Experiment: dtufc_elic-featurecoding_falconmamba_individual Log file: output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/dtufc_elic-featurecoding_falconmamba_individual.log DTUFCCodecConfig: arch: elic-featurecoding handler: falconmamba checkpoint: codec_weights/elic_hybrid/elic2022-official_lambda0.02_epochs600_lr0.0001_bs60_patch256-256_checkpoint_best.pth.tar transform_type: kmeans transform_mapping:featurecoding_utils/transform_mapping/kmeans10samples-8bits/mapping.json bit_depth: 8 device: cuda:0 Loading checkpoint: codec_weights/elic_hybrid/elic2022-official_lambda0.02_epochs600_lr0.0001_bs60_patch256-256_checkpoint_best.pth.tar Checkpoint epoch: 520 Loaded elic-featurecoding (1-channel) on cuda:0 Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_0_conv.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.0.conv_state' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_0_conv.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_0_ssm.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.0.ssm_state' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_0_ssm.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_1_conv.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.1.conv_state' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_1_conv.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_1_ssm.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.1.ssm_state' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_1_ssm.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_2_conv.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.2.conv_state' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_2_conv.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_2_ssm.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.2.ssm_state' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_2_ssm.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_3_conv.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.3.conv_state' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_3_conv.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_3_ssm.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.3.ssm_state' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_3_ssm.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_4_conv.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.4.conv_state' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_4_conv.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_4_ssm.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.4.ssm_state' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_layer_4_ssm.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_output.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.4.output' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/falconmamba/arc_fewshot-8bit_output.json Loaded per-key mappings: model=falconmamba Keys: ['layer.0.conv_state', 'layer.0.ssm_state', 'layer.1.conv_state', 'layer.1.ssm_state', 'layer.2.conv_state', 'layer.2.ssm_state', 'layer.3.conv_state', 'layer.3.ssm_state', 'layer.4.conv_state', 'layer.4.ssm_state', 'layer.4.output'] ---------------- ------------------------------------------------------------------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding Checkpoint codec_weights/elic_hybrid/elic2022-official_lambda0.02_epochs600_lr0.0001_bs60_patch256-256_checkpoint_best.pth.tar Transform type kmeans Transform config featurecoding_utils/transform_mapping/kmeans10samples-8bits/mapping.json Input ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1 Output output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1 ---------------- ------------------------------------------------------------------------------------------------------------------- Files found: 100 ---------------------------------------------------------------------- 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample0-layer4-item1.zst (1/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample0-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 275, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 275, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 275, 4096]) -> torch.Size([1, 1, 275, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,180B, BPFP=4.7107 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 40,980B, BPFP=2.5012 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,516B, BPFP=8.9150 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,272B, BPFP=4.9604 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,116B, BPFP=10.5264 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,300B, BPFP=3.7415 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,532B, BPFP=10.3838 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,860B, BPFP=3.1653 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,272B, BPFP=9.0996 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 327,964B, BPFP=2.3293 ⌛️ [2/4] FRONTEND: Frontend time: 3.143s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 275, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.454s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 275, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000142 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000022 0.00004030 layer.1.conv_state 0.00048975 0.22688635 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00012844 0.04403996 layer.3.ssm_state 0.00000001 0.00000302 layer.3.conv_state 0.00007646 0.03849922 layer.4.ssm_state 0.00000001 0.00000466 layer.4.conv_state 0.00024491 0.08618877 layer.4.output 0.00000149 0.00012584 ------------------------------------------------------------------------------------- TOTAL 0.00001917 0.00807814 (elements=1,945,600) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1945600 Total Bytes 837136 BPFP 3.4422 bits/point EBPFP 5.5358 equivalent bits/point MSE 0.008078 ---------------------- -------------------------------------------------------- Time: 5.609s Load: 0.012s, Pack+Encode: 3.143s, Decode+Unpack: 2.454s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 275, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0081 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample0-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample0-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample1-layer4-item1.zst (2/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample1-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 282, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.013s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 282, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 282, 4096]) -> torch.Size([1, 1, 282, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,660B, BPFP=4.7400 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,360B, BPFP=2.5244 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,464B, BPFP=8.9023 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,224B, BPFP=4.9575 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,160B, BPFP=10.5371 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,532B, BPFP=3.6946 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,480B, BPFP=10.3711 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,612B, BPFP=3.1501 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,088B, BPFP=9.0547 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 334,008B, BPFP=2.3133 ⌛️ [2/4] FRONTEND: Frontend time: 2.681s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 282, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.438s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 282, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000029 0.00003769 layer.1.conv_state 0.00048919 0.22875094 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00011763 0.04332800 layer.3.ssm_state 0.00000001 0.00000280 layer.3.conv_state 0.00011288 0.03791651 layer.4.ssm_state 0.00000001 0.00000457 layer.4.conv_state 0.00028411 0.08773727 layer.4.output 0.00000122 0.00012083 ------------------------------------------------------------------------------------- TOTAL 0.00001982 0.00799469 (elements=1,974,272) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1974272 Total Bytes 842732 BPFP 3.4149 bits/point EBPFP 5.4763 equivalent bits/point MSE 0.007995 ---------------------- -------------------------------------------------------- Time: 5.131s Load: 0.013s, Pack+Encode: 2.681s, Decode+Unpack: 2.438s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 282, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0080 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample1-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample1-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample10-layer4-item1.zst (3/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample10-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 274, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 274, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 274, 4096]) -> torch.Size([1, 1, 274, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,660B, BPFP=4.6790 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,144B, BPFP=2.5112 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,568B, BPFP=8.9277 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,004B, BPFP=4.9441 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,292B, BPFP=10.5693 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,532B, BPFP=3.6946 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,572B, BPFP=10.3936 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,608B, BPFP=3.1499 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,292B, BPFP=9.1045 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 341,268B, BPFP=2.4326 ⌛️ [2/4] FRONTEND: Frontend time: 2.698s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 274, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.435s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 274, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000149 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000027 0.00004056 layer.1.conv_state 0.00049366 0.22744448 layer.2.ssm_state 0.00000001 0.00000212 layer.2.conv_state 0.00012530 0.04391384 layer.3.ssm_state 0.00000001 0.00000291 layer.3.conv_state 0.00007253 0.03788943 layer.4.ssm_state 0.00000001 0.00000468 layer.4.conv_state 0.00034856 0.09100952 layer.4.output 0.00000124 0.00013641 ------------------------------------------------------------------------------------- TOTAL 0.00002076 0.00817941 (elements=1,941,504) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1941504 Total Bytes 849084 BPFP 3.4987 bits/point EBPFP 5.5911 equivalent bits/point MSE 0.008179 ---------------------- -------------------------------------------------------- Time: 5.145s Load: 0.012s, Pack+Encode: 2.698s, Decode+Unpack: 2.435s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 274, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0082 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample10-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample10-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample101-layer4-item1.zst (4/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample101-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 165, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 165, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 165, 4096]) -> torch.Size([1, 1, 165, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,992B, BPFP=4.6382 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,164B, BPFP=2.5125 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,424B, BPFP=8.8926 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,184B, BPFP=4.8940 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,136B, BPFP=10.5312 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,948B, BPFP=3.7200 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,532B, BPFP=10.3838 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,100B, BPFP=3.1799 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,240B, BPFP=9.0918 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 221,228B, BPFP=2.6187 ⌛️ [2/4] FRONTEND: Frontend time: 2.693s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 165, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.322s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 165, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000027 0.00003990 layer.1.conv_state 0.00048753 0.22589837 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00013029 0.04395976 layer.3.ssm_state 0.00000001 0.00000285 layer.3.conv_state 0.00011288 0.03800382 layer.4.ssm_state 0.00000001 0.00000469 layer.4.conv_state 0.00021616 0.08879586 layer.4.output 0.00000217 0.00019821 ------------------------------------------------------------------------------------- TOTAL 0.00002496 0.01053027 (elements=1,495,040) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1495040 Total Bytes 728092 BPFP 3.8960 bits/point EBPFP 6.6083 equivalent bits/point MSE 0.010530 ---------------------- -------------------------------------------------------- Time: 5.024s Load: 0.009s, Pack+Encode: 2.693s, Decode+Unpack: 2.322s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 165, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0105 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample101-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample101-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample102-layer4-item1.zst (5/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample102-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 162, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 162, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 162, 4096]) -> torch.Size([1, 1, 162, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,484B, BPFP=4.6682 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,288B, BPFP=2.5200 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,452B, BPFP=8.8994 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,856B, BPFP=4.9961 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,156B, BPFP=10.5361 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,252B, BPFP=3.6775 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,584B, BPFP=10.3965 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,156B, BPFP=3.1833 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,324B, BPFP=9.1123 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 215,772B, BPFP=2.6014 ⌛️ [2/4] FRONTEND: Frontend time: 2.605s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 162, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.325s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 162, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000145 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00004213 layer.1.conv_state 0.00049337 0.22629075 layer.2.ssm_state 0.00000001 0.00000213 layer.2.conv_state 0.00015958 0.04370106 layer.3.ssm_state 0.00000001 0.00000278 layer.3.conv_state 0.00007099 0.03779954 layer.4.ssm_state 0.00000001 0.00000459 layer.4.conv_state 0.00026744 0.09094489 layer.4.output 0.00000226 0.00017508 ------------------------------------------------------------------------------------- TOTAL 0.00002617 0.01065165 (elements=1,482,752) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1482752 Total Bytes 724468 BPFP 3.9088 bits/point EBPFP 6.6534 equivalent bits/point MSE 0.010652 ---------------------- -------------------------------------------------------- Time: 4.942s Load: 0.012s, Pack+Encode: 2.605s, Decode+Unpack: 2.325s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 162, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0107 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample102-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample102-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample106-layer4-item1.zst (6/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample106-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 185, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 185, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 185, 4096]) -> torch.Size([1, 1, 185, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,932B, BPFP=4.6345 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,204B, BPFP=2.5149 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,512B, BPFP=8.9141 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 78,488B, BPFP=4.7905 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,272B, BPFP=10.5645 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,736B, BPFP=3.6460 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,676B, BPFP=10.4189 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,968B, BPFP=3.1719 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,368B, BPFP=9.1230 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 227,572B, BPFP=2.4026 ⌛️ [2/4] FRONTEND: Frontend time: 2.579s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 185, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.320s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 185, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000146 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000038 0.00004153 layer.1.conv_state 0.00049648 0.22957774 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00012672 0.04448884 layer.3.ssm_state 0.00000001 0.00000273 layer.3.conv_state 0.00007687 0.03782637 layer.4.ssm_state 0.00000001 0.00000448 layer.4.conv_state 0.00032793 0.08423967 layer.4.output 0.00000201 0.00018163 ------------------------------------------------------------------------------------- TOTAL 0.00002540 0.00997477 (elements=1,576,960) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1576960 Total Bytes 731872 BPFP 3.7128 bits/point EBPFP 6.2712 equivalent bits/point MSE 0.009975 ---------------------- -------------------------------------------------------- Time: 4.910s Load: 0.012s, Pack+Encode: 2.579s, Decode+Unpack: 2.320s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 185, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0100 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample106-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample106-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample108-layer4-item1.zst (7/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample108-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 161, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 161, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 161, 4096]) -> torch.Size([1, 1, 161, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,172B, BPFP=4.6492 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,084B, BPFP=2.5076 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,516B, BPFP=8.9150 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,624B, BPFP=4.9209 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,196B, BPFP=10.5459 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,884B, BPFP=3.6550 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,520B, BPFP=10.3809 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,192B, BPFP=3.2466 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,184B, BPFP=9.0781 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 209,804B, BPFP=2.5452 ⌛️ [2/4] FRONTEND: Frontend time: 2.570s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 161, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.321s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 161, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000141 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000024 0.00004135 layer.1.conv_state 0.00050173 0.22834530 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00011340 0.04413517 layer.3.ssm_state 0.00000001 0.00000273 layer.3.conv_state 0.00011848 0.03786190 layer.4.ssm_state 0.00000005 0.00000411 layer.4.conv_state 0.00022485 0.08313771 layer.4.output 0.00000221 0.00017705 ------------------------------------------------------------------------------------- TOTAL 0.00002549 0.01056496 (elements=1,478,656) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1478656 Total Bytes 717320 BPFP 3.8809 bits/point EBPFP 6.6268 equivalent bits/point MSE 0.010565 ---------------------- -------------------------------------------------------- Time: 4.902s Load: 0.011s, Pack+Encode: 2.570s, Decode+Unpack: 2.321s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 161, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0106 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample108-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample108-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample110-layer4-item1.zst (8/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample110-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 198, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 198, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 198, 4096]) -> torch.Size([1, 1, 198, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,196B, BPFP=4.6506 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,268B, BPFP=2.5188 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,524B, BPFP=8.9170 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,348B, BPFP=4.9041 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,292B, BPFP=10.5693 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,140B, BPFP=3.6707 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,656B, BPFP=10.4141 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,060B, BPFP=3.1775 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,432B, BPFP=9.1387 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 263,236B, BPFP=2.5966 ⌛️ [2/4] FRONTEND: Frontend time: 2.742s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 198, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.383s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 198, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000010 0.00004059 layer.1.conv_state 0.00048181 0.22941482 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00010981 0.04392899 layer.3.ssm_state 0.00000001 0.00000291 layer.3.conv_state 0.00007216 0.03816640 layer.4.ssm_state 0.00000002 0.00000469 layer.4.conv_state 0.00032426 0.08382134 layer.4.output 0.00000183 0.00015530 ------------------------------------------------------------------------------------- TOTAL 0.00002372 0.00962564 (elements=1,630,208) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1630208 Total Bytes 770296 BPFP 3.7801 bits/point EBPFP 6.2684 equivalent bits/point MSE 0.009626 ---------------------- -------------------------------------------------------- Time: 5.135s Load: 0.010s, Pack+Encode: 2.742s, Decode+Unpack: 2.383s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 198, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0096 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample110-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample110-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample115-layer4-item1.zst (9/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample115-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 170, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 170, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 170, 4096]) -> torch.Size([1, 1, 170, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,032B, BPFP=4.7627 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,152B, BPFP=2.5117 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,456B, BPFP=8.9004 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,688B, BPFP=4.8638 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,100B, BPFP=10.5225 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,344B, BPFP=3.6831 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,436B, BPFP=10.3604 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,200B, BPFP=3.1860 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,176B, BPFP=9.0762 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 217,700B, BPFP=2.5011 ⌛️ [2/4] FRONTEND: Frontend time: 2.569s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 170, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.323s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 170, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000141 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004103 layer.1.conv_state 0.00049252 0.22750086 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00011876 0.04367338 layer.3.ssm_state 0.00000001 0.00000288 layer.3.conv_state 0.00011952 0.03752621 layer.4.ssm_state 0.00000001 0.00000478 layer.4.conv_state 0.00023872 0.08583158 layer.4.output 0.00000204 0.00016500 ------------------------------------------------------------------------------------- TOTAL 0.00002508 0.01032953 (elements=1,515,520) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1515520 Total Bytes 725428 BPFP 3.8293 bits/point EBPFP 6.5095 equivalent bits/point MSE 0.010330 ---------------------- -------------------------------------------------------- Time: 4.905s Load: 0.012s, Pack+Encode: 2.569s, Decode+Unpack: 2.323s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 170, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0103 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample115-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample115-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample116-layer4-item1.zst (10/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample116-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 190, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 190, 4096]) -> torch.Size([1, 1, 190, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,712B, BPFP=4.6821 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,104B, BPFP=2.5088 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,452B, BPFP=8.8994 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,644B, BPFP=4.9832 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,180B, BPFP=10.5420 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,032B, BPFP=3.6641 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,452B, BPFP=10.3643 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,052B, BPFP=3.2380 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,276B, BPFP=9.1006 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 207,908B, BPFP=2.1372 ⌛️ [2/4] FRONTEND: Frontend time: 2.580s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 190, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.317s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 190, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000145 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000023 0.00004234 layer.1.conv_state 0.00050470 0.22991328 layer.2.ssm_state 0.00000001 0.00000213 layer.2.conv_state 0.00013213 0.04375423 layer.3.ssm_state 0.00000001 0.00000286 layer.3.conv_state 0.00011581 0.03800046 layer.4.ssm_state 0.00000003 0.00000441 layer.4.conv_state 0.00021685 0.08675583 layer.4.output 0.00000199 0.00019850 ------------------------------------------------------------------------------------- TOTAL 0.00002388 0.00990450 (elements=1,597,440) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1597440 Total Bytes 716956 BPFP 3.5905 bits/point EBPFP 6.1398 equivalent bits/point MSE 0.009905 ---------------------- -------------------------------------------------------- Time: 4.909s Load: 0.012s, Pack+Encode: 2.580s, Decode+Unpack: 2.317s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0099 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample116-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample116-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample12-layer4-item1.zst (11/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample12-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 231, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.014s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 231, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 231, 4096]) -> torch.Size([1, 1, 231, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,676B, BPFP=4.7410 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,240B, BPFP=2.5171 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,428B, BPFP=8.8936 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,220B, BPFP=4.8962 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,164B, BPFP=10.5381 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,016B, BPFP=3.6631 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,508B, BPFP=10.3779 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,584B, BPFP=3.2095 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,384B, BPFP=9.1270 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 279,428B, BPFP=2.3626 ⌛️ [2/4] FRONTEND: Frontend time: 2.634s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 231, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.391s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 231, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000148 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000024 0.00004156 layer.1.conv_state 0.00050504 0.22918889 layer.2.ssm_state 0.00000001 0.00000213 layer.2.conv_state 0.00012981 0.04427716 layer.3.ssm_state 0.00000001 0.00000286 layer.3.conv_state 0.00007554 0.03757189 layer.4.ssm_state 0.00000001 0.00000464 layer.4.conv_state 0.00028923 0.08973210 layer.4.output 0.00000157 0.00013048 ------------------------------------------------------------------------------------- TOTAL 0.00002213 0.00898825 (elements=1,765,376) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1765376 Total Bytes 787792 BPFP 3.5700 bits/point EBPFP 5.8737 equivalent bits/point MSE 0.008988 ---------------------- -------------------------------------------------------- Time: 5.039s Load: 0.014s, Pack+Encode: 2.634s, Decode+Unpack: 2.391s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 231, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0090 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample12-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample12-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample120-layer4-item1.zst (12/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample120-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 183, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 183, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) -> torch.Size([1, 1, 183, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,144B, BPFP=4.7695 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,124B, BPFP=2.5100 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,464B, BPFP=8.9023 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,196B, BPFP=4.8948 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,188B, BPFP=10.5439 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,244B, BPFP=3.6160 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,464B, BPFP=10.3672 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,512B, BPFP=3.2051 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,332B, BPFP=9.1143 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 217,224B, BPFP=2.3184 ⌛️ [2/4] FRONTEND: Frontend time: 2.589s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.329s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000024 0.00004280 layer.1.conv_state 0.00048839 0.22643882 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00008466 0.04359639 layer.3.ssm_state 0.00000001 0.00000291 layer.3.conv_state 0.00011958 0.03767862 layer.4.ssm_state 0.00000001 0.00000478 layer.4.conv_state 0.00031804 0.09028312 layer.4.output 0.00000201 0.00016236 ------------------------------------------------------------------------------------- TOTAL 0.00002515 0.01005579 (elements=1,568,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1568768 Total Bytes 725036 BPFP 3.6974 bits/point EBPFP 6.2870 equivalent bits/point MSE 0.010056 ---------------------- -------------------------------------------------------- Time: 4.928s Load: 0.011s, Pack+Encode: 2.589s, Decode+Unpack: 2.329s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 183, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0101 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample120-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample120-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample122-layer4-item1.zst (13/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample122-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 191, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 191, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 191, 4096]) -> torch.Size([1, 1, 191, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,496B, BPFP=4.6689 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,128B, BPFP=2.5103 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,524B, BPFP=8.9170 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,788B, BPFP=4.9309 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,164B, BPFP=10.5381 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,792B, BPFP=3.7104 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,596B, BPFP=10.3994 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 54,000B, BPFP=3.2959 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,196B, BPFP=9.0811 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 220,684B, BPFP=2.2567 ⌛️ [2/4] FRONTEND: Frontend time: 2.601s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 191, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.324s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 191, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000142 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004101 layer.1.conv_state 0.00046948 0.22536540 layer.2.ssm_state 0.00000001 0.00000213 layer.2.conv_state 0.00011482 0.04439262 layer.3.ssm_state 0.00000001 0.00000310 layer.3.conv_state 0.00011212 0.03744067 layer.4.ssm_state 0.00000002 0.00000402 layer.4.conv_state 0.00024277 0.07955752 layer.4.output 0.00000205 0.00019752 ------------------------------------------------------------------------------------- TOTAL 0.00002323 0.00964035 (elements=1,601,536) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1601536 Total Bytes 730512 BPFP 3.6491 bits/point EBPFP 6.1958 equivalent bits/point MSE 0.009640 ---------------------- -------------------------------------------------------- Time: 4.935s Load: 0.010s, Pack+Encode: 2.601s, Decode+Unpack: 2.324s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 191, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0096 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample122-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample122-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample127-layer4-item1.zst (14/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample127-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 175, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 175, 4096]) -> torch.Size([1, 1, 175, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,160B, BPFP=4.7095 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,176B, BPFP=2.5132 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,468B, BPFP=8.9033 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 82,356B, BPFP=5.0266 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,072B, BPFP=10.5156 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,756B, BPFP=3.7083 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,424B, BPFP=10.3574 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,784B, BPFP=3.1606 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,236B, BPFP=9.0908 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 222,608B, BPFP=2.4845 ⌛️ [2/4] FRONTEND: Frontend time: 2.603s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 175, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.334s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 175, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000148 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004268 layer.1.conv_state 0.00051384 0.22689469 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00012234 0.04353094 layer.3.ssm_state 0.00000001 0.00000300 layer.3.conv_state 0.00007815 0.03722347 layer.4.ssm_state 0.00000001 0.00000484 layer.4.conv_state 0.00034266 0.08820990 layer.4.output 0.00000210 0.00017650 ------------------------------------------------------------------------------------- TOTAL 0.00002667 0.01022784 (elements=1,536,000) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1536000 Total Bytes 732184 BPFP 3.8135 bits/point EBPFP 6.4675 equivalent bits/point MSE 0.010228 ---------------------- -------------------------------------------------------- Time: 4.949s Load: 0.012s, Pack+Encode: 2.603s, Decode+Unpack: 2.334s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0102 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample127-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample127-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample128-layer4-item1.zst (15/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample128-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 168, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 168, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) -> torch.Size([1, 1, 168, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,884B, BPFP=4.6926 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,184B, BPFP=2.5137 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,416B, BPFP=8.8906 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,580B, BPFP=4.9182 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,100B, BPFP=10.5225 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,256B, BPFP=3.6777 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,476B, BPFP=10.3701 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,784B, BPFP=3.2217 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,196B, BPFP=9.0811 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 219,548B, BPFP=2.5524 ⌛️ [2/4] FRONTEND: Frontend time: 2.589s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.329s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000149 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004082 layer.1.conv_state 0.00051486 0.22656208 layer.2.ssm_state 0.00000001 0.00000219 layer.2.conv_state 0.00016795 0.04375239 layer.3.ssm_state 0.00000001 0.00000286 layer.3.conv_state 0.00011430 0.03791069 layer.4.ssm_state 0.00000002 0.00000425 layer.4.conv_state 0.00021850 0.08173884 layer.4.output 0.00000215 0.00017768 ------------------------------------------------------------------------------------- TOTAL 0.00002626 0.01029120 (elements=1,507,328) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1507328 Total Bytes 727568 BPFP 3.8615 bits/point EBPFP 6.5578 equivalent bits/point MSE 0.010291 ---------------------- -------------------------------------------------------- Time: 4.927s Load: 0.009s, Pack+Encode: 2.589s, Decode+Unpack: 2.329s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 168, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0103 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample128-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample128-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample129-layer4-item1.zst (16/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample129-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 181, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 181, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 181, 4096]) -> torch.Size([1, 1, 181, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,264B, BPFP=4.7158 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,188B, BPFP=2.5139 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,476B, BPFP=8.9053 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 82,296B, BPFP=5.0229 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,144B, BPFP=10.5332 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,712B, BPFP=3.6445 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,364B, BPFP=10.3428 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,136B, BPFP=3.1821 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,264B, BPFP=9.0977 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 219,028B, BPFP=2.3635 ⌛️ [2/4] FRONTEND: Frontend time: 2.585s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 181, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.320s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 181, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000145 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000027 0.00004029 layer.1.conv_state 0.00048726 0.22867660 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00012801 0.04336672 layer.3.ssm_state 0.00000001 0.00000288 layer.3.conv_state 0.00011944 0.03667283 layer.4.ssm_state 0.00000001 0.00000459 layer.4.conv_state 0.00034131 0.09138912 layer.4.output 0.00000206 0.00018708 ------------------------------------------------------------------------------------- TOTAL 0.00002667 0.01016351 (elements=1,560,576) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1560576 Total Bytes 728016 BPFP 3.7320 bits/point EBPFP 6.3413 equivalent bits/point MSE 0.010164 ---------------------- -------------------------------------------------------- Time: 4.916s Load: 0.011s, Pack+Encode: 2.585s, Decode+Unpack: 2.320s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 181, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0102 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample129-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample129-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample130-layer4-item1.zst (17/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample130-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 166, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) -> torch.Size([1, 1, 166, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,744B, BPFP=4.7451 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,184B, BPFP=2.5137 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,584B, BPFP=8.9316 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,860B, BPFP=4.9353 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,144B, BPFP=10.5332 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,592B, BPFP=3.6982 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,508B, BPFP=10.3779 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,992B, BPFP=3.1733 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,164B, BPFP=9.0732 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 216,972B, BPFP=2.5529 ⌛️ [2/4] FRONTEND: Frontend time: 2.578s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.313s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000147 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000010 0.00003945 layer.1.conv_state 0.00049597 0.22728816 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00012619 0.04375685 layer.3.ssm_state 0.00000001 0.00000299 layer.3.conv_state 0.00006821 0.03734361 layer.4.ssm_state 0.00000001 0.00000478 layer.4.conv_state 0.00029570 0.08547772 layer.4.output 0.00000206 0.00016569 ------------------------------------------------------------------------------------- TOTAL 0.00002569 0.01042626 (elements=1,499,136) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1499136 Total Bytes 725888 BPFP 3.8736 bits/point EBPFP 6.5894 equivalent bits/point MSE 0.010426 ---------------------- -------------------------------------------------------- Time: 4.900s Load: 0.009s, Pack+Encode: 2.578s, Decode+Unpack: 2.313s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0104 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample130-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample130-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample134-layer4-item1.zst (18/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample134-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 182, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 182, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 182, 4096]) -> torch.Size([1, 1, 182, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,252B, BPFP=4.6541 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,356B, BPFP=2.5242 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,484B, BPFP=8.9072 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,972B, BPFP=4.9421 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,084B, BPFP=10.5186 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,460B, BPFP=3.6292 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,392B, BPFP=10.3496 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,612B, BPFP=3.2112 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,048B, BPFP=9.0449 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 224,840B, BPFP=2.4129 ⌛️ [2/4] FRONTEND: Frontend time: 2.583s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 182, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.320s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 182, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000139 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000027 0.00004044 layer.1.conv_state 0.00049604 0.22805680 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00012832 0.04351665 layer.3.ssm_state 0.00000001 0.00000275 layer.3.conv_state 0.00011179 0.03725317 layer.4.ssm_state 0.00000001 0.00000434 layer.4.conv_state 0.00022786 0.08616714 layer.4.output 0.00000216 0.00017167 ------------------------------------------------------------------------------------- TOTAL 0.00002431 0.01002298 (elements=1,564,672) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1564672 Total Bytes 731644 BPFP 3.7408 bits/point EBPFP 6.3321 equivalent bits/point MSE 0.010023 ---------------------- -------------------------------------------------------- Time: 4.912s Load: 0.010s, Pack+Encode: 2.583s, Decode+Unpack: 2.320s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 182, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0100 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample134-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample134-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample135-layer4-item1.zst (19/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample135-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 178, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 178, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 178, 4096]) -> torch.Size([1, 1, 178, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,540B, BPFP=4.6716 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,176B, BPFP=2.5132 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,524B, BPFP=8.9170 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,052B, BPFP=4.8860 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,200B, BPFP=10.5469 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,568B, BPFP=3.6968 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,524B, BPFP=10.3818 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,648B, BPFP=3.1523 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,180B, BPFP=9.0771 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 226,840B, BPFP=2.4890 ⌛️ [2/4] FRONTEND: Frontend time: 2.588s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 178, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.324s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 178, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000145 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00003930 layer.1.conv_state 0.00048610 0.22707574 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00014962 0.04370045 layer.3.ssm_state 0.00000001 0.00000291 layer.3.conv_state 0.00007021 0.03786935 layer.4.ssm_state 0.00000001 0.00000469 layer.4.conv_state 0.00031005 0.08658154 layer.4.output 0.00000198 0.00017905 ------------------------------------------------------------------------------------- TOTAL 0.00002555 0.01013558 (elements=1,548,288) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1548288 Total Bytes 733396 BPFP 3.7895 bits/point EBPFP 6.4068 equivalent bits/point MSE 0.010136 ---------------------- -------------------------------------------------------- Time: 4.923s Load: 0.011s, Pack+Encode: 2.588s, Decode+Unpack: 2.324s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 178, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0101 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample135-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample135-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample142-layer4-item1.zst (20/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample142-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 166, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) -> torch.Size([1, 1, 166, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,056B, BPFP=4.7031 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,040B, BPFP=2.5049 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,492B, BPFP=8.9092 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,608B, BPFP=4.8589 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,264B, BPFP=10.5625 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,488B, BPFP=3.6919 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,524B, BPFP=10.3818 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,544B, BPFP=3.2681 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,256B, BPFP=9.0957 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 216,056B, BPFP=2.5421 ⌛️ [2/4] FRONTEND: Frontend time: 2.587s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.326s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000140 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00004244 layer.1.conv_state 0.00050515 0.22616832 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00013851 0.04433801 layer.3.ssm_state 0.00000001 0.00000283 layer.3.conv_state 0.00011516 0.03802425 layer.4.ssm_state 0.00000001 0.00000413 layer.4.conv_state 0.00022993 0.08829465 layer.4.output 0.00000208 0.00016406 ------------------------------------------------------------------------------------- TOTAL 0.00002578 0.01049037 (elements=1,499,136) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1499136 Total Bytes 724472 BPFP 3.8661 bits/point EBPFP 6.5792 equivalent bits/point MSE 0.010490 ---------------------- -------------------------------------------------------- Time: 4.924s Load: 0.011s, Pack+Encode: 2.587s, Decode+Unpack: 2.326s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0105 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample142-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample142-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample143-layer4-item1.zst (21/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample143-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 154, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 154, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 154, 4096]) -> torch.Size([1, 1, 154, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,608B, BPFP=4.6758 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,144B, BPFP=2.5112 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,440B, BPFP=8.8965 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,512B, BPFP=4.8530 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,132B, BPFP=10.5303 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,032B, BPFP=3.6641 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,552B, BPFP=10.3887 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,308B, BPFP=3.1926 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,272B, BPFP=9.0996 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 210,284B, BPFP=2.6670 ⌛️ [2/4] FRONTEND: Frontend time: 2.600s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 154, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.320s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 154, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000142 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004199 layer.1.conv_state 0.00051446 0.22549982 layer.2.ssm_state 0.00000001 0.00000220 layer.2.conv_state 0.00012903 0.04354233 layer.3.ssm_state 0.00000001 0.00000288 layer.3.conv_state 0.00007035 0.03756766 layer.4.ssm_state 0.00000004 0.00000434 layer.4.conv_state 0.00022206 0.08029702 layer.4.output 0.00000218 0.00018527 ------------------------------------------------------------------------------------- TOTAL 0.00002543 0.01062549 (elements=1,449,984) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1449984 Total Bytes 716428 BPFP 3.9527 bits/point EBPFP 6.7453 equivalent bits/point MSE 0.010625 ---------------------- -------------------------------------------------------- Time: 4.929s Load: 0.010s, Pack+Encode: 2.600s, Decode+Unpack: 2.320s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 154, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0106 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample143-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample143-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample144-layer4-item1.zst (22/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample144-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 166, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) -> torch.Size([1, 1, 166, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,712B, BPFP=4.7432 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,264B, BPFP=2.5186 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,464B, BPFP=8.9023 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,984B, BPFP=4.9429 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,128B, BPFP=10.5293 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,676B, BPFP=3.7034 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,456B, BPFP=10.3652 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,760B, BPFP=3.1592 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,164B, BPFP=9.0732 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 217,440B, BPFP=2.5584 ⌛️ [2/4] FRONTEND: Frontend time: 2.579s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.311s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000148 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00003946 layer.1.conv_state 0.00048448 0.22440261 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00013520 0.04308666 layer.3.ssm_state 0.00000001 0.00000285 layer.3.conv_state 0.00007505 0.03723985 layer.4.ssm_state 0.00000002 0.00000493 layer.4.conv_state 0.00034342 0.08984430 layer.4.output 0.00000207 0.00016318 ------------------------------------------------------------------------------------- TOTAL 0.00002685 0.01044058 (elements=1,499,136) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1499136 Total Bytes 726192 BPFP 3.8753 bits/point EBPFP 6.5902 equivalent bits/point MSE 0.010441 ---------------------- -------------------------------------------------------- Time: 4.899s Load: 0.009s, Pack+Encode: 2.579s, Decode+Unpack: 2.311s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0104 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample144-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample144-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample148-layer4-item1.zst (23/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample148-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 197, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.013s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 197, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 197, 4096]) -> torch.Size([1, 1, 197, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,588B, BPFP=4.6746 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,224B, BPFP=2.5161 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,472B, BPFP=8.9043 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 78,668B, BPFP=4.8015 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,268B, BPFP=10.5635 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,068B, BPFP=3.6663 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,628B, BPFP=10.4072 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,840B, BPFP=3.2251 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,388B, BPFP=9.1279 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 262,764B, BPFP=2.6051 ⌛️ [2/4] FRONTEND: Frontend time: 2.639s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 197, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.379s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 197, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000147 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000027 0.00003979 layer.1.conv_state 0.00049519 0.22790083 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00014288 0.04451084 layer.3.ssm_state 0.00000001 0.00000289 layer.3.conv_state 0.00007032 0.03797150 layer.4.ssm_state 0.00000002 0.00000439 layer.4.conv_state 0.00023237 0.08315530 layer.4.output 0.00000183 0.00017742 ------------------------------------------------------------------------------------- TOTAL 0.00002283 0.00962425 (elements=1,626,112) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1626112 Total Bytes 769052 BPFP 3.7835 bits/point EBPFP 6.2743 equivalent bits/point MSE 0.009624 ---------------------- -------------------------------------------------------- Time: 5.031s Load: 0.013s, Pack+Encode: 2.639s, Decode+Unpack: 2.379s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 197, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0096 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample148-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample148-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample149-layer4-item1.zst (24/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample149-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 168, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 168, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) -> torch.Size([1, 1, 168, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,092B, BPFP=4.7664 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,276B, BPFP=2.5193 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,408B, BPFP=8.8887 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,428B, BPFP=4.9700 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,136B, BPFP=10.5312 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,516B, BPFP=3.6326 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,420B, BPFP=10.3564 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,240B, BPFP=3.1885 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,104B, BPFP=9.0586 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 217,220B, BPFP=2.5253 ⌛️ [2/4] FRONTEND: Frontend time: 2.579s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.325s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000149 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00004064 layer.1.conv_state 0.00048592 0.22926372 layer.2.ssm_state 0.00000001 0.00000219 layer.2.conv_state 0.00011654 0.04366936 layer.3.ssm_state 0.00000001 0.00000279 layer.3.conv_state 0.00011837 0.03750269 layer.4.ssm_state 0.00000001 0.00000427 layer.4.conv_state 0.00020779 0.08413486 layer.4.output 0.00000221 0.00017751 ------------------------------------------------------------------------------------- TOTAL 0.00002440 0.01039125 (elements=1,507,328) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1507328 Total Bytes 725984 BPFP 3.8531 bits/point EBPFP 6.5533 equivalent bits/point MSE 0.010391 ---------------------- -------------------------------------------------------- Time: 4.913s Load: 0.009s, Pack+Encode: 2.579s, Decode+Unpack: 2.325s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 168, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0104 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample149-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample149-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample153-layer4-item1.zst (25/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample153-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 164, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 164, 4096]) -> torch.Size([1, 1, 164, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,260B, BPFP=4.7156 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,136B, BPFP=2.5107 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,528B, BPFP=8.9180 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,424B, BPFP=4.8477 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,308B, BPFP=10.5732 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,544B, BPFP=3.6953 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,556B, BPFP=10.3896 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,272B, BPFP=3.2515 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,308B, BPFP=9.1084 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 214,044B, BPFP=2.5491 ⌛️ [2/4] FRONTEND: Frontend time: 2.588s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 164, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.331s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 164, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000146 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000029 0.00004292 layer.1.conv_state 0.00049823 0.22850658 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00014568 0.04520684 layer.3.ssm_state 0.00000001 0.00000293 layer.3.conv_state 0.00011339 0.03778366 layer.4.ssm_state 0.00000002 0.00000426 layer.4.conv_state 0.00023177 0.08647262 layer.4.output 0.00000211 0.00018886 ------------------------------------------------------------------------------------- TOTAL 0.00002593 0.01058351 (elements=1,490,944) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1490944 Total Bytes 722524 BPFP 3.8769 bits/point EBPFP 6.6052 equivalent bits/point MSE 0.010584 ---------------------- -------------------------------------------------------- Time: 4.931s Load: 0.012s, Pack+Encode: 2.588s, Decode+Unpack: 2.331s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0106 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample153-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample153-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample156-layer4-item1.zst (26/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample156-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 163, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 163, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 163, 4096]) -> torch.Size([1, 1, 163, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,712B, BPFP=4.7432 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,200B, BPFP=2.5146 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,484B, BPFP=8.9072 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,632B, BPFP=4.9214 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,108B, BPFP=10.5244 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,188B, BPFP=3.7346 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,432B, BPFP=10.3594 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,228B, BPFP=3.1877 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,164B, BPFP=9.0732 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 210,972B, BPFP=2.5279 ⌛️ [2/4] FRONTEND: Frontend time: 2.590s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 163, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.322s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 163, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000146 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000043 0.00004172 layer.1.conv_state 0.00049103 0.22615659 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00014094 0.04399301 layer.3.ssm_state 0.00000001 0.00000291 layer.3.conv_state 0.00012075 0.03802117 layer.4.ssm_state 0.00000001 0.00000464 layer.4.conv_state 0.00020319 0.08616225 layer.4.output 0.00000221 0.00016715 ------------------------------------------------------------------------------------- TOTAL 0.00002532 0.01052218 (elements=1,486,848) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1486848 Total Bytes 720264 BPFP 3.8754 bits/point EBPFP 6.6156 equivalent bits/point MSE 0.010522 ---------------------- -------------------------------------------------------- Time: 4.921s Load: 0.009s, Pack+Encode: 2.590s, Decode+Unpack: 2.322s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 163, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0105 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample156-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample156-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample158-layer4-item1.zst (27/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample158-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 162, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 162, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 162, 4096]) -> torch.Size([1, 1, 162, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,836B, BPFP=4.8118 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,212B, BPFP=2.5154 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,448B, BPFP=8.8984 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,276B, BPFP=4.9607 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,144B, BPFP=10.5332 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,736B, BPFP=3.7070 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,604B, BPFP=10.4014 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,480B, BPFP=3.2031 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,312B, BPFP=9.1094 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 209,500B, BPFP=2.5258 ⌛️ [2/4] FRONTEND: Frontend time: 2.574s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 162, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.323s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 162, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00004087 layer.1.conv_state 0.00050753 0.22984603 layer.2.ssm_state 0.00000001 0.00000219 layer.2.conv_state 0.00012648 0.04409989 layer.3.ssm_state 0.00000001 0.00000298 layer.3.conv_state 0.00011960 0.03816242 layer.4.ssm_state 0.00000001 0.00000461 layer.4.conv_state 0.00022111 0.08792534 layer.4.output 0.00000221 0.00016935 ------------------------------------------------------------------------------------- TOTAL 0.00002578 0.01067768 (elements=1,482,752) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1482752 Total Bytes 720692 BPFP 3.8884 bits/point EBPFP 6.6465 equivalent bits/point MSE 0.010678 ---------------------- -------------------------------------------------------- Time: 4.908s Load: 0.012s, Pack+Encode: 2.574s, Decode+Unpack: 2.323s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 162, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0107 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample158-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample158-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample159-layer4-item1.zst (28/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample159-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 162, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 162, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 162, 4096]) -> torch.Size([1, 1, 162, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,504B, BPFP=4.7305 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,204B, BPFP=2.5149 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,440B, BPFP=8.8965 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,992B, BPFP=4.9434 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,084B, BPFP=10.5186 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,708B, BPFP=3.6443 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,384B, BPFP=10.3477 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,544B, BPFP=3.2070 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,088B, BPFP=9.0547 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 210,084B, BPFP=2.5328 ⌛️ [2/4] FRONTEND: Frontend time: 2.590s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 162, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.326s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 162, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000151 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00004134 layer.1.conv_state 0.00049392 0.22728270 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00013642 0.04381878 layer.3.ssm_state 0.00000001 0.00000276 layer.3.conv_state 0.00011766 0.03683697 layer.4.ssm_state 0.00000003 0.00000443 layer.4.conv_state 0.00021047 0.08598651 layer.4.output 0.00000223 0.00017160 ------------------------------------------------------------------------------------- TOTAL 0.00002544 0.01054369 (elements=1,482,752) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1482752 Total Bytes 718176 BPFP 3.8748 bits/point EBPFP 6.6162 equivalent bits/point MSE 0.010544 ---------------------- -------------------------------------------------------- Time: 4.926s Load: 0.011s, Pack+Encode: 2.590s, Decode+Unpack: 2.326s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 162, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0105 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample159-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample159-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample162-layer4-item1.zst (29/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample162-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 151, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 151, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 151, 4096]) -> torch.Size([1, 1, 151, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,656B, BPFP=4.6787 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,268B, BPFP=2.5188 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,428B, BPFP=8.8936 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,684B, BPFP=4.9856 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,172B, BPFP=10.5400 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,296B, BPFP=3.7412 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,452B, BPFP=10.3643 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,284B, BPFP=3.1912 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,180B, BPFP=9.0771 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 210,140B, BPFP=2.7181 ⌛️ [2/4] FRONTEND: Frontend time: 2.580s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 151, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.327s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 151, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000148 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000012 0.00003994 layer.1.conv_state 0.00048080 0.22699024 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00015695 0.04398832 layer.3.ssm_state 0.00000001 0.00000289 layer.3.conv_state 0.00007366 0.03737149 layer.4.ssm_state 0.00000001 0.00000456 layer.4.conv_state 0.00025051 0.08928719 layer.4.output 0.00000211 0.00019914 ------------------------------------------------------------------------------------- TOTAL 0.00002617 0.01096510 (elements=1,437,696) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1437696 Total Bytes 719704 BPFP 4.0048 bits/point EBPFP 6.8402 equivalent bits/point MSE 0.010965 ---------------------- -------------------------------------------------------- Time: 4.918s Load: 0.011s, Pack+Encode: 2.580s, Decode+Unpack: 2.327s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 151, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0110 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample162-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample162-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample166-layer4-item1.zst (30/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample166-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 188, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 188, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 188, 4096]) -> torch.Size([1, 1, 188, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,352B, BPFP=4.5991 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,088B, BPFP=2.5078 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,612B, BPFP=8.9385 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,656B, BPFP=4.9229 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,452B, BPFP=10.6084 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,408B, BPFP=3.6870 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,712B, BPFP=10.4277 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,604B, BPFP=3.1497 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,348B, BPFP=9.1182 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 221,824B, BPFP=2.3045 ⌛️ [2/4] FRONTEND: Frontend time: 2.590s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 188, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.331s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 188, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000142 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00004054 layer.1.conv_state 0.00047809 0.22974256 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00013752 0.04435263 layer.3.ssm_state 0.00000001 0.00000289 layer.3.conv_state 0.00007009 0.03891187 layer.4.ssm_state 0.00000002 0.00000496 layer.4.conv_state 0.00055818 0.09204090 layer.4.output 0.00000209 0.00020371 ------------------------------------------------------------------------------------- TOTAL 0.00002970 0.01009354 (elements=1,589,248) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1589248 Total Bytes 728200 BPFP 3.6656 bits/point EBPFP 6.2146 equivalent bits/point MSE 0.010094 ---------------------- -------------------------------------------------------- Time: 4.933s Load: 0.012s, Pack+Encode: 2.590s, Decode+Unpack: 2.331s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 188, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0101 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample166-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample166-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample167-layer4-item1.zst (31/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample167-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 173, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 173, 4096]) -> torch.Size([1, 1, 173, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,304B, BPFP=4.7183 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,076B, BPFP=2.5071 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,440B, BPFP=8.8965 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,812B, BPFP=4.9324 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,100B, BPFP=10.5225 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,812B, BPFP=3.6506 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,436B, BPFP=10.3604 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,080B, BPFP=3.1787 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,348B, BPFP=9.1182 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 217,416B, BPFP=2.4546 ⌛️ [2/4] FRONTEND: Frontend time: 2.589s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 173, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.316s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 173, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000152 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00004078 layer.1.conv_state 0.00049089 0.22781619 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00013154 0.04296365 layer.3.ssm_state 0.00000001 0.00000288 layer.3.conv_state 0.00011542 0.03771482 layer.4.ssm_state 0.00000001 0.00000451 layer.4.conv_state 0.00026976 0.08812531 layer.4.output 0.00000213 0.00019324 ------------------------------------------------------------------------------------- TOTAL 0.00002576 0.01030562 (elements=1,527,808) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1527808 Total Bytes 724968 BPFP 3.7961 bits/point EBPFP 6.4538 equivalent bits/point MSE 0.010306 ---------------------- -------------------------------------------------------- Time: 4.917s Load: 0.011s, Pack+Encode: 2.589s, Decode+Unpack: 2.316s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0103 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample167-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample167-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample173-layer4-item1.zst (32/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample173-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 181, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 181, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 181, 4096]) -> torch.Size([1, 1, 181, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,136B, BPFP=4.7690 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,428B, BPFP=2.5286 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,460B, BPFP=8.9014 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,012B, BPFP=4.9446 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,048B, BPFP=10.5098 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,220B, BPFP=3.7366 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,336B, BPFP=10.3359 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 50,996B, BPFP=3.1125 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,064B, BPFP=9.0488 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 220,572B, BPFP=2.3801 ⌛️ [2/4] FRONTEND: Frontend time: 2.577s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 181, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.335s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 181, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000147 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000011 0.00003991 layer.1.conv_state 0.00048935 0.22746064 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00014754 0.04388778 layer.3.ssm_state 0.00000001 0.00000286 layer.3.conv_state 0.00011438 0.03732447 layer.4.ssm_state 0.00000005 0.00000518 layer.4.conv_state 0.00050179 0.09319333 layer.4.output 0.00000190 0.00018800 ------------------------------------------------------------------------------------- TOTAL 0.00003030 0.01020093 (elements=1,560,576) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1560576 Total Bytes 729416 BPFP 3.7392 bits/point EBPFP 6.3477 equivalent bits/point MSE 0.010201 ---------------------- -------------------------------------------------------- Time: 4.924s Load: 0.012s, Pack+Encode: 2.577s, Decode+Unpack: 2.335s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 181, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0102 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample173-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample173-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample175-layer4-item1.zst (33/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample175-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 209, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 209, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 209, 4096]) -> torch.Size([1, 1, 209, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,928B, BPFP=4.6343 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,096B, BPFP=2.5083 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,552B, BPFP=8.9238 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,136B, BPFP=4.9521 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,256B, BPFP=10.5605 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,244B, BPFP=3.6770 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,572B, BPFP=10.3936 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,292B, BPFP=3.1917 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,280B, BPFP=9.1016 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 270,536B, BPFP=2.5282 ⌛️ [2/4] FRONTEND: Frontend time: 2.630s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 209, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.376s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 209, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000138 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000014 0.00003995 layer.1.conv_state 0.00048877 0.23051406 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00012277 0.04413624 layer.3.ssm_state 0.00000001 0.00000284 layer.3.conv_state 0.00012099 0.03765269 layer.4.ssm_state 0.00000001 0.00000438 layer.4.conv_state 0.00025427 0.08587961 layer.4.output 0.00000161 0.00016763 ------------------------------------------------------------------------------------- TOTAL 0.00002299 0.00943292 (elements=1,675,264) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1675264 Total Bytes 778036 BPFP 3.7154 bits/point EBPFP 6.1389 equivalent bits/point MSE 0.009433 ---------------------- -------------------------------------------------------- Time: 5.016s Load: 0.010s, Pack+Encode: 2.630s, Decode+Unpack: 2.376s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 209, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0094 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample175-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample175-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample179-layer4-item1.zst (34/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample179-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 175, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 175, 4096]) -> torch.Size([1, 1, 175, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,908B, BPFP=4.7551 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,180B, BPFP=2.5134 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,472B, BPFP=8.9043 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,572B, BPFP=4.8567 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,228B, BPFP=10.5537 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,400B, BPFP=3.6865 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,548B, BPFP=10.3877 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,936B, BPFP=3.2310 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,128B, BPFP=9.0645 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 220,100B, BPFP=2.4565 ⌛️ [2/4] FRONTEND: Frontend time: 2.568s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 175, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.330s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 175, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000142 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00003864 layer.1.conv_state 0.00049563 0.22784603 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00014857 0.04403143 layer.3.ssm_state 0.00000001 0.00000292 layer.3.conv_state 0.00007001 0.03797908 layer.4.ssm_state 0.00000002 0.00000431 layer.4.conv_state 0.00020707 0.08238946 layer.4.output 0.00000204 0.00017248 ------------------------------------------------------------------------------------- TOTAL 0.00002375 0.01014848 (elements=1,536,000) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1536000 Total Bytes 728616 BPFP 3.7949 bits/point EBPFP 6.4434 equivalent bits/point MSE 0.010148 ---------------------- -------------------------------------------------------- Time: 4.908s Load: 0.009s, Pack+Encode: 2.568s, Decode+Unpack: 2.330s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0101 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample179-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample179-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample186-layer4-item1.zst (35/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample186-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 172, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 172, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 172, 4096]) -> torch.Size([1, 1, 172, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,380B, BPFP=4.7229 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,224B, BPFP=2.5161 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,460B, BPFP=8.9014 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,696B, BPFP=4.8643 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,228B, BPFP=10.5537 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,592B, BPFP=3.6372 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,472B, BPFP=10.3691 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,676B, BPFP=3.2151 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,240B, BPFP=9.0918 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 223,828B, BPFP=2.5417 ⌛️ [2/4] FRONTEND: Frontend time: 2.694s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 172, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.326s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 172, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000142 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00004107 layer.1.conv_state 0.00050694 0.22892801 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00012346 0.04391697 layer.3.ssm_state 0.00000001 0.00000285 layer.3.conv_state 0.00011101 0.03766736 layer.4.ssm_state 0.00000002 0.00000430 layer.4.conv_state 0.00022697 0.08570992 layer.4.output 0.00000216 0.00019864 ------------------------------------------------------------------------------------- TOTAL 0.00002499 0.01032675 (elements=1,523,712) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1523712 Total Bytes 730940 BPFP 3.8377 bits/point EBPFP 6.5002 equivalent bits/point MSE 0.010327 ---------------------- -------------------------------------------------------- Time: 5.030s Load: 0.010s, Pack+Encode: 2.694s, Decode+Unpack: 2.326s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 172, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0103 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample186-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample186-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample19-layer4-item1.zst (36/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample19-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 193, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 193, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 193, 4096]) -> torch.Size([1, 1, 193, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,952B, BPFP=4.6968 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,092B, BPFP=2.5081 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,496B, BPFP=8.9102 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,184B, BPFP=4.9551 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,168B, BPFP=10.5391 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,640B, BPFP=3.7012 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,532B, BPFP=10.3838 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,432B, BPFP=3.1392 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,084B, BPFP=9.0537 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 246,880B, BPFP=2.4984 ⌛️ [2/4] FRONTEND: Frontend time: 2.631s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 193, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.386s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 193, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000031 0.00003986 layer.1.conv_state 0.00048110 0.22779810 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00012031 0.04354879 layer.3.ssm_state 0.00000001 0.00000296 layer.3.conv_state 0.00007804 0.03805748 layer.4.ssm_state 0.00000001 0.00000478 layer.4.conv_state 0.00027310 0.08968446 layer.4.output 0.00000181 0.00016393 ------------------------------------------------------------------------------------- TOTAL 0.00002328 0.00982681 (elements=1,609,728) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1609728 Total Bytes 754604 BPFP 3.7502 bits/point EBPFP 6.2735 equivalent bits/point MSE 0.009827 ---------------------- -------------------------------------------------------- Time: 5.027s Load: 0.010s, Pack+Encode: 2.631s, Decode+Unpack: 2.386s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 193, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0098 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample19-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample19-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample190-layer4-item1.zst (37/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample190-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 159, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 159, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 159, 4096]) -> torch.Size([1, 1, 159, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,160B, BPFP=4.7705 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,244B, BPFP=2.5173 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,468B, BPFP=8.9033 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,408B, BPFP=4.9077 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,292B, BPFP=10.5693 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,808B, BPFP=3.7114 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,616B, BPFP=10.4043 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,548B, BPFP=3.2073 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,232B, BPFP=9.0898 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 208,068B, BPFP=2.5559 ⌛️ [2/4] FRONTEND: Frontend time: 2.585s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 159, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.794s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 159, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000038 0.00003990 layer.1.conv_state 0.00048847 0.22737601 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00014206 0.04424906 layer.3.ssm_state 0.00000001 0.00000293 layer.3.conv_state 0.00011550 0.03752349 layer.4.ssm_state 0.00000002 0.00000437 layer.4.conv_state 0.00021972 0.08522289 layer.4.output 0.00000203 0.00017610 ------------------------------------------------------------------------------------- TOTAL 0.00002571 0.01064219 (elements=1,470,464) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1470464 Total Bytes 717988 BPFP 3.9062 bits/point EBPFP 6.6804 equivalent bits/point MSE 0.010642 ---------------------- -------------------------------------------------------- Time: 5.387s Load: 0.009s, Pack+Encode: 2.585s, Decode+Unpack: 2.794s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 159, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0106 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample190-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample190-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample2-layer4-item1.zst (38/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample2-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 310, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.013s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 310, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 310, 4096]) -> torch.Size([1, 1, 310, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,592B, BPFP=4.7358 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,008B, BPFP=2.5029 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,508B, BPFP=8.9131 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,692B, BPFP=4.9861 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,272B, BPFP=10.5645 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,076B, BPFP=3.7278 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,748B, BPFP=10.4365 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,676B, BPFP=3.1541 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,428B, BPFP=9.1377 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 351,172B, BPFP=2.2125 ⌛️ [2/4] FRONTEND: Frontend time: 2.682s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 310, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.425s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 310, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000012 0.00004115 layer.1.conv_state 0.00048398 0.22642902 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00013032 0.04376675 layer.3.ssm_state 0.00000001 0.00000292 layer.3.conv_state 0.00007314 0.03744937 layer.4.ssm_state 0.00000002 0.00000479 layer.4.conv_state 0.00041555 0.08998452 layer.4.output 0.00000121 0.00011483 ------------------------------------------------------------------------------------- TOTAL 0.00002034 0.00755737 (elements=2,088,960) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2088960 Total Bytes 861316 BPFP 3.2985 bits/point EBPFP 5.2522 equivalent bits/point MSE 0.007557 ---------------------- -------------------------------------------------------- Time: 5.120s Load: 0.013s, Pack+Encode: 2.682s, Decode+Unpack: 2.425s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 310, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0076 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample2-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample2-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample20-layer4-item1.zst (39/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample20-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 194, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 194, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 194, 4096]) -> torch.Size([1, 1, 194, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,324B, BPFP=4.6584 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,096B, BPFP=2.5083 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,532B, BPFP=8.9189 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,516B, BPFP=4.9143 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,144B, BPFP=10.5332 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,940B, BPFP=3.6584 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,508B, BPFP=10.3779 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,712B, BPFP=3.1562 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,120B, BPFP=9.0625 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 249,528B, BPFP=2.5122 ⌛️ [2/4] FRONTEND: Frontend time: 2.623s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 194, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.381s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 194, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000145 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00004118 layer.1.conv_state 0.00048732 0.22522081 layer.2.ssm_state 0.00000001 0.00000212 layer.2.conv_state 0.00011647 0.04407288 layer.3.ssm_state 0.00000001 0.00000287 layer.3.conv_state 0.00011908 0.03705123 layer.4.ssm_state 0.00000001 0.00000464 layer.4.conv_state 0.00022797 0.08716037 layer.4.output 0.00000187 0.00015361 ------------------------------------------------------------------------------------- TOTAL 0.00002322 0.00968392 (elements=1,613,824) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1613824 Total Bytes 755564 BPFP 3.7455 bits/point EBPFP 6.2540 equivalent bits/point MSE 0.009684 ---------------------- -------------------------------------------------------- Time: 5.014s Load: 0.010s, Pack+Encode: 2.623s, Decode+Unpack: 2.381s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 194, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0097 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample20-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample20-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample204-layer4-item1.zst (40/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample204-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 172, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 172, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 172, 4096]) -> torch.Size([1, 1, 172, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,324B, BPFP=4.7195 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,172B, BPFP=2.5129 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,396B, BPFP=8.8857 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 82,140B, BPFP=5.0134 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,164B, BPFP=10.5381 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,420B, BPFP=3.6877 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,520B, BPFP=10.3809 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,620B, BPFP=3.2117 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,384B, BPFP=9.1270 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 216,812B, BPFP=2.4620 ⌛️ [2/4] FRONTEND: Frontend time: 2.592s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 172, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.305s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 172, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000140 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00004089 layer.1.conv_state 0.00049938 0.22552301 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00013201 0.04357821 layer.3.ssm_state 0.00000001 0.00000294 layer.3.conv_state 0.00007471 0.03818312 layer.4.ssm_state 0.00000001 0.00000465 layer.4.conv_state 0.00021358 0.08880249 layer.4.output 0.00000218 0.00018909 ------------------------------------------------------------------------------------- TOTAL 0.00002396 0.01031944 (elements=1,523,712) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1523712 Total Bytes 727096 BPFP 3.8175 bits/point EBPFP 6.4967 equivalent bits/point MSE 0.010319 ---------------------- -------------------------------------------------------- Time: 4.907s Load: 0.010s, Pack+Encode: 2.592s, Decode+Unpack: 2.305s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 172, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0103 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample204-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample204-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample21-layer4-item1.zst (41/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample21-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 168, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 168, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) -> torch.Size([1, 1, 168, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,816B, BPFP=4.6885 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,436B, BPFP=2.5291 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,524B, BPFP=8.9170 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,556B, BPFP=4.9778 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,292B, BPFP=10.5693 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,508B, BPFP=3.6931 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,484B, BPFP=10.3721 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,448B, BPFP=3.1401 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,196B, BPFP=9.0811 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 220,668B, BPFP=2.5654 ⌛️ [2/4] FRONTEND: Frontend time: 2.568s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.309s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000036 0.00004012 layer.1.conv_state 0.00050661 0.22862025 layer.2.ssm_state 0.00000001 0.00000219 layer.2.conv_state 0.00012774 0.04443391 layer.3.ssm_state 0.00000001 0.00000284 layer.3.conv_state 0.00007574 0.03757535 layer.4.ssm_state 0.00000011 0.00000552 layer.4.conv_state 0.00082926 0.09560402 layer.4.output 0.00000222 0.00017982 ------------------------------------------------------------------------------------- TOTAL 0.00003770 0.01064590 (elements=1,507,328) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1507328 Total Bytes 729072 BPFP 3.8695 bits/point EBPFP 6.5678 equivalent bits/point MSE 0.010646 ---------------------- -------------------------------------------------------- Time: 4.887s Load: 0.009s, Pack+Encode: 2.568s, Decode+Unpack: 2.309s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 168, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0106 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample21-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample21-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample218-layer4-item1.zst (42/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample218-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 160, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 160, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 160, 4096]) -> torch.Size([1, 1, 160, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,108B, BPFP=4.6453 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,124B, BPFP=2.5100 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,564B, BPFP=8.9268 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,288B, BPFP=4.8394 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,320B, BPFP=10.5762 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,624B, BPFP=3.7002 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,556B, BPFP=10.3896 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,880B, BPFP=3.1665 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,192B, BPFP=9.0801 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 209,452B, BPFP=2.5568 ⌛️ [2/4] FRONTEND: Frontend time: 2.574s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 160, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.315s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 160, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000142 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000030 0.00004072 layer.1.conv_state 0.00048992 0.22722471 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00014897 0.04397518 layer.3.ssm_state 0.00000001 0.00000281 layer.3.conv_state 0.00007436 0.03768810 layer.4.ssm_state 0.00000001 0.00000466 layer.4.conv_state 0.00026192 0.08673804 layer.4.output 0.00000216 0.00018831 ------------------------------------------------------------------------------------- TOTAL 0.00002591 0.01064650 (elements=1,474,560) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1474560 Total Bytes 715252 BPFP 3.8805 bits/point EBPFP 6.6246 equivalent bits/point MSE 0.010647 ---------------------- -------------------------------------------------------- Time: 4.897s Load: 0.009s, Pack+Encode: 2.574s, Decode+Unpack: 2.315s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 160, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0106 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample218-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample218-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample22-layer4-item1.zst (43/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample22-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 184, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 184, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 184, 4096]) -> torch.Size([1, 1, 184, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,996B, BPFP=4.6995 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,308B, BPFP=2.5212 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,612B, BPFP=8.9385 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,148B, BPFP=4.9529 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,196B, BPFP=10.5459 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,800B, BPFP=3.6499 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,412B, BPFP=10.3545 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,412B, BPFP=3.1379 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,072B, BPFP=9.0508 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 223,756B, BPFP=2.3751 ⌛️ [2/4] FRONTEND: Frontend time: 2.573s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 184, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.313s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 184, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000139 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000010 0.00003954 layer.1.conv_state 0.00050597 0.22879204 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00012948 0.04443976 layer.3.ssm_state 0.00000001 0.00000276 layer.3.conv_state 0.00007506 0.03727439 layer.4.ssm_state 0.00000001 0.00000472 layer.4.conv_state 0.00035639 0.08721132 layer.4.output 0.00000208 0.00017762 ------------------------------------------------------------------------------------- TOTAL 0.00002628 0.01003122 (elements=1,572,864) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1572864 Total Bytes 730856 BPFP 3.7173 bits/point EBPFP 6.2966 equivalent bits/point MSE 0.010031 ---------------------- -------------------------------------------------------- Time: 4.895s Load: 0.010s, Pack+Encode: 2.573s, Decode+Unpack: 2.313s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 184, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0100 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample22-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample22-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample23-layer4-item1.zst (44/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample23-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 183, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 183, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) -> torch.Size([1, 1, 183, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,216B, BPFP=4.6519 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,272B, BPFP=2.5190 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,536B, BPFP=8.9199 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,352B, BPFP=4.9043 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,352B, BPFP=10.5840 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,544B, BPFP=3.6343 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,460B, BPFP=10.3662 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,156B, BPFP=3.1833 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,112B, BPFP=9.0605 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 226,064B, BPFP=2.4127 ⌛️ [2/4] FRONTEND: Frontend time: 2.594s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.317s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000010 0.00004169 layer.1.conv_state 0.00049328 0.22639455 layer.2.ssm_state 0.00000001 0.00000220 layer.2.conv_state 0.00013895 0.04407898 layer.3.ssm_state 0.00000001 0.00000269 layer.3.conv_state 0.00011983 0.03712728 layer.4.ssm_state 0.00000001 0.00000440 layer.4.conv_state 0.00020516 0.08292428 layer.4.output 0.00000202 0.00016498 ------------------------------------------------------------------------------------- TOTAL 0.00002402 0.00990083 (elements=1,568,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1568768 Total Bytes 732208 BPFP 3.7339 bits/point EBPFP 6.3150 equivalent bits/point MSE 0.009901 ---------------------- -------------------------------------------------------- Time: 4.920s Load: 0.010s, Pack+Encode: 2.594s, Decode+Unpack: 2.317s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 183, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0099 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample23-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample23-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample234-layer4-item1.zst (45/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample234-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 164, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 164, 4096]) -> torch.Size([1, 1, 164, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,036B, BPFP=4.7629 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,232B, BPFP=2.5166 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,464B, BPFP=8.9023 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,416B, BPFP=4.9082 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,184B, BPFP=10.5430 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,260B, BPFP=3.6169 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,544B, BPFP=10.3867 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,580B, BPFP=3.2092 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,428B, BPFP=9.1377 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 213,452B, BPFP=2.5421 ⌛️ [2/4] FRONTEND: Frontend time: 2.568s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 164, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.311s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 164, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000149 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00004178 layer.1.conv_state 0.00051090 0.22598258 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00012582 0.04479197 layer.3.ssm_state 0.00000001 0.00000288 layer.3.conv_state 0.00007785 0.03816269 layer.4.ssm_state 0.00000001 0.00000481 layer.4.conv_state 0.00021500 0.08516472 layer.4.output 0.00000225 0.00018937 ------------------------------------------------------------------------------------- TOTAL 0.00002468 0.01049868 (elements=1,490,944) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1490944 Total Bytes 721740 BPFP 3.8727 bits/point EBPFP 6.6000 equivalent bits/point MSE 0.010499 ---------------------- -------------------------------------------------------- Time: 4.888s Load: 0.009s, Pack+Encode: 2.568s, Decode+Unpack: 2.311s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0105 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample234-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample234-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample24-layer4-item1.zst (46/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample24-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 196, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 196, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 196, 4096]) -> torch.Size([1, 1, 196, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,044B, BPFP=4.7634 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,220B, BPFP=2.5159 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,400B, BPFP=8.8867 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,616B, BPFP=4.9204 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,084B, BPFP=10.5186 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,716B, BPFP=3.6448 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,408B, BPFP=10.3535 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,984B, BPFP=3.1729 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,268B, BPFP=9.0986 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 251,964B, BPFP=2.5108 ⌛️ [2/4] FRONTEND: Frontend time: 2.619s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 196, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.372s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 196, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000147 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004024 layer.1.conv_state 0.00048758 0.22950111 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00011566 0.04308172 layer.3.ssm_state 0.00000001 0.00000281 layer.3.conv_state 0.00011820 0.03705835 layer.4.ssm_state 0.00000001 0.00000438 layer.4.conv_state 0.00025975 0.09000576 layer.4.output 0.00000193 0.00015861 ------------------------------------------------------------------------------------- TOTAL 0.00002375 0.00976224 (elements=1,622,016) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1622016 Total Bytes 759848 BPFP 3.7477 bits/point EBPFP 6.2526 equivalent bits/point MSE 0.009762 ---------------------- -------------------------------------------------------- Time: 5.000s Load: 0.010s, Pack+Encode: 2.619s, Decode+Unpack: 2.372s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 196, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0098 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample24-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample24-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample25-layer4-item1.zst (47/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample25-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 202, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 202, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 202, 4096]) -> torch.Size([1, 1, 202, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,980B, BPFP=4.6375 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,068B, BPFP=2.5066 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,420B, BPFP=8.8916 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,664B, BPFP=4.8623 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,140B, BPFP=10.5322 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,988B, BPFP=3.6614 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,376B, BPFP=10.3457 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,232B, BPFP=3.1880 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,072B, BPFP=9.0508 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 259,396B, BPFP=2.5081 ⌛️ [2/4] FRONTEND: Frontend time: 2.622s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 202, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.371s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 202, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000011 0.00004111 layer.1.conv_state 0.00050439 0.22465843 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00014180 0.04326224 layer.3.ssm_state 0.00000001 0.00000279 layer.3.conv_state 0.00006834 0.03660763 layer.4.ssm_state 0.00000001 0.00000438 layer.4.conv_state 0.00020995 0.08327228 layer.4.output 0.00000183 0.00014551 ------------------------------------------------------------------------------------- TOTAL 0.00002224 0.00937663 (elements=1,646,592) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1646592 Total Bytes 764480 BPFP 3.7142 bits/point EBPFP 6.1682 equivalent bits/point MSE 0.009377 ---------------------- -------------------------------------------------------- Time: 5.004s Load: 0.010s, Pack+Encode: 2.622s, Decode+Unpack: 2.371s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 202, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0094 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample25-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample25-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample26-layer4-item1.zst (48/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample26-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 196, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 196, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 196, 4096]) -> torch.Size([1, 1, 196, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,104B, BPFP=4.7671 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,336B, BPFP=2.5229 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,428B, BPFP=8.8936 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,452B, BPFP=4.9714 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,152B, BPFP=10.5352 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,800B, BPFP=3.7109 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,540B, BPFP=10.3857 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,144B, BPFP=3.1826 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 255,200B, BPFP=2.5430 ⌛️ [2/4] FRONTEND: Frontend time: 2.626s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 196, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.373s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 196, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000147 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00004009 layer.1.conv_state 0.00047314 0.22846755 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00012900 0.04381490 layer.3.ssm_state 0.00000001 0.00000287 layer.3.conv_state 0.00011898 0.03736930 layer.4.ssm_state 0.00000002 0.00000458 layer.4.conv_state 0.00022845 0.08374249 layer.4.output 0.00000185 0.00016064 ------------------------------------------------------------------------------------- TOTAL 0.00002308 0.00963693 (elements=1,622,016) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1622016 Total Bytes 765444 BPFP 3.7753 bits/point EBPFP 6.2919 equivalent bits/point MSE 0.009637 ---------------------- -------------------------------------------------------- Time: 5.009s Load: 0.010s, Pack+Encode: 2.626s, Decode+Unpack: 2.373s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 196, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0096 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample26-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample26-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample266-layer4-item1.zst (49/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample266-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 157, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 157, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 157, 4096]) -> torch.Size([1, 1, 157, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,464B, BPFP=4.7280 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,088B, BPFP=2.5078 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,404B, BPFP=8.8877 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,716B, BPFP=4.9265 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,140B, BPFP=10.5322 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,820B, BPFP=3.6511 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,504B, BPFP=10.3770 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,100B, BPFP=3.2410 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,340B, BPFP=9.1162 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 207,832B, BPFP=2.5855 ⌛️ [2/4] FRONTEND: Frontend time: 2.603s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 157, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.320s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 157, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000146 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000030 0.00004026 layer.1.conv_state 0.00049487 0.22785461 layer.2.ssm_state 0.00000001 0.00000213 layer.2.conv_state 0.00012267 0.04533950 layer.3.ssm_state 0.00000001 0.00000288 layer.3.conv_state 0.00011738 0.03778110 layer.4.ssm_state 0.00000002 0.00000446 layer.4.conv_state 0.00024516 0.08314132 layer.4.output 0.00000212 0.00019726 ------------------------------------------------------------------------------------- TOTAL 0.00002620 0.01070445 (elements=1,462,272) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1462272 Total Bytes 716552 BPFP 3.9202 bits/point EBPFP 6.7034 equivalent bits/point MSE 0.010704 ---------------------- -------------------------------------------------------- Time: 4.932s Load: 0.009s, Pack+Encode: 2.603s, Decode+Unpack: 2.320s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 157, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0107 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample266-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample266-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample27-layer4-item1.zst (50/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample27-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 195, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 195, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 195, 4096]) -> torch.Size([1, 1, 195, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,996B, BPFP=4.7605 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,200B, BPFP=2.5146 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,468B, BPFP=8.9033 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,404B, BPFP=4.9075 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,180B, BPFP=10.5420 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,612B, BPFP=3.6384 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,452B, BPFP=10.3643 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,544B, BPFP=3.1460 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,112B, BPFP=9.0605 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 257,712B, BPFP=2.5812 ⌛️ [2/4] FRONTEND: Frontend time: 2.633s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 195, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.378s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 195, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000149 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000009 0.00004094 layer.1.conv_state 0.00049297 0.22806650 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00013859 0.04427370 layer.3.ssm_state 0.00000001 0.00000285 layer.3.conv_state 0.00007636 0.03722013 layer.4.ssm_state 0.00000001 0.00000455 layer.4.conv_state 0.00030466 0.08391721 layer.4.output 0.00000195 0.00015240 ------------------------------------------------------------------------------------- TOTAL 0.00002444 0.00965861 (elements=1,617,920) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1617920 Total Bytes 764824 BPFP 3.7818 bits/point EBPFP 6.2892 equivalent bits/point MSE 0.009659 ---------------------- -------------------------------------------------------- Time: 5.021s Load: 0.010s, Pack+Encode: 2.633s, Decode+Unpack: 2.378s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 195, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0097 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample27-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample27-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample283-layer4-item1.zst (51/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample283-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 154, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 154, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 154, 4096]) -> torch.Size([1, 1, 154, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,512B, BPFP=4.7310 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,208B, BPFP=2.5151 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,432B, BPFP=8.8945 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,112B, BPFP=4.8896 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,280B, BPFP=10.5664 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,384B, BPFP=3.6855 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,540B, BPFP=10.3857 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,896B, BPFP=3.2285 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,292B, BPFP=9.1045 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 203,228B, BPFP=2.5775 ⌛️ [2/4] FRONTEND: Frontend time: 2.565s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 154, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.328s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 154, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000149 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000039 0.00004023 layer.1.conv_state 0.00049616 0.22688323 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00014030 0.04425211 layer.3.ssm_state 0.00000001 0.00000301 layer.3.conv_state 0.00011494 0.03773572 layer.4.ssm_state 0.00000003 0.00000433 layer.4.conv_state 0.00023159 0.08364121 layer.4.output 0.00000204 0.00018148 ------------------------------------------------------------------------------------- TOTAL 0.00002644 0.01075037 (elements=1,449,984) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1449984 Total Bytes 712028 BPFP 3.9285 bits/point EBPFP 6.7357 equivalent bits/point MSE 0.010750 ---------------------- -------------------------------------------------------- Time: 4.905s Load: 0.012s, Pack+Encode: 2.565s, Decode+Unpack: 2.328s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 154, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0108 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample283-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample283-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample29-layer4-item1.zst (52/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample29-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 186, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 186, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 186, 4096]) -> torch.Size([1, 1, 186, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,888B, BPFP=4.6929 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,240B, BPFP=2.5171 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,576B, BPFP=8.9297 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,596B, BPFP=4.9192 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,228B, BPFP=10.5537 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,000B, BPFP=3.7231 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,492B, BPFP=10.3740 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,984B, BPFP=3.2339 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,228B, BPFP=9.0889 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 222,460B, BPFP=2.3360 ⌛️ [2/4] FRONTEND: Frontend time: 2.604s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 186, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.318s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 186, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000141 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004098 layer.1.conv_state 0.00049578 0.22640179 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00013264 0.04398356 layer.3.ssm_state 0.00000001 0.00000282 layer.3.conv_state 0.00012070 0.03785419 layer.4.ssm_state 0.00000002 0.00000414 layer.4.conv_state 0.00022222 0.08349981 layer.4.output 0.00000189 0.00016806 ------------------------------------------------------------------------------------- TOTAL 0.00002409 0.00985174 (elements=1,581,056) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1581056 Total Bytes 731836 BPFP 3.7030 bits/point EBPFP 6.2804 equivalent bits/point MSE 0.009852 ---------------------- -------------------------------------------------------- Time: 4.932s Load: 0.009s, Pack+Encode: 2.604s, Decode+Unpack: 2.318s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 186, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0099 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample29-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample29-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample3-layer4-item1.zst (53/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample3-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 276, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.015s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 276, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 276, 4096]) -> torch.Size([1, 1, 276, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,340B, BPFP=4.6594 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,276B, BPFP=2.5193 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,472B, BPFP=8.9043 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,500B, BPFP=4.9744 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,056B, BPFP=10.5117 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,864B, BPFP=3.6538 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,424B, BPFP=10.3574 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,444B, BPFP=3.1399 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,164B, BPFP=9.0732 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 329,824B, BPFP=2.3340 ⌛️ [2/4] FRONTEND: Frontend time: 2.718s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 276, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.439s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 276, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000027 0.00004044 layer.1.conv_state 0.00048777 0.22841613 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00011906 0.04413631 layer.3.ssm_state 0.00000001 0.00000284 layer.3.conv_state 0.00011276 0.03700675 layer.4.ssm_state 0.00000006 0.00000519 layer.4.conv_state 0.00059394 0.09335485 layer.4.output 0.00000125 0.00013911 ------------------------------------------------------------------------------------- TOTAL 0.00002528 0.00819185 (elements=1,949,696) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1949696 Total Bytes 836508 BPFP 3.4324 bits/point EBPFP 5.5114 equivalent bits/point MSE 0.008192 ---------------------- -------------------------------------------------------- Time: 5.171s Load: 0.015s, Pack+Encode: 2.718s, Decode+Unpack: 2.439s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 276, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0082 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample3-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample3-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample30-layer4-item1.zst (54/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample30-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 198, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 198, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 198, 4096]) -> torch.Size([1, 1, 198, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,428B, BPFP=4.6648 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,024B, BPFP=2.5039 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,556B, BPFP=8.9248 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,240B, BPFP=4.8975 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,084B, BPFP=10.5186 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,272B, BPFP=3.6787 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,524B, BPFP=10.3818 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,720B, BPFP=3.2788 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,344B, BPFP=9.1172 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 268,568B, BPFP=2.6492 ⌛️ [2/4] FRONTEND: Frontend time: 2.657s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 198, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.384s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 198, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00003977 layer.1.conv_state 0.00048081 0.22993203 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00011915 0.04367100 layer.3.ssm_state 0.00000001 0.00000295 layer.3.conv_state 0.00011359 0.03804930 layer.4.ssm_state 0.00000002 0.00000401 layer.4.conv_state 0.00023406 0.08256783 layer.4.output 0.00000200 0.00016261 ------------------------------------------------------------------------------------- TOTAL 0.00002301 0.00960683 (elements=1,630,208) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1630208 Total Bytes 776904 BPFP 3.8125 bits/point EBPFP 6.3071 equivalent bits/point MSE 0.009607 ---------------------- -------------------------------------------------------- Time: 5.051s Load: 0.010s, Pack+Encode: 2.657s, Decode+Unpack: 2.384s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 198, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0096 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample30-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample30-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample31-layer4-item1.zst (55/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample31-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 175, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 175, 4096]) -> torch.Size([1, 1, 175, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,136B, BPFP=4.6470 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,012B, BPFP=2.5032 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,620B, BPFP=8.9404 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,300B, BPFP=4.9622 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,324B, BPFP=10.5771 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,748B, BPFP=3.6467 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,580B, BPFP=10.3955 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,980B, BPFP=3.1726 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,316B, BPFP=9.1104 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 226,168B, BPFP=2.5242 ⌛️ [2/4] FRONTEND: Frontend time: 2.590s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 175, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.331s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 175, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004083 layer.1.conv_state 0.00049445 0.22724655 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00015634 0.04347799 layer.3.ssm_state 0.00000001 0.00000282 layer.3.conv_state 0.00006751 0.03754856 layer.4.ssm_state 0.00000001 0.00000443 layer.4.conv_state 0.00021972 0.08583909 layer.4.output 0.00000206 0.00017666 ------------------------------------------------------------------------------------- TOTAL 0.00002411 0.01019044 (elements=1,536,000) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1536000 Total Bytes 733328 BPFP 3.8194 bits/point EBPFP 6.4609 equivalent bits/point MSE 0.010190 ---------------------- -------------------------------------------------------- Time: 4.933s Load: 0.011s, Pack+Encode: 2.590s, Decode+Unpack: 2.331s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0102 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample31-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample31-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample32-layer4-item1.zst (56/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample32-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 225, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.013s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 225, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 225, 4096]) -> torch.Size([1, 1, 225, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,128B, BPFP=4.7075 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,212B, BPFP=2.5154 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,468B, BPFP=8.9033 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,564B, BPFP=4.9172 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,076B, BPFP=10.5166 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,496B, BPFP=3.7534 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,468B, BPFP=10.3682 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,380B, BPFP=3.1970 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,208B, BPFP=9.0840 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 289,960B, BPFP=2.5170 ⌛️ [2/4] FRONTEND: Frontend time: 2.659s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 225, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.395s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 225, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000146 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000027 0.00003906 layer.1.conv_state 0.00049817 0.22676533 layer.2.ssm_state 0.00000001 0.00000221 layer.2.conv_state 0.00012605 0.04449047 layer.3.ssm_state 0.00000001 0.00000293 layer.3.conv_state 0.00007245 0.03821414 layer.4.ssm_state 0.00000001 0.00000455 layer.4.conv_state 0.00021821 0.08429745 layer.4.output 0.00000156 0.00015014 ------------------------------------------------------------------------------------- TOTAL 0.00002082 0.00899170 (elements=1,740,800) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1740800 Total Bytes 799104 BPFP 3.6724 bits/point EBPFP 6.0122 equivalent bits/point MSE 0.008992 ---------------------- -------------------------------------------------------- Time: 5.067s Load: 0.013s, Pack+Encode: 2.659s, Decode+Unpack: 2.395s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 225, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0090 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample32-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample32-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample323-layer4-item1.zst (57/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample323-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 152, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 152, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 152, 4096]) -> torch.Size([1, 1, 152, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,048B, BPFP=4.7026 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,252B, BPFP=2.5178 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,444B, BPFP=8.8975 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 82,124B, BPFP=5.0125 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,132B, BPFP=10.5303 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,712B, BPFP=3.6445 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,556B, BPFP=10.3896 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,560B, BPFP=3.2080 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,288B, BPFP=9.1035 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 208,904B, BPFP=2.6843 ⌛️ [2/4] FRONTEND: Frontend time: 2.577s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 152, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.325s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 152, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000139 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00004082 layer.1.conv_state 0.00048212 0.22730787 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00013258 0.04419329 layer.3.ssm_state 0.00000001 0.00000292 layer.3.conv_state 0.00007465 0.03717476 layer.4.ssm_state 0.00000001 0.00000440 layer.4.conv_state 0.00020440 0.08643479 layer.4.output 0.00000217 0.00020280 ------------------------------------------------------------------------------------- TOTAL 0.00002460 0.01087873 (elements=1,441,792) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1441792 Total Bytes 718164 BPFP 3.9848 bits/point EBPFP 6.8105 equivalent bits/point MSE 0.010879 ---------------------- -------------------------------------------------------- Time: 4.910s Load: 0.009s, Pack+Encode: 2.577s, Decode+Unpack: 2.325s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 152, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0109 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample323-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample323-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample33-layer4-item1.zst (58/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample33-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 205, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.013s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 205, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 205, 4096]) -> torch.Size([1, 1, 205, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,872B, BPFP=4.6919 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,188B, BPFP=2.5139 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,396B, BPFP=8.8857 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,776B, BPFP=4.9912 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,016B, BPFP=10.5020 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,780B, BPFP=3.6487 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,460B, BPFP=10.3662 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,924B, BPFP=3.2302 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,288B, BPFP=9.1035 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 260,484B, BPFP=2.4817 ⌛️ [2/4] FRONTEND: Frontend time: 2.645s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 205, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.395s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 205, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000150 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000024 0.00003967 layer.1.conv_state 0.00049079 0.22743879 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00014720 0.04334477 layer.3.ssm_state 0.00000001 0.00000281 layer.3.conv_state 0.00011227 0.03697424 layer.4.ssm_state 0.00000001 0.00000451 layer.4.conv_state 0.00021844 0.09094729 layer.4.output 0.00000170 0.00016959 ------------------------------------------------------------------------------------- TOTAL 0.00002290 0.00953574 (elements=1,658,880) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1658880 Total Bytes 769328 BPFP 3.7101 bits/point EBPFP 6.1640 equivalent bits/point MSE 0.009536 ---------------------- -------------------------------------------------------- Time: 5.053s Load: 0.013s, Pack+Encode: 2.645s, Decode+Unpack: 2.395s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 205, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0095 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample33-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample33-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample34-layer4-item1.zst (59/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample34-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 223, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.013s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 223, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 223, 4096]) -> torch.Size([1, 1, 223, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,660B, BPFP=4.7400 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 40,936B, BPFP=2.4985 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,436B, BPFP=8.8955 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,844B, BPFP=4.9343 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,132B, BPFP=10.5303 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,312B, BPFP=3.7422 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,564B, BPFP=10.3916 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,616B, BPFP=3.1504 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,368B, BPFP=9.1230 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 271,992B, BPFP=2.3822 ⌛️ [2/4] FRONTEND: Frontend time: 2.655s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 223, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.394s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 223, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00003880 layer.1.conv_state 0.00048351 0.22422057 layer.2.ssm_state 0.00000001 0.00000210 layer.2.conv_state 0.00011966 0.04377229 layer.3.ssm_state 0.00000001 0.00000291 layer.3.conv_state 0.00006977 0.03721070 layer.4.ssm_state 0.00000002 0.00000506 layer.4.conv_state 0.00033313 0.08951908 layer.4.output 0.00000142 0.00013470 ------------------------------------------------------------------------------------- TOTAL 0.00002256 0.00904345 (elements=1,732,608) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1732608 Total Bytes 781004 BPFP 3.6061 bits/point EBPFP 5.9564 equivalent bits/point MSE 0.009043 ---------------------- -------------------------------------------------------- Time: 5.062s Load: 0.013s, Pack+Encode: 2.655s, Decode+Unpack: 2.394s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 223, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0090 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample34-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample34-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample35-layer4-item1.zst (60/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample35-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 183, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 183, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) -> torch.Size([1, 1, 183, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,136B, BPFP=4.7690 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,236B, BPFP=2.5168 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,408B, BPFP=8.8887 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,096B, BPFP=4.8887 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,188B, BPFP=10.5439 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,724B, BPFP=3.7063 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,564B, BPFP=10.3916 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,340B, BPFP=3.1946 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,216B, BPFP=9.0859 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 218,136B, BPFP=2.3281 ⌛️ [2/4] FRONTEND: Frontend time: 2.600s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.326s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000149 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000021 0.00004016 layer.1.conv_state 0.00049749 0.22774586 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00012539 0.04390357 layer.3.ssm_state 0.00000001 0.00000279 layer.3.conv_state 0.00007456 0.03813072 layer.4.ssm_state 0.00000001 0.00000482 layer.4.conv_state 0.00032004 0.09341760 layer.4.output 0.00000203 0.00016096 ------------------------------------------------------------------------------------- TOTAL 0.00002530 0.01016352 (elements=1,568,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1568768 Total Bytes 727188 BPFP 3.7083 bits/point EBPFP 6.3043 equivalent bits/point MSE 0.010164 ---------------------- -------------------------------------------------------- Time: 4.938s Load: 0.011s, Pack+Encode: 2.600s, Decode+Unpack: 2.326s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 183, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0102 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample35-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample35-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample36-layer4-item1.zst (61/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample36-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 181, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 181, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 181, 4096]) -> torch.Size([1, 1, 181, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,144B, BPFP=4.7085 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,200B, BPFP=2.5146 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,492B, BPFP=8.9092 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 82,516B, BPFP=5.0364 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,180B, BPFP=10.5420 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,496B, BPFP=3.6924 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,684B, BPFP=10.4209 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,016B, BPFP=3.1748 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,416B, BPFP=9.1348 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 225,264B, BPFP=2.4308 ⌛️ [2/4] FRONTEND: Frontend time: 2.615s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 181, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.332s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 181, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000145 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000013 0.00004048 layer.1.conv_state 0.00048788 0.22559199 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00012276 0.04464302 layer.3.ssm_state 0.00000001 0.00000294 layer.3.conv_state 0.00007521 0.03778061 layer.4.ssm_state 0.00000001 0.00000480 layer.4.conv_state 0.00025615 0.08873568 layer.4.output 0.00000200 0.00019439 ------------------------------------------------------------------------------------- TOTAL 0.00002381 0.01009659 (elements=1,560,576) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1560576 Total Bytes 735552 BPFP 3.7707 bits/point EBPFP 6.3866 equivalent bits/point MSE 0.010097 ---------------------- -------------------------------------------------------- Time: 4.955s Load: 0.009s, Pack+Encode: 2.615s, Decode+Unpack: 2.332s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 181, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0101 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample36-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample36-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample37-layer4-item1.zst (62/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample37-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 204, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 204, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 204, 4096]) -> torch.Size([1, 1, 204, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,092B, BPFP=4.7053 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,144B, BPFP=2.5112 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,536B, BPFP=8.9199 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,024B, BPFP=4.9453 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,204B, BPFP=10.5479 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,772B, BPFP=3.7092 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,516B, BPFP=10.3799 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,428B, BPFP=3.1389 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,184B, BPFP=9.0781 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 270,040B, BPFP=2.5854 ⌛️ [2/4] FRONTEND: Frontend time: 2.654s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 204, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.391s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 204, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00003781 layer.1.conv_state 0.00048573 0.22822013 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00014000 0.04335044 layer.3.ssm_state 0.00000001 0.00000293 layer.3.conv_state 0.00007101 0.03784892 layer.4.ssm_state 0.00000002 0.00000508 layer.4.conv_state 0.00054851 0.09149114 layer.4.output 0.00000173 0.00018564 ------------------------------------------------------------------------------------- TOTAL 0.00002845 0.00961060 (elements=1,654,784) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1654784 Total Bytes 778084 BPFP 3.7616 bits/point EBPFP 6.2177 equivalent bits/point MSE 0.009611 ---------------------- -------------------------------------------------------- Time: 5.055s Load: 0.010s, Pack+Encode: 2.654s, Decode+Unpack: 2.391s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 204, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0096 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample37-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample37-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample38-layer4-item1.zst (63/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample38-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 189, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 189, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 189, 4096]) -> torch.Size([1, 1, 189, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,836B, BPFP=4.8118 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,240B, BPFP=2.5171 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,424B, BPFP=8.8926 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,916B, BPFP=4.9387 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,160B, BPFP=10.5371 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,232B, BPFP=3.6763 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,488B, BPFP=10.3730 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,356B, BPFP=3.1956 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,248B, BPFP=9.0938 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 218,028B, BPFP=2.2531 ⌛️ [2/4] FRONTEND: Frontend time: 2.602s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 189, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.330s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 189, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000151 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000013 0.00004101 layer.1.conv_state 0.00048943 0.22409496 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00011993 0.04371023 layer.3.ssm_state 0.00000001 0.00000288 layer.3.conv_state 0.00011634 0.03772257 layer.4.ssm_state 0.00000001 0.00000443 layer.4.conv_state 0.00025516 0.08601999 layer.4.output 0.00000196 0.00020034 ------------------------------------------------------------------------------------- TOTAL 0.00002414 0.00978884 (elements=1,593,344) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1593344 Total Bytes 728072 BPFP 3.6556 bits/point EBPFP 6.2164 equivalent bits/point MSE 0.009789 ---------------------- -------------------------------------------------------- Time: 4.941s Load: 0.009s, Pack+Encode: 2.602s, Decode+Unpack: 2.330s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 189, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0098 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample38-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample38-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample39-layer4-item1.zst (64/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample39-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 188, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 188, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 188, 4096]) -> torch.Size([1, 1, 188, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,264B, BPFP=4.7158 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,208B, BPFP=2.5151 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,536B, BPFP=8.9199 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,812B, BPFP=4.8713 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,144B, BPFP=10.5332 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,328B, BPFP=3.6821 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,524B, BPFP=10.3818 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,032B, BPFP=3.2368 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,384B, BPFP=9.1270 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 223,424B, BPFP=2.3211 ⌛️ [2/4] FRONTEND: Frontend time: 2.604s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 188, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.332s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 188, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000145 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000027 0.00004147 layer.1.conv_state 0.00048897 0.22995514 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00011975 0.04344210 layer.3.ssm_state 0.00000001 0.00000279 layer.3.conv_state 0.00006970 0.03794195 layer.4.ssm_state 0.00000001 0.00000406 layer.4.conv_state 0.00027957 0.08319315 layer.4.output 0.00000196 0.00019622 ------------------------------------------------------------------------------------- TOTAL 0.00002374 0.00987309 (elements=1,589,248) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1589248 Total Bytes 731800 BPFP 3.6838 bits/point EBPFP 6.2428 equivalent bits/point MSE 0.009873 ---------------------- -------------------------------------------------------- Time: 4.945s Load: 0.009s, Pack+Encode: 2.604s, Decode+Unpack: 2.332s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 188, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0099 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample39-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample39-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample4-layer4-item1.zst (65/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample4-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 271, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 271, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 271, 4096]) -> torch.Size([1, 1, 271, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,632B, BPFP=4.7383 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,320B, BPFP=2.5220 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,520B, BPFP=8.9160 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,388B, BPFP=4.9065 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,152B, BPFP=10.5352 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,612B, BPFP=3.7605 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,504B, BPFP=10.3770 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,444B, BPFP=3.2009 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,140B, BPFP=9.0674 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 340,228B, BPFP=2.4521 ⌛️ [2/4] FRONTEND: Frontend time: 2.712s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 271, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.437s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 271, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00003953 layer.1.conv_state 0.00049489 0.22902480 layer.2.ssm_state 0.00000001 0.00000211 layer.2.conv_state 0.00013525 0.04441726 layer.3.ssm_state 0.00000001 0.00000289 layer.3.conv_state 0.00011705 0.03763536 layer.4.ssm_state 0.00000001 0.00000459 layer.4.conv_state 0.00022610 0.08576079 layer.4.output 0.00000138 0.00012811 ------------------------------------------------------------------------------------- TOTAL 0.00001983 0.00816771 (elements=1,929,216) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1929216 Total Bytes 850084 BPFP 3.5251 bits/point EBPFP 5.6393 equivalent bits/point MSE 0.008168 ---------------------- -------------------------------------------------------- Time: 5.162s Load: 0.012s, Pack+Encode: 2.712s, Decode+Unpack: 2.437s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 271, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0082 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample4-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample4-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample40-layer4-item1.zst (66/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample40-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 168, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 168, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) -> torch.Size([1, 1, 168, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,888B, BPFP=4.6929 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,272B, BPFP=2.5190 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,464B, BPFP=8.9023 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,392B, BPFP=4.9678 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,196B, BPFP=10.5459 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,120B, BPFP=3.7305 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,488B, BPFP=10.3730 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,952B, BPFP=3.2319 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,204B, BPFP=9.0830 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 224,348B, BPFP=2.6082 ⌛️ [2/4] FRONTEND: Frontend time: 2.588s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.325s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 168, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000142 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000047 0.00004176 layer.1.conv_state 0.00050445 0.22535512 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00012625 0.04427775 layer.3.ssm_state 0.00000001 0.00000277 layer.3.conv_state 0.00011525 0.03819063 layer.4.ssm_state 0.00000001 0.00000439 layer.4.conv_state 0.00022432 0.08364874 layer.4.output 0.00000216 0.00019226 ------------------------------------------------------------------------------------- TOTAL 0.00002530 0.01033072 (elements=1,507,328) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1507328 Total Bytes 734468 BPFP 3.8981 bits/point EBPFP 6.6055 equivalent bits/point MSE 0.010331 ---------------------- -------------------------------------------------------- Time: 4.921s Load: 0.009s, Pack+Encode: 2.588s, Decode+Unpack: 2.325s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 168, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0103 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample40-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample40-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample403-layer4-item1.zst (67/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample403-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 173, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 173, 4096]) -> torch.Size([1, 1, 173, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,368B, BPFP=4.6611 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,332B, BPFP=2.5227 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,420B, BPFP=8.8916 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,636B, BPFP=4.8606 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,112B, BPFP=10.5254 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,420B, BPFP=3.6877 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,556B, BPFP=10.3896 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,452B, BPFP=3.1404 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,352B, BPFP=9.1191 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 219,116B, BPFP=2.4738 ⌛️ [2/4] FRONTEND: Frontend time: 2.604s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 173, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.337s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 173, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000139 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000013 0.00004083 layer.1.conv_state 0.00048463 0.22765255 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00012632 0.04458564 layer.3.ssm_state 0.00000001 0.00000299 layer.3.conv_state 0.00007282 0.03790047 layer.4.ssm_state 0.00000003 0.00000521 layer.4.conv_state 0.00058008 0.09577672 layer.4.output 0.00000205 0.00019780 ------------------------------------------------------------------------------------- TOTAL 0.00003121 0.01050716 (elements=1,527,808) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1527808 Total Bytes 724908 BPFP 3.7958 bits/point EBPFP 6.4443 equivalent bits/point MSE 0.010507 ---------------------- -------------------------------------------------------- Time: 4.949s Load: 0.009s, Pack+Encode: 2.604s, Decode+Unpack: 2.337s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0105 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample403-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample403-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample41-layer4-item1.zst (68/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample41-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 197, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 197, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 197, 4096]) -> torch.Size([1, 1, 197, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,200B, BPFP=4.7119 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,244B, BPFP=2.5173 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,432B, BPFP=8.8945 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,872B, BPFP=4.9360 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,248B, BPFP=10.5586 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,408B, BPFP=3.6870 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,644B, BPFP=10.4111 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,364B, BPFP=3.1960 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,344B, BPFP=9.1172 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 255,636B, BPFP=2.5345 ⌛️ [2/4] FRONTEND: Frontend time: 2.651s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 197, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.392s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 197, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000141 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00003918 layer.1.conv_state 0.00050234 0.22616954 layer.2.ssm_state 0.00000001 0.00000213 layer.2.conv_state 0.00013726 0.04450329 layer.3.ssm_state 0.00000001 0.00000283 layer.3.conv_state 0.00007731 0.03819679 layer.4.ssm_state 0.00000001 0.00000480 layer.4.conv_state 0.00032633 0.09144931 layer.4.output 0.00000190 0.00016251 ------------------------------------------------------------------------------------- TOTAL 0.00002493 0.00975346 (elements=1,626,112) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1626112 Total Bytes 764536 BPFP 3.7613 bits/point EBPFP 6.2649 equivalent bits/point MSE 0.009753 ---------------------- -------------------------------------------------------- Time: 5.053s Load: 0.010s, Pack+Encode: 2.651s, Decode+Unpack: 2.392s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 197, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0098 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample41-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample41-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample42-layer4-item1.zst (69/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample42-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 185, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 185, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 185, 4096]) -> torch.Size([1, 1, 185, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,156B, BPFP=4.7092 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,160B, BPFP=2.5122 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,512B, BPFP=8.9141 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,704B, BPFP=4.9868 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,260B, BPFP=10.5615 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,148B, BPFP=3.6711 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,572B, BPFP=10.3936 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,084B, BPFP=3.1790 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,252B, BPFP=9.0947 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 223,360B, BPFP=2.3581 ⌛️ [2/4] FRONTEND: Frontend time: 2.585s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 185, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.330s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 185, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000145 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000010 0.00004085 layer.1.conv_state 0.00049834 0.22847128 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00012928 0.04390279 layer.3.ssm_state 0.00000001 0.00000272 layer.3.conv_state 0.00007819 0.03755574 layer.4.ssm_state 0.00000001 0.00000464 layer.4.conv_state 0.00030927 0.08722344 layer.4.output 0.00000194 0.00017530 ------------------------------------------------------------------------------------- TOTAL 0.00002507 0.00999289 (elements=1,576,960) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1576960 Total Bytes 732352 BPFP 3.7153 bits/point EBPFP 6.2974 equivalent bits/point MSE 0.009993 ---------------------- -------------------------------------------------------- Time: 4.925s Load: 0.010s, Pack+Encode: 2.585s, Decode+Unpack: 2.330s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 185, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0100 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample42-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample42-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample44-layer4-item1.zst (70/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample44-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 210, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 210, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 210, 4096]) -> torch.Size([1, 1, 210, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,252B, BPFP=4.7761 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,236B, BPFP=2.5168 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,456B, BPFP=8.9004 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,564B, BPFP=4.9783 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,192B, BPFP=10.5449 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,076B, BPFP=3.6667 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,532B, BPFP=10.3838 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,888B, BPFP=3.2280 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,356B, BPFP=9.1201 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 271,000B, BPFP=2.5205 ⌛️ [2/4] FRONTEND: Frontend time: 2.661s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 210, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.393s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 210, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000149 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004133 layer.1.conv_state 0.00049741 0.22620322 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00015034 0.04424143 layer.3.ssm_state 0.00000001 0.00000291 layer.3.conv_state 0.00007531 0.03774595 layer.4.ssm_state 0.00000004 0.00000460 layer.4.conv_state 0.00020482 0.08704112 layer.4.output 0.00000169 0.00015909 ------------------------------------------------------------------------------------- TOTAL 0.00002184 0.00934851 (elements=1,679,360) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1679360 Total Bytes 781696 BPFP 3.7238 bits/point EBPFP 6.1566 equivalent bits/point MSE 0.009349 ---------------------- -------------------------------------------------------- Time: 5.064s Load: 0.010s, Pack+Encode: 2.661s, Decode+Unpack: 2.393s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 210, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0093 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample44-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample44-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample45-layer4-item1.zst (71/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample45-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 184, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 184, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 184, 4096]) -> torch.Size([1, 1, 184, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,940B, BPFP=4.6350 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,324B, BPFP=2.5222 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,516B, BPFP=8.9150 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,344B, BPFP=4.9038 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,240B, BPFP=10.5566 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,924B, BPFP=3.6575 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,560B, BPFP=10.3906 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,220B, BPFP=3.1873 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,296B, BPFP=9.1055 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 224,644B, BPFP=2.3846 ⌛️ [2/4] FRONTEND: Frontend time: 2.596s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 184, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.319s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 184, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000141 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000012 0.00003911 layer.1.conv_state 0.00051907 0.22628322 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00017046 0.04421381 layer.3.ssm_state 0.00000001 0.00000297 layer.3.conv_state 0.00006998 0.03860381 layer.4.ssm_state 0.00000001 0.00000458 layer.4.conv_state 0.00024001 0.09001102 layer.4.output 0.00000196 0.00018041 ------------------------------------------------------------------------------------- TOTAL 0.00002482 0.01006158 (elements=1,572,864) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1572864 Total Bytes 731152 BPFP 3.7188 bits/point EBPFP 6.2951 equivalent bits/point MSE 0.010062 ---------------------- -------------------------------------------------------- Time: 4.924s Load: 0.009s, Pack+Encode: 2.596s, Decode+Unpack: 2.319s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 184, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0101 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample45-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample45-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample46-layer4-item1.zst (72/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample46-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 196, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 196, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 196, 4096]) -> torch.Size([1, 1, 196, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,172B, BPFP=4.6492 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,284B, BPFP=2.5198 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,544B, BPFP=8.9219 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,448B, BPFP=4.9102 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,224B, BPFP=10.5527 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,176B, BPFP=3.6729 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,508B, BPFP=10.3779 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,868B, BPFP=3.1658 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,156B, BPFP=9.0713 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 254,356B, BPFP=2.5346 ⌛️ [2/4] FRONTEND: Frontend time: 2.639s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 196, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.382s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 196, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004011 layer.1.conv_state 0.00047987 0.22706713 layer.2.ssm_state 0.00000001 0.00000212 layer.2.conv_state 0.00014269 0.04362765 layer.3.ssm_state 0.00000001 0.00000285 layer.3.conv_state 0.00007044 0.03762862 layer.4.ssm_state 0.00000001 0.00000460 layer.4.conv_state 0.00031684 0.08547744 layer.4.output 0.00000179 0.00016309 ------------------------------------------------------------------------------------- TOTAL 0.00002426 0.00964635 (elements=1,622,016) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1622016 Total Bytes 760880 BPFP 3.7528 bits/point EBPFP 6.2510 equivalent bits/point MSE 0.009646 ---------------------- -------------------------------------------------------- Time: 5.030s Load: 0.009s, Pack+Encode: 2.639s, Decode+Unpack: 2.382s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 196, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0096 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample46-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample46-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample47-layer4-item1.zst (73/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample47-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 203, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 203, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 203, 4096]) -> torch.Size([1, 1, 203, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,336B, BPFP=4.5981 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,040B, BPFP=2.5049 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,492B, BPFP=8.9092 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,548B, BPFP=4.8552 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,268B, BPFP=10.5635 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,256B, BPFP=3.6777 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,652B, BPFP=10.4131 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,396B, BPFP=3.2590 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,384B, BPFP=9.1270 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 262,416B, BPFP=2.5248 ⌛️ [2/4] FRONTEND: Frontend time: 2.638s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 203, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.377s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 203, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000140 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000043 0.00004006 layer.1.conv_state 0.00048147 0.22534545 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00012906 0.04443418 layer.3.ssm_state 0.00000001 0.00000296 layer.3.conv_state 0.00011463 0.03769767 layer.4.ssm_state 0.00000002 0.00000437 layer.4.conv_state 0.00022714 0.08782594 layer.4.output 0.00000188 0.00014777 ------------------------------------------------------------------------------------- TOTAL 0.00002279 0.00950373 (elements=1,650,688) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1650688 Total Bytes 768932 BPFP 3.7266 bits/point EBPFP 6.1814 equivalent bits/point MSE 0.009504 ---------------------- -------------------------------------------------------- Time: 5.025s Load: 0.010s, Pack+Encode: 2.638s, Decode+Unpack: 2.377s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 203, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0095 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample47-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample47-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample48-layer4-item1.zst (74/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample48-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 184, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 184, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 184, 4096]) -> torch.Size([1, 1, 184, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,376B, BPFP=4.7227 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,104B, BPFP=2.5088 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,472B, BPFP=8.9043 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,092B, BPFP=4.9495 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,288B, BPFP=10.5684 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,084B, BPFP=3.6672 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,620B, BPFP=10.4053 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,888B, BPFP=3.1670 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,248B, BPFP=9.0938 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 221,916B, BPFP=2.3556 ⌛️ [2/4] FRONTEND: Frontend time: 2.576s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 184, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.324s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 184, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00004033 layer.1.conv_state 0.00050109 0.22723258 layer.2.ssm_state 0.00000001 0.00000219 layer.2.conv_state 0.00013610 0.04422471 layer.3.ssm_state 0.00000001 0.00000295 layer.3.conv_state 0.00007718 0.03768325 layer.4.ssm_state 0.00000001 0.00000447 layer.4.conv_state 0.00025147 0.08603761 layer.4.output 0.00000198 0.00017576 ------------------------------------------------------------------------------------- TOTAL 0.00002414 0.00997749 (elements=1,572,864) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1572864 Total Bytes 730232 BPFP 3.7142 bits/point EBPFP 6.2996 equivalent bits/point MSE 0.009977 ---------------------- -------------------------------------------------------- Time: 4.909s Load: 0.009s, Pack+Encode: 2.576s, Decode+Unpack: 2.324s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 184, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0100 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample48-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample48-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample51-layer4-item1.zst (75/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample51-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 192, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 192, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 192, 4096]) -> torch.Size([1, 1, 192, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,816B, BPFP=4.6885 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,232B, BPFP=2.5166 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,488B, BPFP=8.9082 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 82,684B, BPFP=5.0466 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,104B, BPFP=10.5234 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,060B, BPFP=3.6658 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,328B, BPFP=10.3340 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,120B, BPFP=3.2422 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,044B, BPFP=9.0439 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 218,460B, BPFP=2.2223 ⌛️ [2/4] FRONTEND: Frontend time: 2.594s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 192, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.323s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 192, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000024 0.00004123 layer.1.conv_state 0.00048279 0.22546071 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00015517 0.04324983 layer.3.ssm_state 0.00000001 0.00000284 layer.3.conv_state 0.00011498 0.03716317 layer.4.ssm_state 0.00000001 0.00000398 layer.4.conv_state 0.00025157 0.07927671 layer.4.output 0.00000201 0.00017087 ------------------------------------------------------------------------------------- TOTAL 0.00002449 0.00957044 (elements=1,605,632) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1605632 Total Bytes 728480 BPFP 3.6296 bits/point EBPFP 6.1708 equivalent bits/point MSE 0.009570 ---------------------- -------------------------------------------------------- Time: 4.926s Load: 0.009s, Pack+Encode: 2.594s, Decode+Unpack: 2.323s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 192, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0096 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample51-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample51-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample52-layer4-item1.zst (76/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample52-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 170, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 170, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 170, 4096]) -> torch.Size([1, 1, 170, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,768B, BPFP=4.6855 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,288B, BPFP=2.5200 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,464B, BPFP=8.9023 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,800B, BPFP=4.9316 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,144B, BPFP=10.5332 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,840B, BPFP=3.6523 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,564B, BPFP=10.3916 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,392B, BPFP=3.1978 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,260B, BPFP=9.0967 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 224,224B, BPFP=2.5761 ⌛️ [2/4] FRONTEND: Frontend time: 2.575s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 170, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.321s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 170, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000148 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000011 0.00004035 layer.1.conv_state 0.00049192 0.22408612 layer.2.ssm_state 0.00000001 0.00000220 layer.2.conv_state 0.00013328 0.04338669 layer.3.ssm_state 0.00000001 0.00000291 layer.3.conv_state 0.00007075 0.03749568 layer.4.ssm_state 0.00000001 0.00000475 layer.4.conv_state 0.00027001 0.08834796 layer.4.output 0.00000204 0.00017541 ------------------------------------------------------------------------------------- TOTAL 0.00002499 0.01030798 (elements=1,515,520) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1515520 Total Bytes 731888 BPFP 3.8634 bits/point EBPFP 6.5432 equivalent bits/point MSE 0.010308 ---------------------- -------------------------------------------------------- Time: 4.905s Load: 0.009s, Pack+Encode: 2.575s, Decode+Unpack: 2.321s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 170, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0103 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample52-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample52-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample53-layer4-item1.zst (77/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample53-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 200, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.013s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 200, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 200, 4096]) -> torch.Size([1, 1, 200, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,168B, BPFP=4.7100 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,224B, BPFP=2.5161 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,396B, BPFP=8.8857 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,548B, BPFP=4.9773 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,108B, BPFP=10.5244 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,956B, BPFP=3.6594 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,452B, BPFP=10.3643 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,044B, BPFP=3.2375 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,212B, BPFP=9.0850 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 252,548B, BPFP=2.4663 ⌛️ [2/4] FRONTEND: Frontend time: 2.639s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 200, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.379s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 200, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000043 0.00004020 layer.1.conv_state 0.00049042 0.22529593 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00012605 0.04341227 layer.3.ssm_state 0.00000001 0.00000284 layer.3.conv_state 0.00011250 0.03792118 layer.4.ssm_state 0.00000002 0.00000426 layer.4.conv_state 0.00024020 0.08696467 layer.4.output 0.00000182 0.00015843 ------------------------------------------------------------------------------------- TOTAL 0.00002325 0.00954503 (elements=1,638,400) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1638400 Total Bytes 761800 BPFP 3.7197 bits/point EBPFP 6.2063 equivalent bits/point MSE 0.009545 ---------------------- -------------------------------------------------------- Time: 5.031s Load: 0.013s, Pack+Encode: 2.639s, Decode+Unpack: 2.379s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 200, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0095 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample53-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample53-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample55-layer4-item1.zst (78/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample55-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 199, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 199, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 199, 4096]) -> torch.Size([1, 1, 199, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,308B, BPFP=4.6575 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,136B, BPFP=2.5107 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,516B, BPFP=8.9150 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,768B, BPFP=4.8687 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,240B, BPFP=10.5566 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,992B, BPFP=3.6616 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,692B, BPFP=10.4229 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,048B, BPFP=3.2378 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,356B, BPFP=9.1201 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 260,852B, BPFP=2.5602 ⌛️ [2/4] FRONTEND: Frontend time: 2.646s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 199, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.379s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 199, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000043 0.00004033 layer.1.conv_state 0.00048601 0.22152790 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00014442 0.04450006 layer.3.ssm_state 0.00000001 0.00000295 layer.3.conv_state 0.00011344 0.03820748 layer.4.ssm_state 0.00000006 0.00000435 layer.4.conv_state 0.00022676 0.08612607 layer.4.output 0.00000196 0.00015232 ------------------------------------------------------------------------------------- TOTAL 0.00002341 0.00950073 (elements=1,634,304) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1634304 Total Bytes 768052 BPFP 3.7597 bits/point EBPFP 6.2424 equivalent bits/point MSE 0.009501 ---------------------- -------------------------------------------------------- Time: 5.035s Load: 0.009s, Pack+Encode: 2.646s, Decode+Unpack: 2.379s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 199, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0095 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample55-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample55-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample56-layer4-item1.zst (79/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample56-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 215, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 215, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 215, 4096]) -> torch.Size([1, 1, 215, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,248B, BPFP=4.7148 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,132B, BPFP=2.5105 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,540B, BPFP=8.9209 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,552B, BPFP=4.8555 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,284B, BPFP=10.5674 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,412B, BPFP=3.6262 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,656B, BPFP=10.4141 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,500B, BPFP=3.2043 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,400B, BPFP=9.1309 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 279,540B, BPFP=2.5394 ⌛️ [2/4] FRONTEND: Frontend time: 2.626s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 215, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.374s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 215, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000015 0.00004156 layer.1.conv_state 0.00047793 0.22454397 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00012432 0.04499680 layer.3.ssm_state 0.00000001 0.00000277 layer.3.conv_state 0.00007747 0.03863625 layer.4.ssm_state 0.00000001 0.00000456 layer.4.conv_state 0.00032616 0.08383306 layer.4.output 0.00000162 0.00014901 ------------------------------------------------------------------------------------- TOTAL 0.00002306 0.00917047 (elements=1,699,840) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1699840 Total Bytes 786408 BPFP 3.7011 bits/point EBPFP 6.0866 equivalent bits/point MSE 0.009170 ---------------------- -------------------------------------------------------- Time: 5.011s Load: 0.011s, Pack+Encode: 2.626s, Decode+Unpack: 2.374s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 215, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0092 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample56-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample56-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample58-layer4-item1.zst (80/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample58-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 174, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 174, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 174, 4096]) -> torch.Size([1, 1, 174, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,624B, BPFP=4.7988 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,268B, BPFP=2.5188 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,388B, BPFP=8.8838 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,940B, BPFP=4.9402 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,144B, BPFP=10.5332 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,100B, BPFP=3.6682 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,388B, BPFP=10.3486 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,516B, BPFP=3.2053 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,108B, BPFP=9.0596 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 220,504B, BPFP=2.4751 ⌛️ [2/4] FRONTEND: Frontend time: 2.572s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 174, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.319s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 174, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000145 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00003950 layer.1.conv_state 0.00048286 0.22868082 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00013493 0.04372944 layer.3.ssm_state 0.00000001 0.00000281 layer.3.conv_state 0.00011535 0.03758086 layer.4.ssm_state 0.00000001 0.00000447 layer.4.conv_state 0.00022609 0.09130012 layer.4.output 0.00000231 0.00017084 ------------------------------------------------------------------------------------- TOTAL 0.00002474 0.01036796 (elements=1,531,904) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1531904 Total Bytes 730124 BPFP 3.8129 bits/point EBPFP 6.4743 equivalent bits/point MSE 0.010368 ---------------------- -------------------------------------------------------- Time: 4.900s Load: 0.009s, Pack+Encode: 2.572s, Decode+Unpack: 2.319s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 174, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0104 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample58-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample58-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample63-layer4-item1.zst (81/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample63-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 186, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 186, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 186, 4096]) -> torch.Size([1, 1, 186, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,804B, BPFP=4.6877 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,196B, BPFP=2.5144 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,500B, BPFP=8.9111 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,528B, BPFP=4.8540 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,112B, BPFP=10.5254 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,244B, BPFP=3.6770 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,580B, BPFP=10.3955 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,208B, BPFP=3.2476 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,304B, BPFP=9.1074 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 224,452B, BPFP=2.3569 ⌛️ [2/4] FRONTEND: Frontend time: 2.579s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 186, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.321s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 186, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000142 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000027 0.00003979 layer.1.conv_state 0.00049416 0.22652258 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00014839 0.04429742 layer.3.ssm_state 0.00000001 0.00000289 layer.3.conv_state 0.00011359 0.03798821 layer.4.ssm_state 0.00000002 0.00000434 layer.4.conv_state 0.00020754 0.08615682 layer.4.output 0.00000201 0.00016908 ------------------------------------------------------------------------------------- TOTAL 0.00002400 0.00991900 (elements=1,581,056) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1581056 Total Bytes 732072 BPFP 3.7042 bits/point EBPFP 6.2727 equivalent bits/point MSE 0.009919 ---------------------- -------------------------------------------------------- Time: 4.909s Load: 0.009s, Pack+Encode: 2.579s, Decode+Unpack: 2.321s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 186, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0099 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample63-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample63-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample64-layer4-item1.zst (82/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample64-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 185, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 185, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 185, 4096]) -> torch.Size([1, 1, 185, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,036B, BPFP=4.7629 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,456B, BPFP=2.5303 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,436B, BPFP=8.8955 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,728B, BPFP=4.9272 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,104B, BPFP=10.5234 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,836B, BPFP=3.7131 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,448B, BPFP=10.3633 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,424B, BPFP=3.1997 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,048B, BPFP=9.0449 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 222,524B, BPFP=2.3493 ⌛️ [2/4] FRONTEND: Frontend time: 2.608s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 185, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.323s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 185, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000024 0.00003905 layer.1.conv_state 0.00048011 0.22650023 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00013394 0.04334843 layer.3.ssm_state 0.00000001 0.00000289 layer.3.conv_state 0.00011242 0.03738276 layer.4.ssm_state 0.00000002 0.00000438 layer.4.conv_state 0.00023441 0.08367810 layer.4.output 0.00000189 0.00017519 ------------------------------------------------------------------------------------- TOTAL 0.00002393 0.00986294 (elements=1,576,960) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1576960 Total Bytes 732184 BPFP 3.7144 bits/point EBPFP 6.2999 equivalent bits/point MSE 0.009863 ---------------------- -------------------------------------------------------- Time: 4.941s Load: 0.010s, Pack+Encode: 2.608s, Decode+Unpack: 2.323s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 185, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0099 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample64-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample64-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample67-layer4-item1.zst (83/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample67-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 179, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 179, 4096]) -> torch.Size([1, 1, 179, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,396B, BPFP=4.6628 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,244B, BPFP=2.5173 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,472B, BPFP=8.9043 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,556B, BPFP=4.9167 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,160B, BPFP=10.5371 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,400B, BPFP=3.6255 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,332B, BPFP=10.3350 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,528B, BPFP=3.2061 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,076B, BPFP=9.0518 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 222,420B, BPFP=2.4269 ⌛️ [2/4] FRONTEND: Frontend time: 2.585s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 179, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.328s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 179, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000147 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000030 0.00003947 layer.1.conv_state 0.00050588 0.22782342 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00011591 0.04430960 layer.3.ssm_state 0.00000001 0.00000275 layer.3.conv_state 0.00011742 0.03726034 layer.4.ssm_state 0.00000003 0.00000427 layer.4.conv_state 0.00023463 0.08362837 layer.4.output 0.00000192 0.00016882 ------------------------------------------------------------------------------------- TOTAL 0.00002458 0.01005790 (elements=1,552,384) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1552384 Total Bytes 728728 BPFP 3.7554 bits/point EBPFP 6.3646 equivalent bits/point MSE 0.010058 ---------------------- -------------------------------------------------------- Time: 4.922s Load: 0.010s, Pack+Encode: 2.585s, Decode+Unpack: 2.328s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0101 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample67-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample67-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample68-layer4-item1.zst (84/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample68-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 190, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 190, 4096]) -> torch.Size([1, 1, 190, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,116B, BPFP=4.7068 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,308B, BPFP=2.5212 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,404B, BPFP=8.8877 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,792B, BPFP=4.9922 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,152B, BPFP=10.5352 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,504B, BPFP=3.6318 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,468B, BPFP=10.3682 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,192B, BPFP=3.2466 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,272B, BPFP=9.0996 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 212,724B, BPFP=2.1867 ⌛️ [2/4] FRONTEND: Frontend time: 2.587s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 190, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.316s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 190, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000148 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000044 0.00004008 layer.1.conv_state 0.00050610 0.22741947 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00013374 0.04369329 layer.3.ssm_state 0.00000001 0.00000276 layer.3.conv_state 0.00010957 0.03758869 layer.4.ssm_state 0.00000005 0.00000437 layer.4.conv_state 0.00022355 0.08436361 layer.4.output 0.00000189 0.00020864 ------------------------------------------------------------------------------------- TOTAL 0.00002392 0.00979933 (elements=1,597,440) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1597440 Total Bytes 722076 BPFP 3.6162 bits/point EBPFP 6.1670 equivalent bits/point MSE 0.009799 ---------------------- -------------------------------------------------------- Time: 4.912s Load: 0.009s, Pack+Encode: 2.587s, Decode+Unpack: 2.316s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0098 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample68-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample68-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample69-layer4-item1.zst (85/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample69-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 170, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 170, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 170, 4096]) -> torch.Size([1, 1, 170, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,084B, BPFP=4.7048 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,216B, BPFP=2.5156 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,452B, BPFP=8.8994 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,624B, BPFP=4.9209 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,092B, BPFP=10.5205 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,568B, BPFP=3.6357 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,304B, BPFP=10.3281 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,324B, BPFP=3.1936 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,032B, BPFP=9.0410 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 218,300B, BPFP=2.5080 ⌛️ [2/4] FRONTEND: Frontend time: 2.588s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 170, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.316s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 170, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000146 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00003973 layer.1.conv_state 0.00047344 0.22926183 layer.2.ssm_state 0.00000001 0.00000213 layer.2.conv_state 0.00014015 0.04352498 layer.3.ssm_state 0.00000001 0.00000287 layer.3.conv_state 0.00011599 0.03757144 layer.4.ssm_state 0.00000001 0.00000400 layer.4.conv_state 0.00022458 0.08192018 layer.4.output 0.00000213 0.00016722 ------------------------------------------------------------------------------------- TOTAL 0.00002480 0.01028164 (elements=1,515,520) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1515520 Total Bytes 725140 BPFP 3.8278 bits/point EBPFP 6.5033 equivalent bits/point MSE 0.010282 ---------------------- -------------------------------------------------------- Time: 4.914s Load: 0.009s, Pack+Encode: 2.588s, Decode+Unpack: 2.316s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 170, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0103 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample69-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample69-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample7-layer4-item1.zst (86/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample7-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 235, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 235, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 235, 4096]) -> torch.Size([1, 1, 235, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,392B, BPFP=4.7236 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,072B, BPFP=2.5068 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,460B, BPFP=8.9014 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,844B, BPFP=4.9343 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,248B, BPFP=10.5586 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,260B, BPFP=3.6780 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,688B, BPFP=10.4219 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,340B, BPFP=3.1946 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,536B, BPFP=9.1641 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 287,016B, BPFP=2.3854 ⌛️ [2/4] FRONTEND: Frontend time: 2.647s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 235, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.375s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 235, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000149 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000010 0.00004199 layer.1.conv_state 0.00050392 0.22413532 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00010721 0.04426058 layer.3.ssm_state 0.00000001 0.00000289 layer.3.conv_state 0.00007772 0.03808349 layer.4.ssm_state 0.00000001 0.00000462 layer.4.conv_state 0.00021031 0.08756986 layer.4.output 0.00000154 0.00013149 ------------------------------------------------------------------------------------- TOTAL 0.00002006 0.00878378 (elements=1,781,760) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1781760 Total Bytes 796000 BPFP 3.5740 bits/point EBPFP 5.8593 equivalent bits/point MSE 0.008784 ---------------------- -------------------------------------------------------- Time: 5.034s Load: 0.011s, Pack+Encode: 2.647s, Decode+Unpack: 2.375s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 235, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0088 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample7-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample7-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample72-layer4-item1.zst (87/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample72-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 176, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 176, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 176, 4096]) -> torch.Size([1, 1, 176, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 78,332B, BPFP=4.7810 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 40,996B, BPFP=2.5022 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,508B, BPFP=8.9131 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,584B, BPFP=4.9185 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,104B, BPFP=10.5234 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,796B, BPFP=3.6497 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,312B, BPFP=10.3301 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,976B, BPFP=3.1724 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 36,900B, BPFP=9.0088 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 218,136B, BPFP=2.4207 ⌛️ [2/4] FRONTEND: Frontend time: 2.574s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 176, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.316s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 176, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000141 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00004022 layer.1.conv_state 0.00049605 0.22535799 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00014004 0.04331839 layer.3.ssm_state 0.00000001 0.00000278 layer.3.conv_state 0.00011040 0.03651322 layer.4.ssm_state 0.00000002 0.00000483 layer.4.conv_state 0.00070183 0.10162898 layer.4.output 0.00000198 0.00019105 ------------------------------------------------------------------------------------- TOTAL 0.00003488 0.01044087 (elements=1,540,096) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1540096 Total Bytes 725788 BPFP 3.7701 bits/point EBPFP 6.4071 equivalent bits/point MSE 0.010441 ---------------------- -------------------------------------------------------- Time: 4.900s Load: 0.009s, Pack+Encode: 2.574s, Decode+Unpack: 2.316s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 176, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0104 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample72-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample72-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample74-layer4-item1.zst (88/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample74-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 172, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 172, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 172, 4096]) -> torch.Size([1, 1, 172, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,804B, BPFP=4.7488 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,104B, BPFP=2.5088 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,416B, BPFP=8.8906 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,496B, BPFP=4.9741 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,144B, BPFP=10.5332 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,604B, BPFP=3.6379 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,480B, BPFP=10.3711 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,800B, BPFP=3.1616 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,180B, BPFP=9.0771 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 222,472B, BPFP=2.5263 ⌛️ [2/4] FRONTEND: Frontend time: 2.613s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 172, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.331s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 172, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000010 0.00004009 layer.1.conv_state 0.00050574 0.22740231 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00012521 0.04351188 layer.3.ssm_state 0.00000001 0.00000287 layer.3.conv_state 0.00011811 0.03728618 layer.4.ssm_state 0.00000001 0.00000471 layer.4.conv_state 0.00032558 0.09125681 layer.4.output 0.00000215 0.00019407 ------------------------------------------------------------------------------------- TOTAL 0.00002726 0.01039415 (elements=1,523,712) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1523712 Total Bytes 730644 BPFP 3.8361 bits/point EBPFP 6.5042 equivalent bits/point MSE 0.010394 ---------------------- -------------------------------------------------------- Time: 4.952s Load: 0.009s, Pack+Encode: 2.613s, Decode+Unpack: 2.331s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 172, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0104 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample74-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample74-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample75-layer4-item1.zst (89/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample75-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 164, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 164, 4096]) -> torch.Size([1, 1, 164, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,532B, BPFP=4.6711 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,316B, BPFP=2.5217 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,520B, BPFP=8.9160 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,960B, BPFP=4.9414 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,232B, BPFP=10.5547 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,464B, BPFP=3.6904 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,640B, BPFP=10.4102 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,164B, BPFP=3.1838 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,192B, BPFP=9.0801 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 221,368B, BPFP=2.6363 ⌛️ [2/4] FRONTEND: Frontend time: 2.575s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 164, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.319s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 164, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000025 0.00003910 layer.1.conv_state 0.00049170 0.22781828 layer.2.ssm_state 0.00000001 0.00000214 layer.2.conv_state 0.00014440 0.04466504 layer.3.ssm_state 0.00000001 0.00000280 layer.3.conv_state 0.00007111 0.03759450 layer.4.ssm_state 0.00000001 0.00000475 layer.4.conv_state 0.00032601 0.09246288 layer.4.output 0.00000215 0.00019511 ------------------------------------------------------------------------------------- TOTAL 0.00002691 0.01068648 (elements=1,490,944) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1490944 Total Bytes 729532 BPFP 3.9145 bits/point EBPFP 6.6411 equivalent bits/point MSE 0.010686 ---------------------- -------------------------------------------------------- Time: 4.904s Load: 0.010s, Pack+Encode: 2.575s, Decode+Unpack: 2.319s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0107 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample75-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample75-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample76-layer4-item1.zst (90/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample76-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 166, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) -> torch.Size([1, 1, 166, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,472B, BPFP=4.6675 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,140B, BPFP=2.5110 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,468B, BPFP=8.9033 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,348B, BPFP=4.9041 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,124B, BPFP=10.5283 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,932B, BPFP=3.6580 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,440B, BPFP=10.3613 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,556B, BPFP=3.2078 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,256B, BPFP=9.0957 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 219,480B, BPFP=2.5824 ⌛️ [2/4] FRONTEND: Frontend time: 2.571s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.315s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 166, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000146 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000031 0.00004068 layer.1.conv_state 0.00048773 0.22614072 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00012531 0.04404576 layer.3.ssm_state 0.00000001 0.00000283 layer.3.conv_state 0.00012205 0.03743552 layer.4.ssm_state 0.00000002 0.00000419 layer.4.conv_state 0.00022309 0.08185752 layer.4.output 0.00000212 0.00016855 ------------------------------------------------------------------------------------- TOTAL 0.00002513 0.01033171 (elements=1,499,136) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1499136 Total Bytes 726360 BPFP 3.8762 bits/point EBPFP 6.5811 equivalent bits/point MSE 0.010332 ---------------------- -------------------------------------------------------- Time: 4.897s Load: 0.012s, Pack+Encode: 2.571s, Decode+Unpack: 2.315s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0103 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample76-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample76-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample78-layer4-item1.zst (91/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample78-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 180, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 180, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 180, 4096]) -> torch.Size([1, 1, 180, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,748B, BPFP=4.6233 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,036B, BPFP=2.5046 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,604B, BPFP=8.9365 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,024B, BPFP=4.8843 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,284B, BPFP=10.5674 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 61,312B, BPFP=3.7422 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,668B, BPFP=10.4170 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,944B, BPFP=3.2314 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,304B, BPFP=9.1074 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 234,544B, BPFP=2.5450 ⌛️ [2/4] FRONTEND: Frontend time: 2.592s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 180, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.321s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 180, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00004139 layer.1.conv_state 0.00049117 0.22687498 layer.2.ssm_state 0.00000001 0.00000216 layer.2.conv_state 0.00012889 0.04383874 layer.3.ssm_state 0.00000001 0.00000298 layer.3.conv_state 0.00006806 0.03828743 layer.4.ssm_state 0.00000002 0.00000420 layer.4.conv_state 0.00022608 0.08422835 layer.4.output 0.00000211 0.00020721 ------------------------------------------------------------------------------------- TOTAL 0.00002335 0.01005461 (elements=1,556,480) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1556480 Total Bytes 742612 BPFP 3.8169 bits/point EBPFP 6.4282 equivalent bits/point MSE 0.010055 ---------------------- -------------------------------------------------------- Time: 4.922s Load: 0.009s, Pack+Encode: 2.592s, Decode+Unpack: 2.321s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 180, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0101 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample78-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample78-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample80-layer4-item1.zst (92/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample80-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 180, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 180, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 180, 4096]) -> torch.Size([1, 1, 180, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,960B, BPFP=4.6973 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,020B, BPFP=2.5037 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,508B, BPFP=8.9131 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,924B, BPFP=5.0002 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,000B, BPFP=10.4980 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,904B, BPFP=3.6562 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,352B, BPFP=10.3398 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,572B, BPFP=3.2087 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,180B, BPFP=9.0771 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 217,268B, BPFP=2.3575 ⌛️ [2/4] FRONTEND: Frontend time: 2.591s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 180, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.318s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 180, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000146 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000056 0.00004173 layer.1.conv_state 0.00050875 0.22601297 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00009667 0.04417023 layer.3.ssm_state 0.00000001 0.00000280 layer.3.conv_state 0.00011205 0.03758808 layer.4.ssm_state 0.00000001 0.00000431 layer.4.conv_state 0.00021440 0.08738910 layer.4.output 0.00000203 0.00018617 ------------------------------------------------------------------------------------- TOTAL 0.00002371 0.01008532 (elements=1,556,480) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1556480 Total Bytes 725832 BPFP 3.7306 bits/point EBPFP 6.3446 equivalent bits/point MSE 0.010085 ---------------------- -------------------------------------------------------- Time: 4.920s Load: 0.010s, Pack+Encode: 2.591s, Decode+Unpack: 2.318s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 180, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0101 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample80-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample80-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample82-layer4-item1.zst (93/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample82-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 174, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 174, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 174, 4096]) -> torch.Size([1, 1, 174, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,980B, BPFP=4.6985 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,040B, BPFP=2.5049 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,576B, BPFP=8.9297 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,668B, BPFP=4.9236 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,264B, BPFP=10.5625 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,048B, BPFP=3.6650 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,616B, BPFP=10.4043 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,832B, BPFP=3.1636 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,412B, BPFP=9.1338 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 220,760B, BPFP=2.4780 ⌛️ [2/4] FRONTEND: Frontend time: 2.568s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 174, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.316s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 174, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000147 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000011 0.00004156 layer.1.conv_state 0.00049114 0.22722003 layer.2.ssm_state 0.00000001 0.00000215 layer.2.conv_state 0.00011528 0.04422193 layer.3.ssm_state 0.00000001 0.00000295 layer.3.conv_state 0.00007296 0.03777396 layer.4.ssm_state 0.00000001 0.00000481 layer.4.conv_state 0.00026103 0.09120341 layer.4.output 0.00000214 0.00017133 ------------------------------------------------------------------------------------- TOTAL 0.00002425 0.01034975 (elements=1,531,904) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1531904 Total Bytes 728340 BPFP 3.8036 bits/point EBPFP 6.4543 equivalent bits/point MSE 0.010350 ---------------------- -------------------------------------------------------- Time: 4.893s Load: 0.009s, Pack+Encode: 2.568s, Decode+Unpack: 2.316s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 174, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0103 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample82-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample82-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample85-layer4-item1.zst (94/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample85-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 180, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 180, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 180, 4096]) -> torch.Size([1, 1, 180, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,792B, BPFP=4.6870 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,044B, BPFP=2.5051 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,536B, BPFP=8.9199 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 77,272B, BPFP=4.7163 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,256B, BPFP=10.5605 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 58,868B, BPFP=3.5930 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,652B, BPFP=10.4131 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,684B, BPFP=3.2766 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,492B, BPFP=9.1533 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 225,216B, BPFP=2.4438 ⌛️ [2/4] FRONTEND: Frontend time: 2.574s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 180, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.318s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 180, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000149 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000037 0.00004291 layer.1.conv_state 0.00049028 0.22649167 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00010997 0.04367571 layer.3.ssm_state 0.00000001 0.00000287 layer.3.conv_state 0.00007961 0.03825304 layer.4.ssm_state 0.00000002 0.00000393 layer.4.conv_state 0.00024454 0.08526511 layer.4.output 0.00000217 0.00019111 ------------------------------------------------------------------------------------- TOTAL 0.00002360 0.01005668 (elements=1,556,480) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1556480 Total Bytes 729956 BPFP 3.7518 bits/point EBPFP 6.3461 equivalent bits/point MSE 0.010057 ---------------------- -------------------------------------------------------- Time: 4.901s Load: 0.009s, Pack+Encode: 2.574s, Decode+Unpack: 2.318s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 180, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0101 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample85-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample85-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample87-layer4-item1.zst (95/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample87-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 179, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 179, 4096]) -> torch.Size([1, 1, 179, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,636B, BPFP=4.6165 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,024B, BPFP=2.5039 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,504B, BPFP=8.9121 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,156B, BPFP=4.8923 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,148B, BPFP=10.5342 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,364B, BPFP=3.6843 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,540B, BPFP=10.3857 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,924B, BPFP=3.1692 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,200B, BPFP=9.0820 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 221,104B, BPFP=2.4125 ⌛️ [2/4] FRONTEND: Frontend time: 2.577s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 179, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.314s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 179, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000144 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000028 0.00004087 layer.1.conv_state 0.00049519 0.22948354 layer.2.ssm_state 0.00000001 0.00000217 layer.2.conv_state 0.00011503 0.04351832 layer.3.ssm_state 0.00000001 0.00000290 layer.3.conv_state 0.00006908 0.03747980 layer.4.ssm_state 0.00000001 0.00000456 layer.4.conv_state 0.00022394 0.09060962 layer.4.output 0.00000194 0.00016590 ------------------------------------------------------------------------------------- TOTAL 0.00002309 0.01022701 (elements=1,552,384) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1552384 Total Bytes 726744 BPFP 3.7452 bits/point EBPFP 6.3509 equivalent bits/point MSE 0.010227 ---------------------- -------------------------------------------------------- Time: 4.900s Load: 0.009s, Pack+Encode: 2.577s, Decode+Unpack: 2.314s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0102 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample87-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample87-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample9-layer4-item1.zst (96/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample9-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 205, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 205, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 205, 4096]) -> torch.Size([1, 1, 205, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,080B, BPFP=4.6436 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,248B, BPFP=2.5176 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,540B, BPFP=8.9209 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,320B, BPFP=4.9634 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,256B, BPFP=10.5605 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,644B, BPFP=3.6404 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,548B, BPFP=10.3877 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,092B, BPFP=3.1794 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,228B, BPFP=9.0889 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 273,636B, BPFP=2.6071 ⌛️ [2/4] FRONTEND: Frontend time: 2.651s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 205, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.388s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 205, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000010 0.00003985 layer.1.conv_state 0.00050539 0.22558838 layer.2.ssm_state 0.00000001 0.00000219 layer.2.conv_state 0.00015098 0.04398042 layer.3.ssm_state 0.00000001 0.00000272 layer.3.conv_state 0.00007050 0.03775948 layer.4.ssm_state 0.00000001 0.00000446 layer.4.conv_state 0.00024328 0.08788322 layer.4.output 0.00000184 0.00018853 ------------------------------------------------------------------------------------- TOTAL 0.00002299 0.00947632 (elements=1,658,880) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1658880 Total Bytes 780736 BPFP 3.7651 bits/point EBPFP 6.2106 equivalent bits/point MSE 0.009476 ---------------------- -------------------------------------------------------- Time: 5.048s Load: 0.010s, Pack+Encode: 2.651s, Decode+Unpack: 2.388s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 205, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0095 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample9-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample9-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample93-layer4-item1.zst (97/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample93-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 174, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 174, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 174, 4096]) -> torch.Size([1, 1, 174, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,916B, BPFP=4.6335 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,272B, BPFP=2.5190 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,460B, BPFP=8.9014 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,440B, BPFP=4.8486 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,240B, BPFP=10.5566 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,704B, BPFP=3.6440 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,516B, BPFP=10.3799 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 51,924B, BPFP=3.1692 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,212B, BPFP=9.0850 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 222,436B, BPFP=2.4968 ⌛️ [2/4] FRONTEND: Frontend time: 2.599s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 174, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.328s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 174, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000145 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000012 0.00003862 layer.1.conv_state 0.00049634 0.22672346 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00012038 0.04407907 layer.3.ssm_state 0.00000001 0.00000292 layer.3.conv_state 0.00007294 0.03757007 layer.4.ssm_state 0.00000001 0.00000473 layer.4.conv_state 0.00026211 0.08591612 layer.4.output 0.00000210 0.00017307 ------------------------------------------------------------------------------------- TOTAL 0.00002447 0.01021917 (elements=1,531,904) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1531904 Total Bytes 727264 BPFP 3.7980 bits/point EBPFP 6.4343 equivalent bits/point MSE 0.010219 ---------------------- -------------------------------------------------------- Time: 4.937s Load: 0.009s, Pack+Encode: 2.599s, Decode+Unpack: 2.328s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 174, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0102 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample93-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample93-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample94-layer4-item1.zst (98/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample94-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 190, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 190, 4096]) -> torch.Size([1, 1, 190, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 76,552B, BPFP=4.6724 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 40,920B, BPFP=2.4976 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,492B, BPFP=8.9092 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 79,312B, BPFP=4.8408 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,112B, BPFP=10.5254 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 59,320B, BPFP=3.6206 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,172B, BPFP=10.2959 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,704B, BPFP=3.2778 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,080B, BPFP=9.0527 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 213,812B, BPFP=2.1979 ⌛️ [2/4] FRONTEND: Frontend time: 2.589s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 190, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.325s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 190, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000026 0.00003950 layer.1.conv_state 0.00050538 0.22772810 layer.2.ssm_state 0.00000001 0.00000210 layer.2.conv_state 0.00014003 0.04367258 layer.3.ssm_state 0.00000001 0.00000284 layer.3.conv_state 0.00010410 0.03700999 layer.4.ssm_state 0.00000002 0.00000400 layer.4.conv_state 0.00021663 0.08237289 layer.4.output 0.00000194 0.00020652 ------------------------------------------------------------------------------------- TOTAL 0.00002379 0.00975141 (elements=1,597,440) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1597440 Total Bytes 719620 BPFP 3.6039 bits/point EBPFP 6.1370 equivalent bits/point MSE 0.009751 ---------------------- -------------------------------------------------------- Time: 4.923s Load: 0.010s, Pack+Encode: 2.589s, Decode+Unpack: 2.325s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0098 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample94-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample94-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample95-layer4-item1.zst (99/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample95-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 183, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 183, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) -> torch.Size([1, 1, 183, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 75,292B, BPFP=4.5955 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,068B, BPFP=2.5066 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,452B, BPFP=8.8994 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 80,596B, BPFP=4.9192 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,252B, BPFP=10.5596 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,216B, BPFP=3.6753 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,596B, BPFP=10.3994 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 52,356B, BPFP=3.1956 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,216B, BPFP=9.0859 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 219,228B, BPFP=2.3398 ⌛️ [2/4] FRONTEND: Frontend time: 2.584s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.328s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 183, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000143 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000041 0.00004106 layer.1.conv_state 0.00050614 0.22573777 layer.2.ssm_state 0.00000001 0.00000218 layer.2.conv_state 0.00012733 0.04445400 layer.3.ssm_state 0.00000001 0.00000293 layer.3.conv_state 0.00007038 0.03736031 layer.4.ssm_state 0.00000001 0.00000451 layer.4.conv_state 0.00022687 0.08538562 layer.4.output 0.00000194 0.00015914 ------------------------------------------------------------------------------------- TOTAL 0.00002346 0.00994841 (elements=1,568,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1568768 Total Bytes 725416 BPFP 3.6993 bits/point EBPFP 6.2806 equivalent bits/point MSE 0.009948 ---------------------- -------------------------------------------------------- Time: 4.922s Load: 0.009s, Pack+Encode: 2.584s, Decode+Unpack: 2.328s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 183, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0099 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample95-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample95-layer4-item1.zst 💪 Processing: ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample99-layer4-item1.zst (100/100) [1/4] FRONTEND: Loading features from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample99-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 191, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ FalconMamba Features Summary ------------------------------------------------------------ Number of layers: 5 SSM state shape: (1, 8192, 16) Conv state shape: (1, 8192, 4) Output shape: (1, 191, 4096) Data type: torch.float32 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) -> torch.Size([1, 1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) -> torch.Size([1, 1, 8192, 4]) layer.4.output: torch.Size([1, 191, 4096]) -> torch.Size([1, 1, 191, 4096]) Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state layer.0.ssm_state: 77,308B, BPFP=4.7185 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 37,144B, BPFP=9.0684 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 41,296B, BPFP=2.5205 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 36,372B, BPFP=8.8799 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 81,412B, BPFP=4.9690 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 43,112B, BPFP=10.5254 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 60,040B, BPFP=3.6646 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 42,452B, BPFP=10.3643 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 53,248B, BPFP=3.2500 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 37,212B, BPFP=9.0850 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 210,148B, BPFP=2.1489 ⌛️ [2/4] FRONTEND: Frontend time: 2.579s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 191, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.333s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.ssm_state: torch.Size([256])) for layer.0.ssm_state Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.ssm_state: torch.Size([1, 8192, 16]) layer.0.conv_state: torch.Size([1, 8192, 4]) layer.1.ssm_state: torch.Size([1, 8192, 16]) layer.1.conv_state: torch.Size([1, 8192, 4]) layer.2.ssm_state: torch.Size([1, 8192, 16]) layer.2.conv_state: torch.Size([1, 8192, 4]) layer.3.ssm_state: torch.Size([1, 8192, 16]) layer.3.conv_state: torch.Size([1, 8192, 4]) layer.4.ssm_state: torch.Size([1, 8192, 16]) layer.4.conv_state: torch.Size([1, 8192, 4]) layer.4.output: torch.Size([1, 191, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000148 layer.0.conv_state 0.00014612 0.07949343 layer.1.ssm_state 0.00000040 0.00004039 layer.1.conv_state 0.00049989 0.22801819 layer.2.ssm_state 0.00000001 0.00000213 layer.2.conv_state 0.00013469 0.04461259 layer.3.ssm_state 0.00000001 0.00000279 layer.3.conv_state 0.00011145 0.03803531 layer.4.ssm_state 0.00000005 0.00000430 layer.4.conv_state 0.00022056 0.08647152 layer.4.output 0.00000184 0.00018797 ------------------------------------------------------------------------------------- TOTAL 0.00002370 0.00984805 (elements=1,601,536) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1601536 Total Bytes 719744 BPFP 3.5953 bits/point EBPFP 6.1408 equivalent bits/point MSE 0.009848 ---------------------- -------------------------------------------------------- Time: 4.922s Load: 0.010s, Pack+Encode: 2.579s, Decode+Unpack: 2.333s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 191, 4096]), dtype=torch.float32, device=cpu key['ssm_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 16]), dtype=torch.bfloat16, device=cpu key['conv_state']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8192, 4]), dtype=torch.bfloat16, device=cpu 💾 Converting with 0.0098 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample99-layer4-item1.zst to output-fixed/falconmamba/lambda0.02/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample99-layer4-item1.zst ------------------------ ---------------------------- TOTAL PROCESSING SUMMARY ------------------------ ---------------------------- Total files 100 Avg BPFP 3.7558 bits/point Avg EBPFP 6.3246 equivalent bits/point Avg MSE 0.009929 Avg Time 4.972s ------------------------ ----------------------------