thoratsr7's picture
Duplicate from CohereLabs/North-Mini-Code-1.0-fp8
ab98315
Raw
History Blame Contribute Delete
320 Bytes
default_stage:
default_modifiers:
QuantizationModifier:
targets: [Linear]
ignore: ['re:.*lm_head', 're:.*layers\.(?:0)(?:\.|$)', 're:model.multi_modal_projector.*',
're:model.vision_tower.*', 're:.*mlp.gate$', 're:.*self_attn']
scheme: FP8_DYNAMIC
bypass_divisibility_checks: false