| apiversion: multimodal_vlm_modelslim_v1 | |
| spec: | |
| process: | |
| - type: group | |
| configs: | |
| - type: linear_quant | |
| qconfig: | |
| act: | |
| scope: per_token | |
| dtype: int8 | |
| symmetric: true | |
| method: minmax | |
| weight: | |
| scope: per_channel | |
| dtype: int8 | |
| symmetric: true | |
| method: minmax | |
| include: | |
| - '*self_attn*' | |
| exclude: | |
| - '*self_attn.o_proj*' | |
| - mtp* | |
| - type: linear_quant | |
| qconfig: | |
| act: | |
| scope: per_token | |
| dtype: int8 | |
| symmetric: true | |
| method: minmax | |
| weight: | |
| scope: per_channel | |
| dtype: int8 | |
| symmetric: true | |
| method: minmax | |
| include: | |
| - '*mlp.experts*' | |
| exclude: | |
| - mtp* | |
| save: | |
| - type: ascendv1_saver | |
| part_file_size: 4 | |
| dataset: calibImages | |
| default_text: Describe this image in detail. | |
| metadata: | |
| config_id: Qwen3_5_moe_w8a8 | |
| score: 90.0 | |
| label: | |
| w_bit: 8 | |
| a_bit: 8 | |
| is_sparse: false | |
| kv_cache: false | |
| verified_model_types: | |
| - Qwen3.5-397B-A17B | |
| - Qwen3.5-122B-A10B | |
| - Qwen3.5-35B-A3B | |
| verified_tags: {} | |
| default_w8a8_dynamic: | |
| act: | |
| scope: per_token | |
| dtype: int8 | |
| symmetric: true | |
| method: minmax | |
| weight: | |
| scope: per_channel | |
| dtype: int8 | |
| symmetric: true | |
| method: minmax | |