if (len(output) == 0): - r = (out[-1:].to(model_management.intermediate_device()), first_pooled) + r = (out[-1:].to(device=model_management.intermediate_device(), dtype=model_management.intermediate_dtype()), first_pooled) else: - r = (torch.cat(output, dim=-2).to(model_management.intermediate_device()), first_pooled) + r = (torch.cat(output, dim=-2).to(device=model_management.intermediate_device(), dtype=model_management.intermediate_dtype()), first_pooled) if len(o) > 2: extra = {} for k in o[2]: v = o[2][k] if k == "attention_mask": - v = v[:sections].flatten().unsqueeze(dim=0).to(model_management.intermediate_device()) + v = v[:sections].flatten().unsqueeze(dim=0).to(device=model_management.intermediate_device(), dtype=model_management.intermediate_dtype()) extra[k] = v r = r + (extra,)