llada-instruct-iterative-30-snr / pruning_results.json
BashCache's picture
Upload folder using huggingface_hub
aae4f27 verified
Raw History Blame Contribute Delete
14.4 kB
{
"baseline": {},
"pruned": {
"perplexity_c4": 29.359903876589225,
"acc_arc_easy": 0.6839225589225589,
"acc_arc_challenge": 0.4061433447098976,
"acc_winogrande": 0.6274664561957379,
"acc_piqa": 0.7225244831338411,
"sparsity": {
"linear": 0.30720170454545453,
"global": 0.287338988793155,
"per_layer": {
"model.transformer.blocks.0.attn_out": 0.75146484375,
"model.transformer.blocks.0.ff_out": 0.940673828125,
"model.transformer.blocks.0.q_proj": 0.75146484375,
"model.transformer.blocks.0.k_proj": 0.75146484375,
"model.transformer.blocks.0.v_proj": 0.75146484375,
"model.transformer.blocks.0.ff_proj": 0.940673828125,
"model.transformer.blocks.0.up_proj": 0.940673828125,
"model.transformer.blocks.1.attn_out": 0.76171875,
"model.transformer.blocks.1.ff_out": 0.8665364583333334,
"model.transformer.blocks.1.q_proj": 0.76171875,
"model.transformer.blocks.1.k_proj": 0.76171875,
"model.transformer.blocks.1.v_proj": 0.76171875,
"model.transformer.blocks.1.ff_proj": 0.8665364583333334,
"model.transformer.blocks.1.up_proj": 0.8665364583333334,
"model.transformer.blocks.2.attn_out": 0.703125,
"model.transformer.blocks.2.ff_out": 0.7859700520833334,
"model.transformer.blocks.2.q_proj": 0.703125,
"model.transformer.blocks.2.k_proj": 0.703125,
"model.transformer.blocks.2.v_proj": 0.703125,
"model.transformer.blocks.2.ff_proj": 0.7859700520833334,
"model.transformer.blocks.2.up_proj": 0.7859700520833334,
"model.transformer.blocks.3.attn_out": 0.67333984375,
"model.transformer.blocks.3.ff_out": 0.6283365885416666,
"model.transformer.blocks.3.q_proj": 0.67333984375,
"model.transformer.blocks.3.k_proj": 0.67333984375,
"model.transformer.blocks.3.v_proj": 0.67333984375,
"model.transformer.blocks.3.ff_proj": 0.6283365885416666,
"model.transformer.blocks.3.up_proj": 0.6283365885416666,
"model.transformer.blocks.4.attn_out": 0.60888671875,
"model.transformer.blocks.4.ff_out": 0.4276529947916667,
"model.transformer.blocks.4.q_proj": 0.60888671875,
"model.transformer.blocks.4.k_proj": 0.60888671875,
"model.transformer.blocks.4.v_proj": 0.60888671875,
"model.transformer.blocks.4.ff_proj": 0.4276529947916667,
"model.transformer.blocks.4.up_proj": 0.4276529947916667,
"model.transformer.blocks.5.attn_out": 0.4619140625,
"model.transformer.blocks.5.ff_out": 0.2784016927083333,
"model.transformer.blocks.5.q_proj": 0.4619140625,
"model.transformer.blocks.5.k_proj": 0.4619140625,
"model.transformer.blocks.5.v_proj": 0.4619140625,
"model.transformer.blocks.5.ff_proj": 0.2784016927083333,
"model.transformer.blocks.5.up_proj": 0.2784016927083333,
"model.transformer.blocks.6.attn_out": 0.474609375,
"model.transformer.blocks.6.ff_out": 0.2992350260416667,
"model.transformer.blocks.6.q_proj": 0.474609375,
"model.transformer.blocks.6.k_proj": 0.474609375,
"model.transformer.blocks.6.v_proj": 0.474609375,
"model.transformer.blocks.6.ff_proj": 0.2992350260416667,
"model.transformer.blocks.6.up_proj": 0.2992350260416667,
"model.transformer.blocks.7.attn_out": 0.26318359375,
"model.transformer.blocks.7.ff_out": 0.20450846354166666,
"model.transformer.blocks.7.q_proj": 0.26318359375,
"model.transformer.blocks.7.k_proj": 0.26318359375,
"model.transformer.blocks.7.v_proj": 0.26318359375,
"model.transformer.blocks.7.ff_proj": 0.20450846354166666,
"model.transformer.blocks.7.up_proj": 0.20450846354166666,
"model.transformer.blocks.8.attn_out": 0.321533203125,
"model.transformer.blocks.8.ff_out": 0.22574869791666666,
"model.transformer.blocks.8.q_proj": 0.321533203125,
"model.transformer.blocks.8.k_proj": 0.321533203125,
"model.transformer.blocks.8.v_proj": 0.321533203125,
"model.transformer.blocks.8.ff_proj": 0.22574869791666666,
"model.transformer.blocks.8.up_proj": 0.22574869791666666,
"model.transformer.blocks.9.attn_out": 0.343994140625,
"model.transformer.blocks.9.ff_out": 0.2747395833333333,
"model.transformer.blocks.9.q_proj": 0.343994140625,
"model.transformer.blocks.9.k_proj": 0.343994140625,
"model.transformer.blocks.9.v_proj": 0.343994140625,
"model.transformer.blocks.9.ff_proj": 0.2747395833333333,
"model.transformer.blocks.9.up_proj": 0.2747395833333333,
"model.transformer.blocks.10.attn_out": 0.327392578125,
"model.transformer.blocks.10.ff_out": 0.2941080729166667,
"model.transformer.blocks.10.q_proj": 0.327392578125,
"model.transformer.blocks.10.k_proj": 0.327392578125,
"model.transformer.blocks.10.v_proj": 0.327392578125,
"model.transformer.blocks.10.ff_proj": 0.2941080729166667,
"model.transformer.blocks.10.up_proj": 0.2941080729166667,
"model.transformer.blocks.11.attn_out": 0.33154296875,
"model.transformer.blocks.11.ff_out": 0.2783203125,
"model.transformer.blocks.11.q_proj": 0.33154296875,
"model.transformer.blocks.11.k_proj": 0.33154296875,
"model.transformer.blocks.11.v_proj": 0.33154296875,
"model.transformer.blocks.11.ff_proj": 0.2783203125,
"model.transformer.blocks.11.up_proj": 0.2783203125,
"model.transformer.blocks.12.attn_out": 0.275146484375,
"model.transformer.blocks.12.ff_out": 0.2825520833333333,
"model.transformer.blocks.12.q_proj": 0.275146484375,
"model.transformer.blocks.12.k_proj": 0.275146484375,
"model.transformer.blocks.12.v_proj": 0.275146484375,
"model.transformer.blocks.12.ff_proj": 0.2825520833333333,
"model.transformer.blocks.12.up_proj": 0.2825520833333333,
"model.transformer.blocks.13.attn_out": 0.178955078125,
"model.transformer.blocks.13.ff_out": 0.221923828125,
"model.transformer.blocks.13.q_proj": 0.178955078125,
"model.transformer.blocks.13.k_proj": 0.178955078125,
"model.transformer.blocks.13.v_proj": 0.178955078125,
"model.transformer.blocks.13.ff_proj": 0.221923828125,
"model.transformer.blocks.13.up_proj": 0.221923828125,
"model.transformer.blocks.14.attn_out": 0.2978515625,
"model.transformer.blocks.14.ff_out": 0.240966796875,
"model.transformer.blocks.14.q_proj": 0.2978515625,
"model.transformer.blocks.14.k_proj": 0.2978515625,
"model.transformer.blocks.14.v_proj": 0.2978515625,
"model.transformer.blocks.14.ff_proj": 0.240966796875,
"model.transformer.blocks.14.up_proj": 0.240966796875,
"model.transformer.blocks.15.attn_out": 0.2509765625,
"model.transformer.blocks.15.ff_out": 0.2666829427083333,
"model.transformer.blocks.15.q_proj": 0.2509765625,
"model.transformer.blocks.15.k_proj": 0.2509765625,
"model.transformer.blocks.15.v_proj": 0.2509765625,
"model.transformer.blocks.15.ff_proj": 0.2666829427083333,
"model.transformer.blocks.15.up_proj": 0.2666829427083333,
"model.transformer.blocks.16.attn_out": 0.147216796875,
"model.transformer.blocks.16.ff_out": 0.2578125,
"model.transformer.blocks.16.q_proj": 0.147216796875,
"model.transformer.blocks.16.k_proj": 0.147216796875,
"model.transformer.blocks.16.v_proj": 0.147216796875,
"model.transformer.blocks.16.ff_proj": 0.2578125,
"model.transformer.blocks.16.up_proj": 0.2578125,
"model.transformer.blocks.17.attn_out": 0.19189453125,
"model.transformer.blocks.17.ff_out": 0.24267578125,
"model.transformer.blocks.17.q_proj": 0.19189453125,
"model.transformer.blocks.17.k_proj": 0.19189453125,
"model.transformer.blocks.17.v_proj": 0.19189453125,
"model.transformer.blocks.17.ff_proj": 0.24267578125,
"model.transformer.blocks.17.up_proj": 0.24267578125,
"model.transformer.blocks.18.attn_out": 0.247802734375,
"model.transformer.blocks.18.ff_out": 0.24894205729166666,
"model.transformer.blocks.18.q_proj": 0.247802734375,
"model.transformer.blocks.18.k_proj": 0.247802734375,
"model.transformer.blocks.18.v_proj": 0.247802734375,
"model.transformer.blocks.18.ff_proj": 0.24894205729166666,
"model.transformer.blocks.18.up_proj": 0.24894205729166666,
"model.transformer.blocks.19.attn_out": 0.152587890625,
"model.transformer.blocks.19.ff_out": 0.21883138020833334,
"model.transformer.blocks.19.q_proj": 0.152587890625,
"model.transformer.blocks.19.k_proj": 0.152587890625,
"model.transformer.blocks.19.v_proj": 0.152587890625,
"model.transformer.blocks.19.ff_proj": 0.21883138020833334,
"model.transformer.blocks.19.up_proj": 0.21883138020833334,
"model.transformer.blocks.20.attn_out": 0.139892578125,
"model.transformer.blocks.20.ff_out": 0.231689453125,
"model.transformer.blocks.20.q_proj": 0.139892578125,
"model.transformer.blocks.20.k_proj": 0.139892578125,
"model.transformer.blocks.20.v_proj": 0.139892578125,
"model.transformer.blocks.20.ff_proj": 0.231689453125,
"model.transformer.blocks.20.up_proj": 0.231689453125,
"model.transformer.blocks.21.attn_out": 0.122802734375,
"model.transformer.blocks.21.ff_out": 0.21663411458333334,
"model.transformer.blocks.21.q_proj": 0.122802734375,
"model.transformer.blocks.21.k_proj": 0.122802734375,
"model.transformer.blocks.21.v_proj": 0.122802734375,
"model.transformer.blocks.21.ff_proj": 0.21663411458333334,
"model.transformer.blocks.21.up_proj": 0.21663411458333334,
"model.transformer.blocks.22.attn_out": 0.088134765625,
"model.transformer.blocks.22.ff_out": 0.21630859375,
"model.transformer.blocks.22.q_proj": 0.088134765625,
"model.transformer.blocks.22.k_proj": 0.088134765625,
"model.transformer.blocks.22.v_proj": 0.088134765625,
"model.transformer.blocks.22.ff_proj": 0.21630859375,
"model.transformer.blocks.22.up_proj": 0.21630859375,
"model.transformer.blocks.23.attn_out": 0.14111328125,
"model.transformer.blocks.23.ff_out": 0.156494140625,
"model.transformer.blocks.23.q_proj": 0.14111328125,
"model.transformer.blocks.23.k_proj": 0.14111328125,
"model.transformer.blocks.23.v_proj": 0.14111328125,
"model.transformer.blocks.23.ff_proj": 0.156494140625,
"model.transformer.blocks.23.up_proj": 0.156494140625,
"model.transformer.blocks.24.attn_out": 0.0703125,
"model.transformer.blocks.24.ff_out": 0.14469401041666666,
"model.transformer.blocks.24.q_proj": 0.0703125,
"model.transformer.blocks.24.k_proj": 0.0703125,
"model.transformer.blocks.24.v_proj": 0.0703125,
"model.transformer.blocks.24.ff_proj": 0.14469401041666666,
"model.transformer.blocks.24.up_proj": 0.14469401041666666,
"model.transformer.blocks.25.attn_out": 0.15380859375,
"model.transformer.blocks.25.ff_out": 0.23209635416666666,
"model.transformer.blocks.25.q_proj": 0.15380859375,
"model.transformer.blocks.25.k_proj": 0.15380859375,
"model.transformer.blocks.25.v_proj": 0.15380859375,
"model.transformer.blocks.25.ff_proj": 0.23209635416666666,
"model.transformer.blocks.25.up_proj": 0.23209635416666666,
"model.transformer.blocks.26.attn_out": 0.134765625,
"model.transformer.blocks.26.ff_out": 0.27685546875,
"model.transformer.blocks.26.q_proj": 0.134765625,
"model.transformer.blocks.26.k_proj": 0.134765625,
"model.transformer.blocks.26.v_proj": 0.134765625,
"model.transformer.blocks.26.ff_proj": 0.27685546875,
"model.transformer.blocks.26.up_proj": 0.27685546875,
"model.transformer.blocks.27.attn_out": 0.291015625,
"model.transformer.blocks.27.ff_out": 0.3339029947916667,
"model.transformer.blocks.27.q_proj": 0.291015625,
"model.transformer.blocks.27.k_proj": 0.291015625,
"model.transformer.blocks.27.v_proj": 0.291015625,
"model.transformer.blocks.27.ff_proj": 0.3339029947916667,
"model.transformer.blocks.27.up_proj": 0.3339029947916667,
"model.transformer.blocks.28.attn_out": 0.473876953125,
"model.transformer.blocks.28.ff_out": 0.3470865885416667,
"model.transformer.blocks.28.q_proj": 0.473876953125,
"model.transformer.blocks.28.k_proj": 0.473876953125,
"model.transformer.blocks.28.v_proj": 0.473876953125,
"model.transformer.blocks.28.ff_proj": 0.3470865885416667,
"model.transformer.blocks.28.up_proj": 0.3470865885416667,
"model.transformer.blocks.29.attn_out": 0.5244140625,
"model.transformer.blocks.29.ff_out": 0.319091796875,
"model.transformer.blocks.29.q_proj": 0.5244140625,
"model.transformer.blocks.29.k_proj": 0.5244140625,
"model.transformer.blocks.29.v_proj": 0.5244140625,
"model.transformer.blocks.29.ff_proj": 0.319091796875,
"model.transformer.blocks.29.up_proj": 0.319091796875,
"model.transformer.blocks.30.attn_out": 0.471435546875,
"model.transformer.blocks.30.ff_out": 0.288330078125,
"model.transformer.blocks.30.q_proj": 0.471435546875,
"model.transformer.blocks.30.k_proj": 0.471435546875,
"model.transformer.blocks.30.v_proj": 0.471435546875,
"model.transformer.blocks.30.ff_proj": 0.288330078125,
"model.transformer.blocks.30.up_proj": 0.288330078125,
"model.transformer.blocks.31.attn_out": 0.183349609375,
"model.transformer.blocks.31.ff_out": 0.312255859375,
"model.transformer.blocks.31.q_proj": 0.183349609375,
"model.transformer.blocks.31.k_proj": 0.183349609375,
"model.transformer.blocks.31.v_proj": 0.183349609375,
"model.transformer.blocks.31.ff_proj": 0.312255859375,
"model.transformer.blocks.31.up_proj": 0.312255859375,
"model.transformer.ff_out": 0.0
}
}
}
}