BananaMind-2-Pro-Preview-Chat-ternary / quantization_report.json
fontlab's picture
Add a quantized checkpoint
b9df603 verified
Raw
History Blame Contribute Delete
51 kB
{
"summary": {
"group_size": 64,
"tensors_ternary": 168,
"tensors_int8": 1,
"tensors_total": 266,
"tensors_quantized": 169,
"float_mb": 555.89,
"result_mb": 66.66,
"ratio": 8.34,
"worst_relative_error": 1.8704,
"mean_relative_error": 0.5312,
"mean_zero_fraction": 0.4473,
"method": "ternary",
"quality": {
"tokens_compared": 63,
"top1_agreement": 0.4127,
"top5_agreement": 0.6667,
"mean_kl_divergence": 1.4209,
"reference_perplexity": 38.325,
"candidate_perplexity": 131.063,
"perplexity_ratio": 3.42,
"greedy_identical_prefix": [
0,
0,
0,
0,
0
],
"greedy_identical_answers": "0/8"
}
},
"tensors": [
{
"name": "transformer.h.0.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.0.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52544,
"zero_fraction": 0.44605
},
{
"name": "transformer.h.0.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.58326,
"zero_fraction": 0.45928
},
{
"name": "transformer.h.0.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.0.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.54151,
"zero_fraction": 0.44389
},
{
"name": "transformer.h.0.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51277,
"zero_fraction": 0.45881
},
{
"name": "transformer.h.0.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.0.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.0.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.57151,
"zero_fraction": 0.46311
},
{
"name": "transformer.h.0.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53645,
"zero_fraction": 0.44798
},
{
"name": "transformer.h.0.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52619,
"zero_fraction": 0.4502
},
{
"name": "transformer.h.1.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.1.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.68262,
"zero_fraction": 0.44455
},
{
"name": "transformer.h.1.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53616,
"zero_fraction": 0.44108
},
{
"name": "transformer.h.1.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.1.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.67879,
"zero_fraction": 0.44681
},
{
"name": "transformer.h.1.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51314,
"zero_fraction": 0.45068
},
{
"name": "transformer.h.1.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.1.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.1.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.6958,
"zero_fraction": 0.45045
},
{
"name": "transformer.h.1.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53017,
"zero_fraction": 0.4457
},
{
"name": "transformer.h.1.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.5231,
"zero_fraction": 0.45012
},
{
"name": "transformer.h.10.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.10.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.56254,
"zero_fraction": 0.44968
},
{
"name": "transformer.h.10.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52379,
"zero_fraction": 0.44042
},
{
"name": "transformer.h.10.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.10.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.5291,
"zero_fraction": 0.4485
},
{
"name": "transformer.h.10.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49261,
"zero_fraction": 0.45534
},
{
"name": "transformer.h.10.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.10.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.10.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.58624,
"zero_fraction": 0.45217
},
{
"name": "transformer.h.10.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50431,
"zero_fraction": 0.45001
},
{
"name": "transformer.h.10.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50098,
"zero_fraction": 0.44821
},
{
"name": "transformer.h.11.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.11.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.63709,
"zero_fraction": 0.44739
},
{
"name": "transformer.h.11.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50918,
"zero_fraction": 0.44098
},
{
"name": "transformer.h.11.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.11.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.55463,
"zero_fraction": 0.44762
},
{
"name": "transformer.h.11.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49221,
"zero_fraction": 0.45365
},
{
"name": "transformer.h.11.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.11.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.11.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 1.87045,
"zero_fraction": 0.45876
},
{
"name": "transformer.h.11.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50126,
"zero_fraction": 0.44869
},
{
"name": "transformer.h.11.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50149,
"zero_fraction": 0.45038
},
{
"name": "transformer.h.12.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.12.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.54634,
"zero_fraction": 0.4489
},
{
"name": "transformer.h.12.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51433,
"zero_fraction": 0.4421
},
{
"name": "transformer.h.12.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.12.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.54998,
"zero_fraction": 0.44856
},
{
"name": "transformer.h.12.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49165,
"zero_fraction": 0.45395
},
{
"name": "transformer.h.12.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.12.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.12.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.53657,
"zero_fraction": 0.45076
},
{
"name": "transformer.h.12.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50015,
"zero_fraction": 0.448
},
{
"name": "transformer.h.12.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49888,
"zero_fraction": 0.44753
},
{
"name": "transformer.h.13.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.13.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51966,
"zero_fraction": 0.4542
},
{
"name": "transformer.h.13.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50759,
"zero_fraction": 0.44081
},
{
"name": "transformer.h.13.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.13.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53785,
"zero_fraction": 0.45039
},
{
"name": "transformer.h.13.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.48911,
"zero_fraction": 0.45612
},
{
"name": "transformer.h.13.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.13.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.13.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54092,
"zero_fraction": 0.45187
},
{
"name": "transformer.h.13.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50146,
"zero_fraction": 0.44808
},
{
"name": "transformer.h.13.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50046,
"zero_fraction": 0.44936
},
{
"name": "transformer.h.14.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.14.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.55382,
"zero_fraction": 0.45092
},
{
"name": "transformer.h.14.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51405,
"zero_fraction": 0.44356
},
{
"name": "transformer.h.14.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.14.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.54023,
"zero_fraction": 0.44899
},
{
"name": "transformer.h.14.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.48846,
"zero_fraction": 0.4538
},
{
"name": "transformer.h.14.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.14.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.14.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.53874,
"zero_fraction": 0.45206
},
{
"name": "transformer.h.14.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50055,
"zero_fraction": 0.44787
},
{
"name": "transformer.h.14.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49837,
"zero_fraction": 0.45022
},
{
"name": "transformer.h.15.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.15.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51955,
"zero_fraction": 0.45194
},
{
"name": "transformer.h.15.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.5129,
"zero_fraction": 0.4436
},
{
"name": "transformer.h.15.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.15.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53742,
"zero_fraction": 0.44888
},
{
"name": "transformer.h.15.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.48876,
"zero_fraction": 0.45426
},
{
"name": "transformer.h.15.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.15.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.15.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54381,
"zero_fraction": 0.45291
},
{
"name": "transformer.h.15.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50612,
"zero_fraction": 0.45066
},
{
"name": "transformer.h.15.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.5027,
"zero_fraction": 0.45157
},
{
"name": "transformer.h.16.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.16.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.54182,
"zero_fraction": 0.45058
},
{
"name": "transformer.h.16.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51856,
"zero_fraction": 0.44282
},
{
"name": "transformer.h.16.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.16.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53377,
"zero_fraction": 0.44859
},
{
"name": "transformer.h.16.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.48916,
"zero_fraction": 0.45847
},
{
"name": "transformer.h.16.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.16.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.16.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.55314,
"zero_fraction": 0.4547
},
{
"name": "transformer.h.16.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50235,
"zero_fraction": 0.4504
},
{
"name": "transformer.h.16.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49997,
"zero_fraction": 0.45177
},
{
"name": "transformer.h.17.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.17.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53927,
"zero_fraction": 0.46846
},
{
"name": "transformer.h.17.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52639,
"zero_fraction": 0.44798
},
{
"name": "transformer.h.17.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.17.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.54907,
"zero_fraction": 0.46303
},
{
"name": "transformer.h.17.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51894,
"zero_fraction": 0.47726
},
{
"name": "transformer.h.17.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.17.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.17.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54836,
"zero_fraction": 0.45213
},
{
"name": "transformer.h.17.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50459,
"zero_fraction": 0.44843
},
{
"name": "transformer.h.17.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49796,
"zero_fraction": 0.44906
},
{
"name": "transformer.h.18.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.18.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51919,
"zero_fraction": 0.44947
},
{
"name": "transformer.h.18.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52167,
"zero_fraction": 0.44327
},
{
"name": "transformer.h.18.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.18.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53997,
"zero_fraction": 0.45149
},
{
"name": "transformer.h.18.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49286,
"zero_fraction": 0.45904
},
{
"name": "transformer.h.18.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.18.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.18.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54553,
"zero_fraction": 0.45147
},
{
"name": "transformer.h.18.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50114,
"zero_fraction": 0.44643
},
{
"name": "transformer.h.18.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49899,
"zero_fraction": 0.44907
},
{
"name": "transformer.h.19.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.19.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52591,
"zero_fraction": 0.45297
},
{
"name": "transformer.h.19.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52533,
"zero_fraction": 0.44801
},
{
"name": "transformer.h.19.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.19.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53258,
"zero_fraction": 0.45617
},
{
"name": "transformer.h.19.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50129,
"zero_fraction": 0.4608
},
{
"name": "transformer.h.19.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.19.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.19.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54184,
"zero_fraction": 0.45056
},
{
"name": "transformer.h.19.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49854,
"zero_fraction": 0.44656
},
{
"name": "transformer.h.19.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49828,
"zero_fraction": 0.44865
},
{
"name": "transformer.h.2.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.2.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.5935,
"zero_fraction": 0.4504
},
{
"name": "transformer.h.2.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52771,
"zero_fraction": 0.43996
},
{
"name": "transformer.h.2.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.2.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.58501,
"zero_fraction": 0.44924
},
{
"name": "transformer.h.2.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50744,
"zero_fraction": 0.45421
},
{
"name": "transformer.h.2.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.2.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.2.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.5635,
"zero_fraction": 0.45022
},
{
"name": "transformer.h.2.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52057,
"zero_fraction": 0.44746
},
{
"name": "transformer.h.2.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51459,
"zero_fraction": 0.44995
},
{
"name": "transformer.h.20.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.20.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.5241,
"zero_fraction": 0.45346
},
{
"name": "transformer.h.20.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52259,
"zero_fraction": 0.44242
},
{
"name": "transformer.h.20.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.20.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.54122,
"zero_fraction": 0.45545
},
{
"name": "transformer.h.20.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.497,
"zero_fraction": 0.46026
},
{
"name": "transformer.h.20.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.20.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.20.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.53824,
"zero_fraction": 0.4495
},
{
"name": "transformer.h.20.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50115,
"zero_fraction": 0.44605
},
{
"name": "transformer.h.20.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49922,
"zero_fraction": 0.44725
},
{
"name": "transformer.h.21.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.21.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51615,
"zero_fraction": 0.45089
},
{
"name": "transformer.h.21.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51021,
"zero_fraction": 0.44634
},
{
"name": "transformer.h.21.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.21.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53312,
"zero_fraction": 0.45279
},
{
"name": "transformer.h.21.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49333,
"zero_fraction": 0.45693
},
{
"name": "transformer.h.21.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.21.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.21.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54052,
"zero_fraction": 0.45054
},
{
"name": "transformer.h.21.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50032,
"zero_fraction": 0.44563
},
{
"name": "transformer.h.21.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49858,
"zero_fraction": 0.44714
},
{
"name": "transformer.h.22.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.22.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52323,
"zero_fraction": 0.4506
},
{
"name": "transformer.h.22.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50313,
"zero_fraction": 0.44684
},
{
"name": "transformer.h.22.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.22.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53909,
"zero_fraction": 0.44854
},
{
"name": "transformer.h.22.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49226,
"zero_fraction": 0.45206
},
{
"name": "transformer.h.22.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.22.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.22.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.57357,
"zero_fraction": 0.45229
},
{
"name": "transformer.h.22.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50022,
"zero_fraction": 0.44722
},
{
"name": "transformer.h.22.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49896,
"zero_fraction": 0.44802
},
{
"name": "transformer.h.23.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.23.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51427,
"zero_fraction": 0.4504
},
{
"name": "transformer.h.23.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50223,
"zero_fraction": 0.45219
},
{
"name": "transformer.h.23.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.23.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52447,
"zero_fraction": 0.45045
},
{
"name": "transformer.h.23.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49304,
"zero_fraction": 0.45653
},
{
"name": "transformer.h.23.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.23.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.23.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.60155,
"zero_fraction": 0.4593
},
{
"name": "transformer.h.23.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49875,
"zero_fraction": 0.45031
},
{
"name": "transformer.h.23.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50014,
"zero_fraction": 0.44877
},
{
"name": "transformer.h.3.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.3.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.64971,
"zero_fraction": 0.44776
},
{
"name": "transformer.h.3.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53296,
"zero_fraction": 0.43921
},
{
"name": "transformer.h.3.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.3.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.56119,
"zero_fraction": 0.44774
},
{
"name": "transformer.h.3.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.4952,
"zero_fraction": 0.45133
},
{
"name": "transformer.h.3.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.3.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.3.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.5111,
"zero_fraction": 0.45072
},
{
"name": "transformer.h.3.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51403,
"zero_fraction": 0.44681
},
{
"name": "transformer.h.3.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.51179,
"zero_fraction": 0.44924
},
{
"name": "transformer.h.4.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.4.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.58958,
"zero_fraction": 0.44988
},
{
"name": "transformer.h.4.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52625,
"zero_fraction": 0.44262
},
{
"name": "transformer.h.4.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.4.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.55556,
"zero_fraction": 0.44876
},
{
"name": "transformer.h.4.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50009,
"zero_fraction": 0.45511
},
{
"name": "transformer.h.4.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.4.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.4.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54572,
"zero_fraction": 0.45089
},
{
"name": "transformer.h.4.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50837,
"zero_fraction": 0.44698
},
{
"name": "transformer.h.4.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50481,
"zero_fraction": 0.44935
},
{
"name": "transformer.h.5.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.5.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.55943,
"zero_fraction": 0.44953
},
{
"name": "transformer.h.5.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52182,
"zero_fraction": 0.43993
},
{
"name": "transformer.h.5.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.5.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53754,
"zero_fraction": 0.44802
},
{
"name": "transformer.h.5.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49815,
"zero_fraction": 0.45029
},
{
"name": "transformer.h.5.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.5.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.5.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54146,
"zero_fraction": 0.44955
},
{
"name": "transformer.h.5.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50303,
"zero_fraction": 0.44787
},
{
"name": "transformer.h.5.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50337,
"zero_fraction": 0.4479
},
{
"name": "transformer.h.6.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.6.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.55472,
"zero_fraction": 0.44853
},
{
"name": "transformer.h.6.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49489,
"zero_fraction": 0.44197
},
{
"name": "transformer.h.6.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.6.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.54333,
"zero_fraction": 0.44813
},
{
"name": "transformer.h.6.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49503,
"zero_fraction": 0.45183
},
{
"name": "transformer.h.6.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.6.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.6.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54083,
"zero_fraction": 0.44993
},
{
"name": "transformer.h.6.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50204,
"zero_fraction": 0.44694
},
{
"name": "transformer.h.6.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50145,
"zero_fraction": 0.44842
},
{
"name": "transformer.h.7.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.7.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.52404,
"zero_fraction": 0.44938
},
{
"name": "transformer.h.7.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50169,
"zero_fraction": 0.44117
},
{
"name": "transformer.h.7.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.7.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.5452,
"zero_fraction": 0.44963
},
{
"name": "transformer.h.7.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49302,
"zero_fraction": 0.45467
},
{
"name": "transformer.h.7.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.7.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.7.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54576,
"zero_fraction": 0.44958
},
{
"name": "transformer.h.7.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49993,
"zero_fraction": 0.44659
},
{
"name": "transformer.h.7.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50031,
"zero_fraction": 0.44807
},
{
"name": "transformer.h.8.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.8.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.54765,
"zero_fraction": 0.454
},
{
"name": "transformer.h.8.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50266,
"zero_fraction": 0.44229
},
{
"name": "transformer.h.8.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.8.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.54446,
"zero_fraction": 0.45176
},
{
"name": "transformer.h.8.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49281,
"zero_fraction": 0.45563
},
{
"name": "transformer.h.8.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.8.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.8.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.56901,
"zero_fraction": 0.45218
},
{
"name": "transformer.h.8.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49997,
"zero_fraction": 0.44706
},
{
"name": "transformer.h.8.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50011,
"zero_fraction": 0.44825
},
{
"name": "transformer.h.9.attn.k_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.9.attn.k_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.57334,
"zero_fraction": 0.44753
},
{
"name": "transformer.h.9.attn.o_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.47812,
"zero_fraction": 0.44168
},
{
"name": "transformer.h.9.attn.q_norm.weight",
"shape": [
80
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.9.attn.q_proj.weight",
"shape": [
640,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.53186,
"zero_fraction": 0.44458
},
{
"name": "transformer.h.9.attn.v_proj.weight",
"shape": [
320,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.49707,
"zero_fraction": 0.45616
},
{
"name": "transformer.h.9.ln_1.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.9.ln_2.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.h.9.mlp.w_down.weight",
"shape": [
640,
1920
],
"quantized": true,
"reason": "",
"relative_error": 0.54528,
"zero_fraction": 0.45042
},
{
"name": "transformer.h.9.mlp.w_gate.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.50142,
"zero_fraction": 0.44719
},
{
"name": "transformer.h.9.mlp.w_up.weight",
"shape": [
1920,
640
],
"quantized": true,
"reason": "",
"relative_error": 0.4995,
"zero_fraction": 0.44919
},
{
"name": "transformer.ln_f.weight",
"shape": [
640
],
"quantized": false,
"reason": "not a matrix",
"relative_error": 0.0,
"zero_fraction": 0.0
},
{
"name": "transformer.wte.weight",
"shape": [
32768,
640
],
"quantized": true,
"reason": "8-bit",
"relative_error": 0.0065,
"zero_fraction": 0.00893
}
]
}