The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
attention: use flag based OOM fallback (#11038)
rattus277237ccc149
diff --git a/comfy/ldm/modules/attention.py b/comfy/ldm/modules/attention.pyindex 7437e0567..a8800ded0 100644--- a/comfy/ldm/modules/attention.py+++ b/comfy/ldm/modules/attention.py@@ -517,6 +517,7 @@ def attention_pytorch(q, k, v, heads, mask=None, attn_precision=None, skip_resha@wrap_attndef attention_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape=False, skip_output_reshape=False, **kwargs):+ exception_fallback = Falseif skip_reshape:b, _, _, dim_head = q.shapetensor_layout = "HND"@@ -541,6 +542,8 @@ def attention_sage(q, k, v, heads, mask=None, attn_precision=None, skip_reshape=out = sageattn(q, k, v, attn_mask=mask, is_causal=False, tensor_layout=tensor_layout)except Exception as e:logging.error("Error running sage attention: {}, using pytorch attention instead.".format(e))+ exception_fallback = True+ if exception_fallback:if tensor_layout == "NHD":q, k, v = map(lambda t: t.transpose(1, 2),diff --git a/comfy/ldm/modules/diffusionmodules/model.py b/comfy/ldm/modules/diffusionmodules/model.pyindex 4245eedca..de1e01cc8 100644--- a/comfy/ldm/modules/diffusionmodules/model.py+++ b/comfy/ldm/modules/diffusionmodules/model.py@@ -279,6 +279,7 @@ def pytorch_attention(q, k, v):orig_shape = q.shapeB = orig_shape[0]C = orig_shape[1]+ oom_fallback = Falseq, k, v = map(lambda t: t.view(B, 1, C, -1).transpose(2, 3).contiguous(),(q, k, v),@@ -289,6 +290,8 @@ def pytorch_attention(q, k, v):out = out.transpose(2, 3).reshape(orig_shape)except model_management.OOM_EXCEPTION:logging.warning("scaled_dot_product_attention OOMed: switched to slice attention")+ oom_fallback = True+ if oom_fallback:out = slice_attention(q.view(B, -1, C), k.view(B, -1, C).transpose(1, 2), v.view(B, -1, C).transpose(1, 2)).reshape(orig_shape)return out