ComfyUI Node
Diffusion Model Loader Resident
KJ-compatible diffusion-model loader with GPU-resident ingest and residency tracking. It mirrors the KJ node's weight dtype, compute dtype, cublas_ops, SageAttention, fp16 accumulation, and optional extra-state-dict features.
Diffusion Model Loader Resident
- MODEL
◄model_name▾►
◄weight_dtype▾►
◄compute_dtypedefault►
◄patch_cublaslinearfalse►
◄sage_attentiondisabled►
◄enable_fp16_accumulationfalse►
◄extra_state_dict—►
◄policy_override—►
CategoryGPU Resident Loader/loaders
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| model_name | COMBO | The diffusion model file to load. | |
| weight_dtype | COMBO | 7 options: default, fp8_e4m3fn, fp8_e4m3fn_fast, fp8_e5m2, fp16, bf16, +1 | |
| compute_dtype | COMBO | default | Compute dtype to apply after model creation. |
| patch_cublaslinear | BOOLEAN | false | Toggle ComfyUI's cublas_ops performance feature. |
| sage_attention | COMBO | disabled | Patch optimized attention override to a SageAttention variant. |
| enable_fp16_accumulation | BOOLEAN | false | Set torch.backends.cuda.matmul.allow_fp16_accumulation. |
| extra_state_dictopt | STRING | Optional absolute path to a second state dict merged into the main diffusion state dict before model detection. | |
| policy_overrideopt | STRING | Optional residency policy override. Connect Set Global Residency Policy here, not to extra_state_dict. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| MODEL | MODEL | — |