sparse_only_stage: | |
pruning_modifiers: | |
SparseGPTModifier: | |
sparsity: 0.5 | |
sequential_update: true | |
mask_structure: '2:4' | |
targets: ['re:model.layers.\d*$'] | |
quant_stage: | |
quant_modifiers: | |
QuantizationModifier: | |
ignore: [lm_head] | |
targets: [Linear] | |
scheme: FP8_DYNAMIC | |
constant_pruning_modifiers: | |
ConstantPruningModifier: | |
targets: ['re:.*q_proj.weight', 're:.*k_proj.weight', 're:.*v_proj.weight', 're:.*o_proj.weight', | |
're:.*gate_proj.weight', 're:.*up_proj.weight', 're:.*down_proj.weight'] | |
start: 0 | |