ComfyUI Node
Load MiniCPM-V Model
A ComfyUI node in orama/MiniCPM-V with 5 inputs and 1 output.
Load MiniCPM-V Model
- model
◄model_idopenbmb/MiniCPM-V-4.6►
◄dtypeauto►
◄device_mapauto►
◄use_flash_attention_2false►
◄visual_token_compression16x (fast)►
Categoryorama/MiniCPM-V
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| model_id | COMBO | openbmb/MiniCPM-V-4.6 | 2 options: openbmb/MiniCPM-V-4.6, huihui-ai/Huihui-MiniCPM-V-4.6-abliterated |
| dtype | COMBO | auto | 4 options: auto, bfloat16, float16, float32 |
| device_map | COMBO | auto | 4 options: auto, cuda, cpu, mps |
| use_flash_attention_2 | BOOLEAN | false | Recommended for multi-image tasks. Requires flash-attn package. |
| visual_token_compression | COMBO | 16x (fast) | 16x = fewer tokens, faster. 4x = more detail. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| model | MINICPM_V_MODEL | — |