MOSS-Image-SFT-NVFP4-Core / benchmark /weight_quantization_metrics.json
WiktorMatuszek's picture
MOSS-Image-SFT MOSS-Image-SFT-NVFP4-Core experimental release
efa7c77
Raw History Blame Contribute Delete
89.3 kB
{
"all_linear_weights": {
"MAE": 0.0027748894747125826,
"MSE": 1.6767324196748098e-05,
"NMSE": 0.007748817962681275,
"RMSE": 0.004094792326449303,
"SQNR_dB": 21.10764541550128,
"cosine_similarity": 0.9961264739410771,
"layers": 269,
"max_absolute_error": 0.201171875,
"reconstructed_sum_sq": 11001423.685776627,
"scope": "parameter_weighted_aggregate",
"source_sum_sq": 10996549.497795047,
"squared_error_sum": 85210.26027602801,
"weight_elements": 5081923584
},
"elapsed_seconds": 1044.72,
"linear_layers_compared": 269,
"measured_utc": "2026-10-21T12:53:13.222640+00:00",
"measured_weights": {
"MAE": 0.003276925399413872,
"MSE": 1.980088614023871e-05,
"NMSE": 0.008972728083962562,
"RMSE": 0.0044498186637478504,
"SQNR_dB": 20.470754932234925,
"cosine_similarity": 0.9955148197561109,
"layers": 228,
"max_absolute_error": 0.201171875,
"reconstructed_sum_sq": 9501457.554675689,
"scope": "parameter_weighted_aggregate",
"source_sum_sq": 9496583.366694113,
"squared_error_sum": 85210.26027602801,
"weight_elements": 4303355904
},
"method": "nvfp4",
"method_notes": "True decoded saved TorchAO tensors vs original BF16 source weights; ratios aggregate parameter-weighted sums. No image/activation loss estimated.",
"original_linear_layers_expected": 269,
"original_revision": "fc5de0d56fc910cbca6f6625a70f5f582b26deb1",
"original_source": "OpenMOSS-Team/MOSS-Image-SFT",
"per_layer": [
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.0.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.0.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.0.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.0.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.0.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.0.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.1.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.1.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.1.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.1.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.1.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.1.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0032988988753851066,
"MSE": 1.935879046891327e-05,
"NMSE": 0.008955136329686095,
"SQNR_dB": 20.479277982259248,
"max_absolute_error": 0.07421875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.10.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003411518869634827,
"MSE": 2.0910621774360973e-05,
"NMSE": 0.008986604576283417,
"SQNR_dB": 20.464043674799687,
"max_absolute_error": 0.0712890625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.10.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0027018909104999426,
"MSE": 1.3092056201418037e-05,
"NMSE": 0.008996525442651504,
"SQNR_dB": 20.459251874982943,
"max_absolute_error": 0.029296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.10.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002672588826904042,
"MSE": 1.3298164306643187e-05,
"NMSE": 0.009048508066198708,
"SQNR_dB": 20.43423022124184,
"max_absolute_error": 0.0458984375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.10.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026562780283499892,
"MSE": 1.2720501639464145e-05,
"NMSE": 0.008999660309040678,
"SQNR_dB": 20.45773882637711,
"max_absolute_error": 0.0302734375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.10.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002682691214309804,
"MSE": 1.297555137224514e-05,
"NMSE": 0.008958809483324566,
"SQNR_dB": 20.477496989511614,
"max_absolute_error": 0.04052734375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.10.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0032971796808349504,
"MSE": 1.933732765056731e-05,
"NMSE": 0.00896023184265181,
"SQNR_dB": 20.476807529850056,
"max_absolute_error": 0.05322265625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.11.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003423167387213176,
"MSE": 2.1073621871890174e-05,
"NMSE": 0.009031566158051514,
"SQNR_dB": 20.442369324304693,
"max_absolute_error": 0.0712890625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.11.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0027031015516676285,
"MSE": 1.3113428236541955e-05,
"NMSE": 0.009004522282484611,
"SQNR_dB": 20.455393228871138,
"max_absolute_error": 0.0390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.11.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002777870747871364,
"MSE": 1.4363852426621739e-05,
"NMSE": 0.009056675014664632,
"SQNR_dB": 20.43031215990276,
"max_absolute_error": 0.037109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.11.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002693411246440324,
"MSE": 1.3039613596358732e-05,
"NMSE": 0.009019662441860817,
"SQNR_dB": 20.448097154935745,
"max_absolute_error": 0.0390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.11.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0027822910902357777,
"MSE": 1.4066131065828645e-05,
"NMSE": 0.008958679203539292,
"SQNR_dB": 20.47756014545142,
"max_absolute_error": 0.0732421875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.11.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0032999214989740525,
"MSE": 1.9410027003149536e-05,
"NMSE": 0.008976082709792622,
"SQNR_dB": 20.469131542785284,
"max_absolute_error": 0.059814453125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.12.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003400065288685685,
"MSE": 2.0828883659494408e-05,
"NMSE": 0.009017008730353857,
"SQNR_dB": 20.449375098367042,
"max_absolute_error": 0.0537109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.12.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002739804166556322,
"MSE": 1.3425541904456114e-05,
"NMSE": 0.00902013183892794,
"SQNR_dB": 20.44787114730734,
"max_absolute_error": 0.0390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.12.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026804683958591086,
"MSE": 1.3305652292987015e-05,
"NMSE": 0.009040460547919692,
"SQNR_dB": 20.43809444709875,
"max_absolute_error": 0.0390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.12.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002675327037653441,
"MSE": 1.2873458577595128e-05,
"NMSE": 0.0090245786318313,
"SQNR_dB": 20.44573066668027,
"max_absolute_error": 0.03369140625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.12.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0027215466959022716,
"MSE": 1.3390791825230881e-05,
"NMSE": 0.008966122142718014,
"SQNR_dB": 20.473953491700794,
"max_absolute_error": 0.0380859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.12.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003273073964822553,
"MSE": 1.911599150217392e-05,
"NMSE": 0.008964686518669392,
"SQNR_dB": 20.474648924478544,
"max_absolute_error": 0.09130859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.13.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003405436007725984,
"MSE": 2.0864616222633032e-05,
"NMSE": 0.009028827706993354,
"SQNR_dB": 20.4436863434146,
"max_absolute_error": 0.07421875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.13.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0027122736170775677,
"MSE": 1.3189044504385161e-05,
"NMSE": 0.009029025546882735,
"SQNR_dB": 20.443591181746097,
"max_absolute_error": 0.0341796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.13.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0027269046016569457,
"MSE": 1.3860168021579532e-05,
"NMSE": 0.009049021257619362,
"SQNR_dB": 20.43398391557412,
"max_absolute_error": 0.0380859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.13.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026556283324664634,
"MSE": 1.2729319855151931e-05,
"NMSE": 0.009021226142785931,
"SQNR_dB": 20.447344302124048,
"max_absolute_error": 0.03125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.13.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.00274107860126932,
"MSE": 1.3676485419328468e-05,
"NMSE": 0.00897076778565807,
"SQNR_dB": 20.47170385182796,
"max_absolute_error": 0.0361328125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.13.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003272581703900999,
"MSE": 1.9130126301017647e-05,
"NMSE": 0.008944016171310968,
"SQNR_dB": 20.484674242267968,
"max_absolute_error": 0.08447265625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.14.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034082447341519232,
"MSE": 2.094216814464314e-05,
"NMSE": 0.009010217708433136,
"SQNR_dB": 20.45264715299639,
"max_absolute_error": 0.060546875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.14.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0026517875712909,
"MSE": 1.2640052441449323e-05,
"NMSE": 0.009010538178252467,
"SQNR_dB": 20.4524926885832,
"max_absolute_error": 0.028076171875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.14.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0028024879004787773,
"MSE": 1.4725848044289196e-05,
"NMSE": 0.009045339071560788,
"SQNR_dB": 20.435751486330492,
"max_absolute_error": 0.048828125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.14.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026187286992022723,
"MSE": 1.2400471724076914e-05,
"NMSE": 0.009017565477578307,
"SQNR_dB": 20.449106955360527,
"max_absolute_error": 0.03125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.14.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0028412058547300995,
"MSE": 1.4709014984964608e-05,
"NMSE": 0.009015342973120469,
"SQNR_dB": 20.450177466453702,
"max_absolute_error": 0.0380859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.14.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.00332189498055655,
"MSE": 1.9758627402822207e-05,
"NMSE": 0.008924333786519235,
"SQNR_dB": 20.49424194670978,
"max_absolute_error": 0.0478515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.15.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0033934620301608073,
"MSE": 2.077650907210628e-05,
"NMSE": 0.009021557324066079,
"SQNR_dB": 20.44718486973629,
"max_absolute_error": 0.09375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.15.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0026923068817163256,
"MSE": 1.295092894666217e-05,
"NMSE": 0.00902587042925243,
"SQNR_dB": 20.44510905279045,
"max_absolute_error": 0.0361328125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.15.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002750168611740816,
"MSE": 1.4102008663662923e-05,
"NMSE": 0.009043053353675262,
"SQNR_dB": 20.43684906830109,
"max_absolute_error": 0.04296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.15.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026421368165608482,
"MSE": 1.2526769728796028e-05,
"NMSE": 0.009030801549029794,
"SQNR_dB": 20.442737011957313,
"max_absolute_error": 0.03125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.15.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0027686857953824243,
"MSE": 1.3920998094582366e-05,
"NMSE": 0.008954257611067752,
"SQNR_dB": 20.479704152623988,
"max_absolute_error": 0.04150390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.15.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0033308681751435137,
"MSE": 1.988591201518775e-05,
"NMSE": 0.008925230274345677,
"SQNR_dB": 20.493805701079225,
"max_absolute_error": 0.08056640625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.16.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.00338321267829759,
"MSE": 2.066866687368697e-05,
"NMSE": 0.008976774612159568,
"SQNR_dB": 20.468796788963385,
"max_absolute_error": 0.0693359375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.16.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.00265226660314004,
"MSE": 1.2582835259318463e-05,
"NMSE": 0.009007885491206373,
"SQNR_dB": 20.453771432357676,
"max_absolute_error": 0.0302734375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.16.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0027676738650758395,
"MSE": 1.4319250377366995e-05,
"NMSE": 0.009091809053234206,
"SQNR_dB": 20.413496939322776,
"max_absolute_error": 0.04248046875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.16.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002625022927865107,
"MSE": 1.2370503125092802e-05,
"NMSE": 0.00907148134712562,
"SQNR_dB": 20.42321788094071,
"max_absolute_error": 0.027099609375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.16.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0027618197137846707,
"MSE": 1.3856569575889914e-05,
"NMSE": 0.008938184165233749,
"SQNR_dB": 20.4875070123968,
"max_absolute_error": 0.046875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.16.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0032985324818268747,
"MSE": 1.9528530989620325e-05,
"NMSE": 0.008907703612336445,
"SQNR_dB": 20.50234241757916,
"max_absolute_error": 0.07958984375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.17.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0033867868600563366,
"MSE": 2.0679641165488134e-05,
"NMSE": 0.008963662877446064,
"SQNR_dB": 20.475144856061984,
"max_absolute_error": 0.0517578125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.17.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002652782754429878,
"MSE": 1.261704641964737e-05,
"NMSE": 0.009001740161663089,
"SQNR_dB": 20.456735272767407,
"max_absolute_error": 0.03515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.17.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002810586640019103,
"MSE": 1.4884758457739351e-05,
"NMSE": 0.00905383253829179,
"SQNR_dB": 20.43167542574188,
"max_absolute_error": 0.05078125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.17.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002597699198286316,
"MSE": 1.214196744422614e-05,
"NMSE": 0.008996940998133576,
"SQNR_dB": 20.459051276111776,
"max_absolute_error": 0.03173828125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.17.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002840729090391255,
"MSE": 1.4622075050420682e-05,
"NMSE": 0.008916583968279329,
"SQNR_dB": 20.498014962823966,
"max_absolute_error": 0.0537109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.17.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003273684828647181,
"MSE": 1.930297398801573e-05,
"NMSE": 0.008933776044528137,
"SQNR_dB": 20.48964938795535,
"max_absolute_error": 0.0654296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.18.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034217638055102154,
"MSE": 2.107501857502682e-05,
"NMSE": 0.009000948787398174,
"SQNR_dB": 20.457117092925014,
"max_absolute_error": 0.060546875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.18.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0026955538289565872,
"MSE": 1.2978056875959265e-05,
"NMSE": 0.009048758854852184,
"SQNR_dB": 20.434109853753412,
"max_absolute_error": 0.029052734375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.18.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0029092613723884896,
"MSE": 1.5834897243515306e-05,
"NMSE": 0.009045898744011373,
"SQNR_dB": 20.435482778727927,
"max_absolute_error": 0.0556640625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.18.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026641471851573642,
"MSE": 1.2691896722547058e-05,
"NMSE": 0.00900739736892087,
"SQNR_dB": 20.45400677566789,
"max_absolute_error": 0.02880859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.18.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002980924186615361,
"MSE": 1.600727592885488e-05,
"NMSE": 0.008939638165279354,
"SQNR_dB": 20.486800590567327,
"max_absolute_error": 0.041015625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.18.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002689253992418992,
"MSE": 1.302095096944188e-05,
"NMSE": 0.009008986791164965,
"SQNR_dB": 20.453240498366426,
"max_absolute_error": 0.0341796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.19.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0029522600566302434,
"MSE": 1.6330531456071775e-05,
"NMSE": 0.009044653650943282,
"SQNR_dB": 20.436080590268425,
"max_absolute_error": 0.0458984375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.19.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002653130072650929,
"MSE": 1.2694079015312069e-05,
"NMSE": 0.008994813551262033,
"SQNR_dB": 20.460078344859983,
"max_absolute_error": 0.03515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.19.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0030625537754158908,
"MSE": 1.6869408230765253e-05,
"NMSE": 0.0089483267066781,
"SQNR_dB": 20.48258168035701,
"max_absolute_error": 0.0419921875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.19.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003286816788486213,
"MSE": 1.945328524503975e-05,
"NMSE": 0.008967466570381997,
"SQNR_dB": 20.473302336456605,
"max_absolute_error": 0.0908203125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.2.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034593767187891394,
"MSE": 2.135157238784443e-05,
"NMSE": 0.008996326989466934,
"SQNR_dB": 20.45934767649446,
"max_absolute_error": 0.0703125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.2.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0025880968945509006,
"MSE": 1.201968192360417e-05,
"NMSE": 0.008993780949705626,
"SQNR_dB": 20.460576942077708,
"max_absolute_error": 0.02880859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.2.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002639492186337519,
"MSE": 1.2806213059650674e-05,
"NMSE": 0.009045307680047325,
"SQNR_dB": 20.43576655838542,
"max_absolute_error": 0.0341796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.2.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.00257682754887395,
"MSE": 1.194544665879564e-05,
"NMSE": 0.009004099870677842,
"SQNR_dB": 20.455596965853616,
"max_absolute_error": 0.028076171875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.2.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.00241062248189187,
"MSE": 1.0555518660154027e-05,
"NMSE": 0.008880335957096235,
"SQNR_dB": 20.515706038663176,
"max_absolute_error": 0.03271484375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.2.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0033067507186814965,
"MSE": 1.9642245885164452e-05,
"NMSE": 0.008975354868813538,
"SQNR_dB": 20.469483712153014,
"max_absolute_error": 0.07177734375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.3.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003493249396597967,
"MSE": 2.1793632019928566e-05,
"NMSE": 0.009003467552361808,
"SQNR_dB": 20.455901962468786,
"max_absolute_error": 0.0859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.3.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002633858304936232,
"MSE": 1.2436331585425417e-05,
"NMSE": 0.009015965893818718,
"SQNR_dB": 20.44987739836212,
"max_absolute_error": 0.0322265625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.3.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002753831045629392,
"MSE": 1.39961427485718e-05,
"NMSE": 0.009054762923119197,
"SQNR_dB": 20.431229161427446,
"max_absolute_error": 0.05078125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.3.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026043510339284325,
"MSE": 1.2198936884943998e-05,
"NMSE": 0.00900556672400813,
"SQNR_dB": 20.454889516545585,
"max_absolute_error": 0.029296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.3.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026012185854969757,
"MSE": 1.2350232291749836e-05,
"NMSE": 0.008907218218802105,
"SQNR_dB": 20.502579077313694,
"max_absolute_error": 0.068359375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.3.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.00331184490313215,
"MSE": 1.9609267233168572e-05,
"NMSE": 0.008961360929231525,
"SQNR_dB": 20.47626030608702,
"max_absolute_error": 0.056640625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.4.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003493634484935656,
"MSE": 2.1775019562978517e-05,
"NMSE": 0.009028058903437036,
"SQNR_dB": 20.44405616037096,
"max_absolute_error": 0.05078125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.4.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002646604266179959,
"MSE": 1.2621542455091984e-05,
"NMSE": 0.009005316170036103,
"SQNR_dB": 20.455010348165132,
"max_absolute_error": 0.03271484375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.4.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026646712533252817,
"MSE": 1.3045695583796375e-05,
"NMSE": 0.00905004467796558,
"SQNR_dB": 20.433492767786618,
"max_absolute_error": 0.0419921875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.4.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002616233580940233,
"MSE": 1.2401774023423457e-05,
"NMSE": 0.008975865224348753,
"SQNR_dB": 20.469236771174447,
"max_absolute_error": 0.03125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.4.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0025382369353142125,
"MSE": 1.1529867111533029e-05,
"NMSE": 0.008920641856133331,
"SQNR_dB": 20.496038962350656,
"max_absolute_error": 0.0478515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.4.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003307444441389661,
"MSE": 1.955622119242165e-05,
"NMSE": 0.008965003192069873,
"SQNR_dB": 20.474495514673215,
"max_absolute_error": 0.07568359375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.5.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003487523039173041,
"MSE": 2.1735617493892896e-05,
"NMSE": 0.00900906143293643,
"SQNR_dB": 20.45320451610124,
"max_absolute_error": 0.103515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.5.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002679457663474571,
"MSE": 1.2866114951358325e-05,
"NMSE": 0.009008400655620879,
"SQNR_dB": 20.453523064785433,
"max_absolute_error": 0.029296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.5.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0027273670547606397,
"MSE": 1.3603851214767321e-05,
"NMSE": 0.009022999649484775,
"SQNR_dB": 20.446490595033175,
"max_absolute_error": 0.0400390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.5.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026237155825490385,
"MSE": 1.2376661318828058e-05,
"NMSE": 0.008983410071094385,
"SQNR_dB": 20.465587753664224,
"max_absolute_error": 0.0302734375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.5.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026246277597585183,
"MSE": 1.2309963743075106e-05,
"NMSE": 0.0089441257571814,
"SQNR_dB": 20.484621030996486,
"max_absolute_error": 0.0380859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.5.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0033136554667292325,
"MSE": 1.9570318841016612e-05,
"NMSE": 0.009010362993023274,
"SQNR_dB": 20.452577126067904,
"max_absolute_error": 0.062744140625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.6.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034616714456439258,
"MSE": 2.1454168072301056e-05,
"NMSE": 0.008992071690942691,
"SQNR_dB": 20.461402392687088,
"max_absolute_error": 0.08203125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.6.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0026715010907464743,
"MSE": 1.2825639088315656e-05,
"NMSE": 0.009005244048154183,
"SQNR_dB": 20.45504513013189,
"max_absolute_error": 0.0361328125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.6.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.00270662344669827,
"MSE": 1.3538858424789353e-05,
"NMSE": 0.009053520490976205,
"SQNR_dB": 20.431825111261396,
"max_absolute_error": 0.046875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.6.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026261057336014145,
"MSE": 1.2420078262064477e-05,
"NMSE": 0.008991267305588463,
"SQNR_dB": 20.46179090799113,
"max_absolute_error": 0.0283203125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.6.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026366849799877887,
"MSE": 1.2551637397155289e-05,
"NMSE": 0.008918790149726784,
"SQNR_dB": 20.49694054472962,
"max_absolute_error": 0.0400390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.6.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003325811038770214,
"MSE": 1.9731385883589962e-05,
"NMSE": 0.00896993858198076,
"SQNR_dB": 20.47210530601188,
"max_absolute_error": 0.06494140625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.7.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034737185954052314,
"MSE": 2.1621093070495604e-05,
"NMSE": 0.009034188718220273,
"SQNR_dB": 20.44110841555712,
"max_absolute_error": 0.0908203125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.7.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0026452656738269404,
"MSE": 1.2553023721627193e-05,
"NMSE": 0.009016419268773657,
"SQNR_dB": 20.449659015447793,
"max_absolute_error": 0.029052734375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.7.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002754316078114124,
"MSE": 1.3952477802204109e-05,
"NMSE": 0.009039102745159643,
"SQNR_dB": 20.438746770650763,
"max_absolute_error": 0.05078125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.7.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026019584025298668,
"MSE": 1.2177613578083004e-05,
"NMSE": 0.009028046744320135,
"SQNR_dB": 20.444062009514155,
"max_absolute_error": 0.0341796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.7.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002696466199705483,
"MSE": 1.3068277862034441e-05,
"NMSE": 0.008952599865095659,
"SQNR_dB": 20.480508257901636,
"max_absolute_error": 0.0361328125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.7.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0033260200799639793,
"MSE": 1.9655339415100488e-05,
"NMSE": 0.008950180676233726,
"SQNR_dB": 20.481681975445,
"max_absolute_error": 0.08544921875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.8.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034387435622556133,
"MSE": 2.1205088723814263e-05,
"NMSE": 0.009000810095103234,
"SQNR_dB": 20.45718401227303,
"max_absolute_error": 0.126953125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.8.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002636613479651054,
"MSE": 1.2464963993155023e-05,
"NMSE": 0.008991806855377621,
"SQNR_dB": 20.461530303498133,
"max_absolute_error": 0.029296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.8.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0027243488988408534,
"MSE": 1.3709203815105218e-05,
"NMSE": 0.009046724544489872,
"SQNR_dB": 20.435086329207593,
"max_absolute_error": 0.0341796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.8.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0025145852570024685,
"MSE": 1.1441139479820114e-05,
"NMSE": 0.008996516776181209,
"SQNR_dB": 20.459256058600342,
"max_absolute_error": 0.029296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.8.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002721299860398627,
"MSE": 1.3363300464126135e-05,
"NMSE": 0.008941016606227023,
"SQNR_dB": 20.486130985000823,
"max_absolute_error": 0.037109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.8.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003320040366362018,
"MSE": 1.9559612244160853e-05,
"NMSE": 0.008956628399406646,
"SQNR_dB": 20.478554438088644,
"max_absolute_error": 0.0693359375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.9.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034254351639014962,
"MSE": 2.1050955865840473e-05,
"NMSE": 0.00901349935025063,
"SQNR_dB": 20.451065682377397,
"max_absolute_error": 0.07275390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.9.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0026919665851323524,
"MSE": 1.299581350507369e-05,
"NMSE": 0.008980611688769279,
"SQNR_dB": 20.466940815944067,
"max_absolute_error": 0.0341796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.9.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002614129607295156,
"MSE": 1.269382105095844e-05,
"NMSE": 0.00906946220730134,
"SQNR_dB": 20.42418464574527,
"max_absolute_error": 0.04296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.9.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026067337703628876,
"MSE": 1.2266913151552261e-05,
"NMSE": 0.0089938118868775,
"SQNR_dB": 20.460562003065963,
"max_absolute_error": 0.037109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.9.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026634054056220028,
"MSE": 1.2731469208838082e-05,
"NMSE": 0.008912194724290381,
"SQNR_dB": 20.500153330813728,
"max_absolute_error": 0.056640625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.9.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "shared_block_modulation.weight",
"weight_elements": 56623104
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.blocks.0.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.blocks.0.ffn_gate.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.blocks.0.ffn_up.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.blocks.0.proj.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.blocks.0.qkv.weight",
"weight_elements": 28311552
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.blocks.1.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.blocks.1.ffn_gate.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.blocks.1.ffn_up.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.blocks.1.proj.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.blocks.1.qkv.weight",
"weight_elements": 28311552
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "text_embedder.proj.weight",
"weight_elements": 12582912
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "time_embedding.0.weight",
"weight_elements": 786432
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "time_embedding.2.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "video_embedder.weight",
"weight_elements": 393216
},
{
"MAE": 0.0032296321883062457,
"MSE": 1.878892822630606e-05,
"NMSE": 0.008931320529165883,
"SQNR_dB": 20.490843242994835,
"max_absolute_error": 0.060546875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.19.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034355206256639903,
"MSE": 2.1257601455449735e-05,
"NMSE": 0.009000220957055454,
"SQNR_dB": 20.457468284215654,
"max_absolute_error": 0.0751953125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.19.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0032201885086067957,
"MSE": 1.8773042392963178e-05,
"NMSE": 0.008971371292592055,
"SQNR_dB": 20.471411690753772,
"max_absolute_error": 0.0537109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.20.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0033909375457948383,
"MSE": 2.0761400708988163e-05,
"NMSE": 0.008950206991477744,
"SQNR_dB": 20.481669206374878,
"max_absolute_error": 0.103515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.20.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0026785449849001724,
"MSE": 1.2838439168898807e-05,
"NMSE": 0.009047453457009564,
"SQNR_dB": 20.434736423635496,
"max_absolute_error": 0.030029296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.20.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0029449244202353706,
"MSE": 1.6323114774432105e-05,
"NMSE": 0.009036676096157043,
"SQNR_dB": 20.439912839661595,
"max_absolute_error": 0.048828125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.20.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026756552001411618,
"MSE": 1.2818224511460356e-05,
"NMSE": 0.00901757275946146,
"SQNR_dB": 20.449103448338146,
"max_absolute_error": 0.02880859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.20.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0030287466526869758,
"MSE": 1.6406702361514223e-05,
"NMSE": 0.008943846157733764,
"SQNR_dB": 20.48475679652411,
"max_absolute_error": 0.037109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.20.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003190950086517935,
"MSE": 1.851031893973908e-05,
"NMSE": 0.008960963360776385,
"SQNR_dB": 20.476452983982455,
"max_absolute_error": 0.0693359375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.21.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0033984523110561203,
"MSE": 2.083959322101803e-05,
"NMSE": 0.008962869123338265,
"SQNR_dB": 20.47552945139723,
"max_absolute_error": 0.0595703125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.21.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002712106389531954,
"MSE": 1.3234111221843783e-05,
"NMSE": 0.009005632577087497,
"SQNR_dB": 20.454857758939102,
"max_absolute_error": 0.0380859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.21.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003061595143980877,
"MSE": 1.7706083861412497e-05,
"NMSE": 0.009106708661700342,
"SQNR_dB": 20.40638556982934,
"max_absolute_error": 0.0458984375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.21.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026958005392093264,
"MSE": 1.3104075744048732e-05,
"NMSE": 0.008986547573794006,
"SQNR_dB": 20.464071222406695,
"max_absolute_error": 0.033203125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.21.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003152384133062574,
"MSE": 1.7771315338338784e-05,
"NMSE": 0.008974274076621248,
"SQNR_dB": 20.470006711360618,
"max_absolute_error": 0.03515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.21.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003216642740495651,
"MSE": 1.8815813970750026e-05,
"NMSE": 0.008958190469884502,
"SQNR_dB": 20.477797077832264,
"max_absolute_error": 0.078125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.22.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034271501591202403,
"MSE": 2.122790038520696e-05,
"NMSE": 0.008977848009150797,
"SQNR_dB": 20.468277512785242,
"max_absolute_error": 0.068359375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.22.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002663045176867909,
"MSE": 1.267431280086269e-05,
"NMSE": 0.009014352122898339,
"SQNR_dB": 20.450654813166448,
"max_absolute_error": 0.02880859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.22.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003077458646018978,
"MSE": 1.774152098885203e-05,
"NMSE": 0.009031538290713702,
"SQNR_dB": 20.442382724693395,
"max_absolute_error": 0.03955078125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.22.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002699910084111493,
"MSE": 1.3032942669025086e-05,
"NMSE": 0.009005825721845148,
"SQNR_dB": 20.454764616339464,
"max_absolute_error": 0.029296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.22.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003182029534224353,
"MSE": 1.8039556532454735e-05,
"NMSE": 0.008941317042890634,
"SQNR_dB": 20.485985055517865,
"max_absolute_error": 0.033203125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.22.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0032915418204595877,
"MSE": 1.964858627074616e-05,
"NMSE": 0.008922613847330799,
"SQNR_dB": 20.495079019935282,
"max_absolute_error": 0.0576171875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.23.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003439350541220441,
"MSE": 2.1515199651390445e-05,
"NMSE": 0.009009157375048498,
"SQNR_dB": 20.453158266102506,
"max_absolute_error": 0.07421875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.23.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002653993972405967,
"MSE": 1.2647027048189624e-05,
"NMSE": 0.008978934298604711,
"SQNR_dB": 20.467752062837615,
"max_absolute_error": 0.03076171875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.23.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0031696473388776757,
"MSE": 1.8829473376031982e-05,
"NMSE": 0.009048042620313706,
"SQNR_dB": 20.43445362357196,
"max_absolute_error": 0.0400390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.23.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0026886894762102598,
"MSE": 1.3006040381698006e-05,
"NMSE": 0.00898383981941103,
"SQNR_dB": 20.465380000865437,
"max_absolute_error": 0.029296875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.23.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0033313196431259545,
"MSE": 1.9801135246383597e-05,
"NMSE": 0.008958953661561634,
"SQNR_dB": 20.477427097067714,
"max_absolute_error": 0.04541015625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.23.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0033789736620785786,
"MSE": 2.0752706021113192e-05,
"NMSE": 0.008867008044993407,
"SQNR_dB": 20.522228975018052,
"max_absolute_error": 0.0576171875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.24.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0033869689639952363,
"MSE": 2.1093577527979168e-05,
"NMSE": 0.008966509911301143,
"SQNR_dB": 20.473765671245406,
"max_absolute_error": 0.08984375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.24.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002681685052999762,
"MSE": 1.293388678339159e-05,
"NMSE": 0.008987152349280731,
"SQNR_dB": 20.463778961316436,
"max_absolute_error": 0.0361328125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.24.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0032992276349187504,
"MSE": 2.041406837574193e-05,
"NMSE": 0.0090775436425075,
"SQNR_dB": 20.420316544332884,
"max_absolute_error": 0.044921875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.24.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002729252628499476,
"MSE": 1.3480852464777571e-05,
"NMSE": 0.00897512985270449,
"SQNR_dB": 20.469592593061826,
"max_absolute_error": 0.03271484375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.24.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003301354952729664,
"MSE": 1.9436624450668608e-05,
"NMSE": 0.008967637862048882,
"SQNR_dB": 20.473219380680124,
"max_absolute_error": 0.0419921875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.24.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003382402370777282,
"MSE": 2.086884511685739e-05,
"NMSE": 0.008845547204824162,
"SQNR_dB": 20.532752955202568,
"max_absolute_error": 0.0830078125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.25.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.00329939451718987,
"MSE": 2.014387024451953e-05,
"NMSE": 0.008956849962432754,
"SQNR_dB": 20.478447006580716,
"max_absolute_error": 0.091796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.25.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002702014776840198,
"MSE": 1.3176592826157173e-05,
"NMSE": 0.00899550007045539,
"SQNR_dB": 20.459746887050176,
"max_absolute_error": 0.0302734375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.25.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0033157053354364056,
"MSE": 2.0741211336279633e-05,
"NMSE": 0.00903312971054089,
"SQNR_dB": 20.44161753504133,
"max_absolute_error": 0.06640625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.25.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002769698412118049,
"MSE": 1.386941054351448e-05,
"NMSE": 0.009013933376482089,
"SQNR_dB": 20.45085656197715,
"max_absolute_error": 0.03125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.25.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003509667642545979,
"MSE": 2.2088977017554008e-05,
"NMSE": 0.00892768959342862,
"SQNR_dB": 20.49260918115797,
"max_absolute_error": 0.0517578125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.25.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003338864540850233,
"MSE": 2.0375080749728797e-05,
"NMSE": 0.008905841067052184,
"SQNR_dB": 20.50325059521225,
"max_absolute_error": 0.050537109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.26.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0032764938241447474,
"MSE": 1.9753066744146965e-05,
"NMSE": 0.008941386868651098,
"SQNR_dB": 20.485951140129437,
"max_absolute_error": 0.091796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.26.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0028670020975600555,
"MSE": 1.4808945974286308e-05,
"NMSE": 0.008988680391738717,
"SQNR_dB": 20.46304061398224,
"max_absolute_error": 0.033203125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.26.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003406160527725149,
"MSE": 2.183569339618736e-05,
"NMSE": 0.00903017426128205,
"SQNR_dB": 20.443038687359994,
"max_absolute_error": 0.041015625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.26.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0028982740893388073,
"MSE": 1.511579322035562e-05,
"NMSE": 0.009006723439374332,
"SQNR_dB": 20.454331725063426,
"max_absolute_error": 0.03173828125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.26.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0034873822076053546,
"MSE": 2.17413342375025e-05,
"NMSE": 0.00896884692526377,
"SQNR_dB": 20.47263388191655,
"max_absolute_error": 0.0380859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.26.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0032874291643880706,
"MSE": 1.9769928938945805e-05,
"NMSE": 0.008962921890308494,
"SQNR_dB": 20.475503883316044,
"max_absolute_error": 0.06005859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.27.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003295692875126864,
"MSE": 1.985036134474442e-05,
"NMSE": 0.008965374976784278,
"SQNR_dB": 20.474315413563886,
"max_absolute_error": 0.103515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.27.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0028909331237540223,
"MSE": 1.5040711128009795e-05,
"NMSE": 0.009001038981751903,
"SQNR_dB": 20.457073574497354,
"max_absolute_error": 0.0341796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.27.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0033973760248780147,
"MSE": 2.1707347050630537e-05,
"NMSE": 0.009070795037029941,
"SQNR_dB": 20.4235464622954,
"max_absolute_error": 0.041015625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.27.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0029393553556364724,
"MSE": 1.559248895055844e-05,
"NMSE": 0.009007347954921387,
"SQNR_dB": 20.454030600847673,
"max_absolute_error": 0.03125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.27.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003622642551311123,
"MSE": 2.3599458697789508e-05,
"NMSE": 0.00894065055677181,
"SQNR_dB": 20.48630879085872,
"max_absolute_error": 0.056640625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.27.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0032997633661421173,
"MSE": 1.9894862383521076e-05,
"NMSE": 0.008940402890474108,
"SQNR_dB": 20.48642909708874,
"max_absolute_error": 0.08203125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.28.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0032772102286423715,
"MSE": 1.9713872681640326e-05,
"NMSE": 0.008952283024800499,
"SQNR_dB": 20.480661961215638,
"max_absolute_error": 0.1015625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.28.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0028916020688305732,
"MSE": 1.5113415621412599e-05,
"NMSE": 0.008990370467755651,
"SQNR_dB": 20.462224118499744,
"max_absolute_error": 0.03466796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.28.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003361315682485092,
"MSE": 2.1184504379361013e-05,
"NMSE": 0.00904310667310792,
"SQNR_dB": 20.436823461610288,
"max_absolute_error": 0.044921875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.28.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002903862377273439,
"MSE": 1.5310471607260345e-05,
"NMSE": 0.008972103908263373,
"SQNR_dB": 20.471057053828623,
"max_absolute_error": 0.02978515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.28.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0035655266841186175,
"MSE": 2.280682355227596e-05,
"NMSE": 0.008988148899854073,
"SQNR_dB": 20.463297415656214,
"max_absolute_error": 0.0390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.28.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003375321158616041,
"MSE": 2.075929193967565e-05,
"NMSE": 0.008921405455161452,
"SQNR_dB": 20.495667226037313,
"max_absolute_error": 0.0478515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.29.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0032924737484414297,
"MSE": 1.9965448438414884e-05,
"NMSE": 0.008961207413700276,
"SQNR_dB": 20.476334704964273,
"max_absolute_error": 0.16796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.29.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002915812971190217,
"MSE": 1.5376022035262266e-05,
"NMSE": 0.009014042104713611,
"SQNR_dB": 20.450804176648553,
"max_absolute_error": 0.0322265625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.29.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003444230334482576,
"MSE": 2.222855977558027e-05,
"NMSE": 0.009034264419490895,
"SQNR_dB": 20.441072024346592,
"max_absolute_error": 0.0400390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.29.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0029611362794738296,
"MSE": 1.5963502508498513e-05,
"NMSE": 0.008989910204906117,
"SQNR_dB": 20.462446461653848,
"max_absolute_error": 0.03271484375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.29.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003643647782630711,
"MSE": 2.379905414086111e-05,
"NMSE": 0.008982535071999073,
"SQNR_dB": 20.466010784321433,
"max_absolute_error": 0.0380859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.29.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0034880157621842976,
"MSE": 2.21427627792807e-05,
"NMSE": 0.00894450798035875,
"SQNR_dB": 20.484435441175325,
"max_absolute_error": 0.05078125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.30.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003340264588262937,
"MSE": 2.0615320373825727e-05,
"NMSE": 0.008955225771551098,
"SQNR_dB": 20.479234606130724,
"max_absolute_error": 0.13671875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.30.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0028624576412476254,
"MSE": 1.4841847092878529e-05,
"NMSE": 0.0089667615712508,
"SQNR_dB": 20.47364378101744,
"max_absolute_error": 0.03173828125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.30.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0035186456646366638,
"MSE": 2.3049592254359107e-05,
"NMSE": 0.009079200894798077,
"SQNR_dB": 20.41952374193878,
"max_absolute_error": 0.04248046875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.30.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002963735316747639,
"MSE": 1.5929034958043665e-05,
"NMSE": 0.009024364659014357,
"SQNR_dB": 20.44583363915038,
"max_absolute_error": 0.03466796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.30.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0035883785911866705,
"MSE": 2.308046392660323e-05,
"NMSE": 0.008971660568869668,
"SQNR_dB": 20.471271657461635,
"max_absolute_error": 0.0390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.30.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0035596940979458563,
"MSE": 2.300359821053585e-05,
"NMSE": 0.008892814201384389,
"SQNR_dB": 20.509607813553636,
"max_absolute_error": 0.078125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.31.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.00340224472233866,
"MSE": 2.1426887431069908e-05,
"NMSE": 0.008927961977437499,
"SQNR_dB": 20.492476679835846,
"max_absolute_error": 0.1005859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.31.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0028067383613988536,
"MSE": 1.4318355824182843e-05,
"NMSE": 0.008987595685073249,
"SQNR_dB": 20.463564729341677,
"max_absolute_error": 0.03515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.31.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0036891545071781866,
"MSE": 2.5397594999271635e-05,
"NMSE": 0.009060797241542164,
"SQNR_dB": 20.428335879335975,
"max_absolute_error": 0.04150390625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.31.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.00297336865913938,
"MSE": 1.606871691374897e-05,
"NMSE": 0.00896443586430433,
"SQNR_dB": 20.474770355751414,
"max_absolute_error": 0.046875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.31.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0036211164032625143,
"MSE": 2.345061028122898e-05,
"NMSE": 0.00894867599930931,
"SQNR_dB": 20.4824121593877,
"max_absolute_error": 0.037109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.31.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003653659779082524,
"MSE": 2.4149397565142894e-05,
"NMSE": 0.008896598702414412,
"SQNR_dB": 20.507759986480558,
"max_absolute_error": 0.076171875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.32.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034623859914590994,
"MSE": 2.232207029842234e-05,
"NMSE": 0.008957194899546776,
"SQNR_dB": 20.47827975871861,
"max_absolute_error": 0.12890625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.32.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0028471477606864873,
"MSE": 1.482887399718194e-05,
"NMSE": 0.009002711730624115,
"SQNR_dB": 20.456266558649926,
"max_absolute_error": 0.044921875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.32.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0036126442400342624,
"MSE": 2.45284713571752e-05,
"NMSE": 0.009017733723454782,
"SQNR_dB": 20.449025927311094,
"max_absolute_error": 0.0703125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.32.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002998427413717119,
"MSE": 1.63936329037207e-05,
"NMSE": 0.00900231432338138,
"SQNR_dB": 20.45645827375447,
"max_absolute_error": 0.0361328125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.32.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0036286430401956562,
"MSE": 2.369823481231816e-05,
"NMSE": 0.00893609319210017,
"SQNR_dB": 20.488523107290835,
"max_absolute_error": 0.072265625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.32.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.00377254661764135,
"MSE": 2.5673741200061356e-05,
"NMSE": 0.008895842848315008,
"SQNR_dB": 20.508128978295147,
"max_absolute_error": 0.0751953125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.33.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034756185254608916,
"MSE": 2.2640024049546046e-05,
"NMSE": 0.008940097606031181,
"SQNR_dB": 20.486577396459907,
"max_absolute_error": 0.109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.33.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0028732288851141424,
"MSE": 1.5106750101689894e-05,
"NMSE": 0.008992688798242662,
"SQNR_dB": 20.461104355584638,
"max_absolute_error": 0.052734375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.33.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0038010190761716273,
"MSE": 2.6829962672218867e-05,
"NMSE": 0.009032849881004182,
"SQNR_dB": 20.441752073468166,
"max_absolute_error": 0.0419921875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.33.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.002991871277942682,
"MSE": 1.6304777878922024e-05,
"NMSE": 0.008996791434604766,
"SQNR_dB": 20.459123473045345,
"max_absolute_error": 0.0361328125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.33.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0037752394914703177,
"MSE": 2.5525868929093765e-05,
"NMSE": 0.008977021115152422,
"SQNR_dB": 20.46867753296738,
"max_absolute_error": 0.0625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.33.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003815376119687132,
"MSE": 2.6180711474175134e-05,
"NMSE": 0.008878517616327627,
"SQNR_dB": 20.516595392653873,
"max_absolute_error": 0.06640625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.34.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0035139856215606335,
"MSE": 2.311906132729048e-05,
"NMSE": 0.008954249129213384,
"SQNR_dB": 20.479708266448334,
"max_absolute_error": 0.103515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.34.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002844755744451812,
"MSE": 1.4790881130404716e-05,
"NMSE": 0.009038815290620188,
"SQNR_dB": 20.438884883811852,
"max_absolute_error": 0.0361328125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.34.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003943863561651495,
"MSE": 2.8876546504186167e-05,
"NMSE": 0.009012787011826316,
"SQNR_dB": 20.45140891962646,
"max_absolute_error": 0.046875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.34.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003039801394212017,
"MSE": 1.6883958930230243e-05,
"NMSE": 0.009001468178321948,
"SQNR_dB": 20.456866494782954,
"max_absolute_error": 0.03515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.34.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0038614471492190933,
"MSE": 2.676567003031431e-05,
"NMSE": 0.00895082101823631,
"SQNR_dB": 20.481371269929053,
"max_absolute_error": 0.0595703125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.34.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.00385207100211174,
"MSE": 2.6563360551601968e-05,
"NMSE": 0.00889071063679083,
"SQNR_dB": 20.510635243726092,
"max_absolute_error": 0.0546875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.35.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034862615472487143,
"MSE": 2.310689078094821e-05,
"NMSE": 0.008956297756828026,
"SQNR_dB": 20.478714765045996,
"max_absolute_error": 0.201171875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.35.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002792049321980017,
"MSE": 1.4357017992471206e-05,
"NMSE": 0.008978183877801681,
"SQNR_dB": 20.468115042699804,
"max_absolute_error": 0.0419921875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.35.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0040977816191864105,
"MSE": 3.100277227023867e-05,
"NMSE": 0.009012641581388484,
"SQNR_dB": 20.451478998000823,
"max_absolute_error": 0.0556640625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.35.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0029966085173114693,
"MSE": 1.6346758147019782e-05,
"NMSE": 0.0089889399854727,
"SQNR_dB": 20.46291519123568,
"max_absolute_error": 0.0419921875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.35.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.004011742724875065,
"MSE": 2.8792448021377245e-05,
"NMSE": 0.008982358182926123,
"SQNR_dB": 20.466096308846634,
"max_absolute_error": 0.04736328125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.35.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003968096528236104,
"MSE": 2.809183987437281e-05,
"NMSE": 0.008934194404295331,
"SQNR_dB": 20.489446016969374,
"max_absolute_error": 0.10546875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.36.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003431574598219469,
"MSE": 2.2436078278394187e-05,
"NMSE": 0.008933512372270335,
"SQNR_dB": 20.489777567905133,
"max_absolute_error": 0.099609375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.36.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0028358899432854687,
"MSE": 1.4859102956476997e-05,
"NMSE": 0.009022240339849387,
"SQNR_dB": 20.446856080873314,
"max_absolute_error": 0.0478515625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.36.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.004327245570602105,
"MSE": 3.4430506193314404e-05,
"NMSE": 0.009042029888482926,
"SQNR_dB": 20.4373406173639,
"max_absolute_error": 0.109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.36.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003123316511505452,
"MSE": 1.8035800854991226e-05,
"NMSE": 0.009000632973676455,
"SQNR_dB": 20.457269475263953,
"max_absolute_error": 0.046875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.36.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.004182853492700052,
"MSE": 3.1419884922627314e-05,
"NMSE": 0.008950542816908397,
"SQNR_dB": 20.48150625551193,
"max_absolute_error": 0.0673828125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.36.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0039556616850222985,
"MSE": 2.7828834524140962e-05,
"NMSE": 0.008850550259835345,
"SQNR_dB": 20.530297273382047,
"max_absolute_error": 0.0615234375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.37.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0034676514078768195,
"MSE": 2.3052427002281958e-05,
"NMSE": 0.008957409940085531,
"SQNR_dB": 20.478175496391632,
"max_absolute_error": 0.1484375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.37.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.002732102777653073,
"MSE": 1.3892819682677367e-05,
"NMSE": 0.008998569680486904,
"SQNR_dB": 20.45826516031502,
"max_absolute_error": 0.04541015625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.37.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.004250096905789341,
"MSE": 3.294119506710229e-05,
"NMSE": 0.009078766126779058,
"SQNR_dB": 20.419731713845962,
"max_absolute_error": 0.15234375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.37.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003035385449063988,
"MSE": 1.705895056125665e-05,
"NMSE": 0.008969086827391833,
"SQNR_dB": 20.47251771673068,
"max_absolute_error": 0.0546875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.37.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003888103898401923,
"MSE": 2.704388039265744e-05,
"NMSE": 0.008936951559534367,
"SQNR_dB": 20.488105960413417,
"max_absolute_error": 0.0673828125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.37.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.004141834944848645,
"MSE": 3.0631862497923625e-05,
"NMSE": 0.008835714684762876,
"SQNR_dB": 20.537583164350156,
"max_absolute_error": 0.064453125,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.38.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003371744836650392,
"MSE": 2.2388271945339585e-05,
"NMSE": 0.008967042439660209,
"SQNR_dB": 20.47350774785874,
"max_absolute_error": 0.09375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.38.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0027128781644056915,
"MSE": 1.3760749410356896e-05,
"NMSE": 0.008969496837071647,
"SQNR_dB": 20.472319189416787,
"max_absolute_error": 0.037109375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.38.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0045712991342280644,
"MSE": 3.816539398089178e-05,
"NMSE": 0.009048812696617167,
"SQNR_dB": 20.434084012516802,
"max_absolute_error": 0.119140625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.38.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.003047848164128539,
"MSE": 1.7302382998069733e-05,
"NMSE": 0.009019421562476008,
"SQNR_dB": 20.448213139303476,
"max_absolute_error": 0.0458984375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.38.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.00434180867724736,
"MSE": 3.3663459425812185e-05,
"NMSE": 0.008956433983484797,
"SQNR_dB": 20.478648708694365,
"max_absolute_error": 0.0498046875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.38.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.004259752158087101,
"MSE": 3.246349143834902e-05,
"NMSE": 0.008848511579882026,
"SQNR_dB": 20.53129776421713,
"max_absolute_error": 0.083984375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.39.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.003342910395490711,
"MSE": 2.2791060164330294e-05,
"NMSE": 0.00895064972238132,
"SQNR_dB": 20.481454383596834,
"max_absolute_error": 0.09765625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.39.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0026594535059637847,
"MSE": 1.3146836643730972e-05,
"NMSE": 0.008991502414749285,
"SQNR_dB": 20.46167734749652,
"max_absolute_error": 0.03466796875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.39.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.004959142127730881,
"MSE": 4.5344492063796256e-05,
"NMSE": 0.009066672124825988,
"SQNR_dB": 20.425520892390196,
"max_absolute_error": 0.123046875,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.39.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0030388489959714982,
"MSE": 1.7196021193677367e-05,
"NMSE": 0.008966714720074228,
"SQNR_dB": 20.47366647288947,
"max_absolute_error": 0.056640625,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.39.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.004647489862815887,
"MSE": 3.8506453060848786e-05,
"NMSE": 0.008984087984435445,
"SQNR_dB": 20.465260035222332,
"max_absolute_error": 0.0859375,
"quantized": true,
"subclass": "NVFP4Tensor",
"tensor": "layers.39.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.40.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.40.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.40.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.40.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.40.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.40.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.41.proj.k.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.41.proj.o.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.41.proj.q.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.41.proj.v.weight",
"weight_elements": 9437184
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "final_linear_video.weight",
"weight_elements": 393216
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.41.proj.ffn_down.weight",
"weight_elements": 25165824
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "layers.41.proj.ffn_up_gate.weight",
"weight_elements": 50331648
},
{
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"SQNR_dB": null,
"max_absolute_error": 0.0,
"quantized": false,
"subclass": "Tensor",
"tensor": "video_head_modulation.weight",
"weight_elements": 18874368
}
],
"protected_bf16_weights": {
"MAE": 0.0,
"MSE": 0.0,
"NMSE": 0.0,
"RMSE": 0.0,
"SQNR_dB": null,
"cosine_similarity": 1.0,
"layers": 41,
"max_absolute_error": 0.0,
"reconstructed_sum_sq": 1499966.1311009347,
"scope": "parameter_weighted_aggregate",
"source_sum_sq": 1499966.1311009347,
"squared_error_sum": 0.0,
"weight_elements": 778567680
},
"quantized_linear_layers_expected": 228,
"quantized_linear_layers_observed": 228,
"schema_version": 1,
"scope": "core",
"status": "COMPLETE_269_LINEAR_WEIGHTS",
"variant": "nvfp4-core"
}