Minimax H3 Multi Ref workflow for low VRAM

Minor update in V3.4<ol><li>recontruct the attention part of model subgraph.</li><li>add sol attn into workflow.</li><li>recontruct the front panel of model subgraph</li><li>add download link of lightx2v lora in markdown.</li><li>renew the attention combination</li><li>use Chunk FeedForward to replace Low VRAM attention</li></ol>Remark:<ol><li>Make sure you had SWITCH OFF ACC Lora if you chose normal Turbo Lora, vice versa.</li><li>Beware you had to choose either (Sage attention + pytouch attention) or kitchen attention + Specturn, Sage attention and kitchen attention could not use together, SWITCH OFF those which are not using.</li><li>disable the int8_qk in Sol attn patch if you feel too much noise.</li><li>Sol attention/Fused Modulation/Chunk Feed Forward/Spectrum could be compatible with Sage attention and Kitchen attention.</li><li>Sol attention now support 30 series RTX display card.</li></ol>

minimaxH3MultiRef_v34.json

v3.4

v3.3

Files with SHA256: 7dac2fb1c11813125934d584c97ca936eac4347618118d675b74af3b2f2cdc85