Experiment: dtufc_elic-featurecoding_falconmamba_individual Log file: output-fixed/falconmamba/lambda0.004/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_lambda0.004_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_lambda0.004_epochs600_lr0.0001_bs60_patch256-256_checkpoint_best.pth.tar Checkpoint epoch: 255 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_lambda0.004_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/lambda0.004/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: 21,732B, BPFP=1.3264 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,812B, BPFP=0.4768 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,848B, BPFP=2.6484 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,956B, BPFP=1.2791 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,220B, BPFP=2.9834 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,744B, BPFP=0.9609 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,940B, BPFP=2.9150 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,476B, BPFP=0.7004 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,428B, BPFP=2.5459 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 142,120B, BPFP=1.0094 ⌛️ [2/4] FRONTEND: Frontend time: 3.097s (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: 2.440s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000022 0.00012998 layer.1.conv_state 0.00048975 0.84709102 layer.2.ssm_state 0.00000001 0.00000885 layer.2.conv_state 0.00012844 0.17149240 layer.3.ssm_state 0.00000001 0.00001074 layer.3.conv_state 0.00007646 0.16027279 layer.4.ssm_state 0.00000001 0.00001514 layer.4.conv_state 0.00024491 0.33880082 layer.4.output 0.00000149 0.00058721 ------------------------------------------------------------------------------------- TOTAL 0.00001917 0.03086537 (elements=1,945,600) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1945600 Total Bytes 275044 BPFP 1.1309 bits/point EBPFP 1.6775 equivalent bits/point MSE 0.030865 ---------------------- -------------------------------------------------------- Time: 5.549s Load: 0.012s, Pack+Encode: 3.097s, Decode+Unpack: 2.440s ---------------------- -------------------------------------------------------- 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 0.0309 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample0-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.014s ------------------------------------------------------------ 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: 21,980B, BPFP=1.3416 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,836B, BPFP=0.4783 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,876B, BPFP=2.6553 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,900B, BPFP=1.2756 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,240B, BPFP=2.9883 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,528B, BPFP=0.9478 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,916B, BPFP=2.9092 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,384B, BPFP=0.6948 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,352B, BPFP=2.5273 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 108,044B, BPFP=0.7483 ⌛️ [2/4] FRONTEND: Frontend time: 2.668s (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: 2.400s [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 0.00000566 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000029 0.00012995 layer.1.conv_state 0.00048919 0.84803665 layer.2.ssm_state 0.00000001 0.00000877 layer.2.conv_state 0.00011763 0.17015107 layer.3.ssm_state 0.00000001 0.00001022 layer.3.conv_state 0.00011288 0.15808040 layer.4.ssm_state 0.00000001 0.00001512 layer.4.conv_state 0.00028411 0.34516537 layer.4.output 0.00000122 0.00058627 ------------------------------------------------------------------------------------- TOTAL 0.00001982 0.03048773 (elements=1,974,272) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1974272 Total Bytes 240824 BPFP 0.9758 bits/point EBPFP 1.5139 equivalent bits/point MSE 0.030488 ---------------------- -------------------------------------------------------- Time: 5.083s Load: 0.014s, Pack+Encode: 2.668s, Decode+Unpack: 2.400s ---------------------- -------------------------------------------------------- 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 0.0305 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample1-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,956B, BPFP=1.3401 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,748B, BPFP=0.4729 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,884B, BPFP=2.6572 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,884B, BPFP=1.2747 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,292B, BPFP=3.0010 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,648B, BPFP=0.9551 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,960B, BPFP=2.9199 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,412B, BPFP=0.6965 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,376B, BPFP=2.5332 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 143,632B, BPFP=1.0238 ⌛️ [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, 274, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.408s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000027 0.00013092 layer.1.conv_state 0.00049366 0.85064042 layer.2.ssm_state 0.00000001 0.00000877 layer.2.conv_state 0.00012530 0.17278293 layer.3.ssm_state 0.00000001 0.00001039 layer.3.conv_state 0.00007253 0.16037445 layer.4.ssm_state 0.00000001 0.00001512 layer.4.conv_state 0.00034856 0.35265401 layer.4.output 0.00000124 0.00060088 ------------------------------------------------------------------------------------- TOTAL 0.00002076 0.03125440 (elements=1,941,504) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1941504 Total Bytes 276560 BPFP 1.1396 bits/point EBPFP 1.6873 equivalent bits/point MSE 0.031254 ---------------------- -------------------------------------------------------- Time: 5.082s Load: 0.012s, Pack+Encode: 2.661s, Decode+Unpack: 2.408s ---------------------- -------------------------------------------------------- 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 0.0313 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample10-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,956B, BPFP=1.3401 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,744B, BPFP=0.4727 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,816B, BPFP=2.6406 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,068B, BPFP=1.2859 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,220B, BPFP=2.9834 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,716B, BPFP=0.9592 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,888B, BPFP=2.9023 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,688B, BPFP=0.7134 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,324B, BPFP=2.5205 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 92,576B, BPFP=1.0958 ⌛️ [2/4] FRONTEND: Frontend time: 2.664s (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: 2.293s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000027 0.00012919 layer.1.conv_state 0.00048753 0.84248018 layer.2.ssm_state 0.00000001 0.00000880 layer.2.conv_state 0.00013029 0.17106727 layer.3.ssm_state 0.00000001 0.00001042 layer.3.conv_state 0.00011288 0.15956604 layer.4.ssm_state 0.00000001 0.00001546 layer.4.conv_state 0.00021616 0.34660256 layer.4.output 0.00000217 0.00083440 ------------------------------------------------------------------------------------- TOTAL 0.00002496 0.04014710 (elements=1,495,040) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1495040 Total Bytes 225764 BPFP 1.2081 bits/point EBPFP 1.9208 equivalent bits/point MSE 0.040147 ---------------------- -------------------------------------------------------- Time: 4.967s Load: 0.009s, Pack+Encode: 2.664s, Decode+Unpack: 2.293s ---------------------- -------------------------------------------------------- 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 0.0401 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample101-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 21,712B, BPFP=1.3252 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,808B, BPFP=0.4766 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,828B, BPFP=2.6436 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,016B, BPFP=1.2827 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,256B, BPFP=2.9922 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,552B, BPFP=0.9492 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,920B, BPFP=2.9102 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,668B, BPFP=0.7122 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,396B, BPFP=2.5381 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,060B, BPFP=1.1340 ⌛️ [2/4] FRONTEND: Frontend time: 2.545s (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: 2.295s [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 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00013172 layer.1.conv_state 0.00049337 0.84760839 layer.2.ssm_state 0.00000001 0.00000871 layer.2.conv_state 0.00015958 0.17221496 layer.3.ssm_state 0.00000001 0.00001018 layer.3.conv_state 0.00007099 0.16085777 layer.4.ssm_state 0.00000001 0.00001512 layer.4.conv_state 0.00026744 0.34552938 layer.4.output 0.00000226 0.00083070 ------------------------------------------------------------------------------------- TOTAL 0.00002617 0.04061492 (elements=1,482,752) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1482752 Total Bytes 226984 BPFP 1.2247 bits/point EBPFP 1.9418 equivalent bits/point MSE 0.040615 ---------------------- -------------------------------------------------------- Time: 4.850s Load: 0.009s, Pack+Encode: 2.545s, Decode+Unpack: 2.295s ---------------------- -------------------------------------------------------- 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 0.0406 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample102-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,828B, BPFP=1.3323 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,792B, BPFP=0.4756 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,868B, BPFP=2.6533 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,768B, BPFP=1.2676 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,284B, BPFP=2.9990 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,300B, BPFP=0.9338 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,948B, BPFP=2.9170 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,480B, BPFP=0.7007 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,352B, BPFP=2.5273 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 98,280B, BPFP=1.0376 ⌛️ [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, 185, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.301s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000038 0.00013032 layer.1.conv_state 0.00049648 0.84387994 layer.2.ssm_state 0.00000001 0.00000879 layer.2.conv_state 0.00012672 0.17349575 layer.3.ssm_state 0.00000001 0.00001011 layer.3.conv_state 0.00007687 0.16125292 layer.4.ssm_state 0.00000001 0.00001492 layer.4.conv_state 0.00032793 0.34995458 layer.4.output 0.00000201 0.00073009 ------------------------------------------------------------------------------------- TOTAL 0.00002540 0.03823903 (elements=1,576,960) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1576960 Total Bytes 230668 BPFP 1.1702 bits/point EBPFP 1.8418 equivalent bits/point MSE 0.038239 ---------------------- -------------------------------------------------------- Time: 4.915s Load: 0.012s, Pack+Encode: 2.602s, Decode+Unpack: 2.301s ---------------------- -------------------------------------------------------- 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 0.0382 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample106-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.008s ------------------------------------------------------------ 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: 21,928B, BPFP=1.3384 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,832B, BPFP=0.4780 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,880B, BPFP=2.6562 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,888B, BPFP=1.2749 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,248B, BPFP=2.9902 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,156B, BPFP=0.9250 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,928B, BPFP=2.9121 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,108B, BPFP=0.7390 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,384B, BPFP=2.5352 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,740B, BPFP=1.1493 ⌛️ [2/4] FRONTEND: Frontend time: 2.552s (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: 2.298s [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 0.00000558 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000024 0.00013002 layer.1.conv_state 0.00050173 0.84366870 layer.2.ssm_state 0.00000001 0.00000877 layer.2.conv_state 0.00011340 0.17176215 layer.3.ssm_state 0.00000001 0.00001022 layer.3.conv_state 0.00011848 0.15998013 layer.4.ssm_state 0.00000005 0.00001458 layer.4.conv_state 0.00022485 0.33010754 layer.4.output 0.00000221 0.00081800 ------------------------------------------------------------------------------------- TOTAL 0.00002549 0.04026072 (elements=1,478,656) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1478656 Total Bytes 227860 BPFP 1.2328 bits/point EBPFP 1.9530 equivalent bits/point MSE 0.040261 ---------------------- -------------------------------------------------------- Time: 4.858s Load: 0.008s, Pack+Encode: 2.552s, Decode+Unpack: 2.298s ---------------------- -------------------------------------------------------- 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 0.0403 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample108-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 21,892B, BPFP=1.3362 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,792B, BPFP=0.4756 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,904B, BPFP=2.6621 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,800B, BPFP=1.2695 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,304B, BPFP=3.0039 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,520B, BPFP=0.9473 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,948B, BPFP=2.9170 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,544B, BPFP=0.7046 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,368B, BPFP=2.5312 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 143,548B, BPFP=1.4160 ⌛️ [2/4] FRONTEND: Frontend time: 2.714s (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: 2.356s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000010 0.00013029 layer.1.conv_state 0.00048181 0.85004663 layer.2.ssm_state 0.00000001 0.00000878 layer.2.conv_state 0.00010981 0.17400871 layer.3.ssm_state 0.00000001 0.00001051 layer.3.conv_state 0.00007216 0.16306888 layer.4.ssm_state 0.00000002 0.00001514 layer.4.conv_state 0.00032426 0.35578671 layer.4.output 0.00000183 0.00073840 ------------------------------------------------------------------------------------- TOTAL 0.00002372 0.03730604 (elements=1,630,208) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1630208 Total Bytes 276388 BPFP 1.3563 bits/point EBPFP 2.0082 equivalent bits/point MSE 0.037306 ---------------------- -------------------------------------------------------- Time: 5.082s Load: 0.011s, Pack+Encode: 2.714s, Decode+Unpack: 2.356s ---------------------- -------------------------------------------------------- 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 0.0373 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample110-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 21,976B, BPFP=1.3413 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,788B, BPFP=0.4753 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,828B, BPFP=2.6436 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,796B, BPFP=1.2693 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,248B, BPFP=2.9902 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,420B, BPFP=0.9412 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,848B, BPFP=2.8926 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,704B, BPFP=0.7144 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,244B, BPFP=2.5010 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 103,088B, BPFP=1.1844 ⌛️ [2/4] FRONTEND: Frontend time: 2.543s (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: 2.299s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00013007 layer.1.conv_state 0.00049252 0.84563148 layer.2.ssm_state 0.00000001 0.00000882 layer.2.conv_state 0.00011876 0.17028493 layer.3.ssm_state 0.00000001 0.00001028 layer.3.conv_state 0.00011952 0.15952493 layer.4.ssm_state 0.00000001 0.00001534 layer.4.conv_state 0.00023872 0.34289217 layer.4.output 0.00000204 0.00082849 ------------------------------------------------------------------------------------- TOTAL 0.00002508 0.03958329 (elements=1,515,520) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1515520 Total Bytes 235708 BPFP 1.2442 bits/point EBPFP 1.9443 equivalent bits/point MSE 0.039583 ---------------------- -------------------------------------------------------- Time: 4.851s Load: 0.009s, Pack+Encode: 2.543s, Decode+Unpack: 2.299s ---------------------- -------------------------------------------------------- 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 0.0396 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample115-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 21,788B, BPFP=1.3298 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,784B, BPFP=0.4751 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,856B, BPFP=2.6504 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,056B, BPFP=1.2852 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,260B, BPFP=2.9932 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,324B, BPFP=0.9353 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,928B, BPFP=2.9121 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,000B, BPFP=0.7324 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,296B, BPFP=2.5137 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 72,228B, BPFP=0.7425 ⌛️ [2/4] FRONTEND: Frontend time: 2.550s (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: 2.291s [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 0.00000558 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000023 0.00013146 layer.1.conv_state 0.00050470 0.84758639 layer.2.ssm_state 0.00000001 0.00000879 layer.2.conv_state 0.00013213 0.17102093 layer.3.ssm_state 0.00000001 0.00001039 layer.3.conv_state 0.00011581 0.15979056 layer.4.ssm_state 0.00000003 0.00001487 layer.4.conv_state 0.00021685 0.33889610 layer.4.output 0.00000199 0.00069777 ------------------------------------------------------------------------------------- TOTAL 0.00002388 0.03751094 (elements=1,597,440) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1597440 Total Bytes 205288 BPFP 1.0281 bits/point EBPFP 1.6945 equivalent bits/point MSE 0.037511 ---------------------- -------------------------------------------------------- Time: 4.852s Load: 0.010s, Pack+Encode: 2.550s, Decode+Unpack: 2.291s ---------------------- -------------------------------------------------------- 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 0.0375 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample116-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.012s ------------------------------------------------------------ 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: 22,144B, BPFP=1.3516 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,812B, BPFP=0.4768 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,816B, BPFP=2.6406 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,768B, BPFP=1.2676 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,280B, BPFP=2.9980 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,276B, BPFP=0.9324 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,924B, BPFP=2.9111 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,740B, BPFP=0.7166 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,320B, BPFP=2.5195 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 112,468B, BPFP=0.9509 ⌛️ [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, 231, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.345s [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 0.00000567 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000024 0.00013141 layer.1.conv_state 0.00050504 0.84504354 layer.2.ssm_state 0.00000001 0.00000867 layer.2.conv_state 0.00012981 0.17155042 layer.3.ssm_state 0.00000001 0.00001037 layer.3.conv_state 0.00007554 0.16035262 layer.4.ssm_state 0.00000001 0.00001527 layer.4.conv_state 0.00028923 0.35242984 layer.4.output 0.00000157 0.00066683 ------------------------------------------------------------------------------------- TOTAL 0.00002213 0.03421669 (elements=1,765,376) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1765376 Total Bytes 245316 BPFP 1.1117 bits/point EBPFP 1.7137 equivalent bits/point MSE 0.034217 ---------------------- -------------------------------------------------------- Time: 4.958s Load: 0.012s, Pack+Encode: 2.601s, Decode+Unpack: 2.345s ---------------------- -------------------------------------------------------- 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 0.0342 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample12-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 22,004B, BPFP=1.3430 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,820B, BPFP=0.4773 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,860B, BPFP=2.6514 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,804B, BPFP=1.2698 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,248B, BPFP=2.9902 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 14,992B, BPFP=0.9150 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,924B, BPFP=2.9111 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,716B, BPFP=0.7151 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,320B, BPFP=2.5195 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 92,428B, BPFP=0.9865 ⌛️ [2/4] FRONTEND: Frontend time: 2.560s (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: 2.291s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000024 0.00013065 layer.1.conv_state 0.00048839 0.84312081 layer.2.ssm_state 0.00000001 0.00000882 layer.2.conv_state 0.00008466 0.17138578 layer.3.ssm_state 0.00000001 0.00001046 layer.3.conv_state 0.00011958 0.16026792 layer.4.ssm_state 0.00000001 0.00001542 layer.4.conv_state 0.00031804 0.35003611 layer.4.output 0.00000201 0.00071990 ------------------------------------------------------------------------------------- TOTAL 0.00002515 0.03835133 (elements=1,568,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1568768 Total Bytes 224884 BPFP 1.1468 bits/point EBPFP 1.8223 equivalent bits/point MSE 0.038351 ---------------------- -------------------------------------------------------- Time: 4.862s Load: 0.010s, Pack+Encode: 2.560s, Decode+Unpack: 2.291s ---------------------- -------------------------------------------------------- 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 0.0384 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample120-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,980B, BPFP=1.3416 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,780B, BPFP=0.4749 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,828B, BPFP=2.6436 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,784B, BPFP=1.2686 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,276B, BPFP=2.9971 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,872B, BPFP=0.9688 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,888B, BPFP=2.9023 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,592B, BPFP=0.7686 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,236B, BPFP=2.4990 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 89,872B, BPFP=0.9190 ⌛️ [2/4] FRONTEND: Frontend time: 2.556s (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: 2.295s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00012999 layer.1.conv_state 0.00046948 0.84862822 layer.2.ssm_state 0.00000001 0.00000877 layer.2.conv_state 0.00011482 0.17247064 layer.3.ssm_state 0.00000001 0.00001091 layer.3.conv_state 0.00011212 0.15927970 layer.4.ssm_state 0.00000002 0.00001420 layer.4.conv_state 0.00024277 0.32811603 layer.4.output 0.00000205 0.00070555 ------------------------------------------------------------------------------------- TOTAL 0.00002323 0.03724041 (elements=1,601,536) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1601536 Total Bytes 223876 BPFP 1.1183 bits/point EBPFP 1.7877 equivalent bits/point MSE 0.037240 ---------------------- -------------------------------------------------------- Time: 4.861s Load: 0.010s, Pack+Encode: 2.556s, Decode+Unpack: 2.295s ---------------------- -------------------------------------------------------- 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 0.0372 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample122-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 22,048B, BPFP=1.3457 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,796B, BPFP=0.4758 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,856B, BPFP=2.6504 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,960B, BPFP=1.2793 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,192B, BPFP=2.9766 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,548B, BPFP=0.9490 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,904B, BPFP=2.9062 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,432B, BPFP=0.6978 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,296B, BPFP=2.5137 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 93,912B, BPFP=1.0481 ⌛️ [2/4] FRONTEND: Frontend time: 2.556s (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: 2.295s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00013042 layer.1.conv_state 0.00051384 0.83763754 layer.2.ssm_state 0.00000001 0.00000885 layer.2.conv_state 0.00012234 0.17074743 layer.3.ssm_state 0.00000001 0.00001055 layer.3.conv_state 0.00007815 0.15927114 layer.4.ssm_state 0.00000001 0.00001548 layer.4.conv_state 0.00034266 0.35327441 layer.4.output 0.00000210 0.00076297 ------------------------------------------------------------------------------------- TOTAL 0.00002667 0.03909146 (elements=1,536,000) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1536000 Total Bytes 226712 BPFP 1.1808 bits/point EBPFP 1.8725 equivalent bits/point MSE 0.039091 ---------------------- -------------------------------------------------------- Time: 4.860s Load: 0.010s, Pack+Encode: 2.556s, Decode+Unpack: 2.295s ---------------------- -------------------------------------------------------- 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 0.0391 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample127-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,052B, BPFP=1.3459 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,772B, BPFP=0.4744 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,900B, BPFP=1.2756 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,200B, BPFP=2.9785 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,412B, BPFP=0.9407 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,868B, BPFP=2.8975 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,952B, BPFP=0.7295 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,280B, BPFP=2.5098 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 93,804B, BPFP=1.0905 ⌛️ [2/4] FRONTEND: Frontend time: 2.539s (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: 2.301s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00012876 layer.1.conv_state 0.00051486 0.83929944 layer.2.ssm_state 0.00000001 0.00000879 layer.2.conv_state 0.00016795 0.17169404 layer.3.ssm_state 0.00000001 0.00001037 layer.3.conv_state 0.00011430 0.15929253 layer.4.ssm_state 0.00000002 0.00001460 layer.4.conv_state 0.00021850 0.33533791 layer.4.output 0.00000215 0.00079740 ------------------------------------------------------------------------------------- TOTAL 0.00002626 0.03950325 (elements=1,507,328) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1507328 Total Bytes 226860 BPFP 1.2040 bits/point EBPFP 1.9102 equivalent bits/point MSE 0.039503 ---------------------- -------------------------------------------------------- Time: 4.850s Load: 0.009s, Pack+Encode: 2.539s, Decode+Unpack: 2.301s ---------------------- -------------------------------------------------------- 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 0.0395 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample128-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 22,056B, BPFP=1.3462 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,768B, BPFP=0.4741 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,856B, BPFP=2.6504 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,956B, BPFP=1.2791 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,188B, BPFP=2.9756 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,312B, BPFP=0.9346 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,900B, BPFP=2.9053 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,592B, BPFP=0.7075 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,328B, BPFP=2.5215 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 93,216B, BPFP=1.0059 ⌛️ [2/4] FRONTEND: Frontend time: 2.547s (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: 2.294s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000027 0.00013037 layer.1.conv_state 0.00048726 0.84087145 layer.2.ssm_state 0.00000001 0.00000883 layer.2.conv_state 0.00012801 0.17027299 layer.3.ssm_state 0.00000001 0.00001044 layer.3.conv_state 0.00011944 0.15869400 layer.4.ssm_state 0.00000001 0.00001519 layer.4.conv_state 0.00034131 0.35069105 layer.4.output 0.00000206 0.00072314 ------------------------------------------------------------------------------------- TOTAL 0.00002667 0.03846048 (elements=1,560,576) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1560576 Total Bytes 225940 BPFP 1.1582 bits/point EBPFP 1.8386 equivalent bits/point MSE 0.038460 ---------------------- -------------------------------------------------------- Time: 4.851s Load: 0.010s, Pack+Encode: 2.547s, Decode+Unpack: 2.294s ---------------------- -------------------------------------------------------- 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 0.0385 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample129-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,932B, BPFP=1.3386 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,808B, BPFP=0.4766 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,880B, BPFP=2.6562 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,932B, BPFP=1.2776 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,280B, BPFP=2.9980 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,600B, BPFP=0.9521 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,956B, BPFP=2.9189 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,464B, BPFP=0.6997 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,200B, BPFP=2.4902 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 95,064B, BPFP=1.1185 ⌛️ [2/4] FRONTEND: Frontend time: 2.542s (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: 2.291s [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 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000010 0.00012735 layer.1.conv_state 0.00049597 0.84607518 layer.2.ssm_state 0.00000001 0.00000877 layer.2.conv_state 0.00012619 0.17201307 layer.3.ssm_state 0.00000001 0.00001061 layer.3.conv_state 0.00006821 0.15951800 layer.4.ssm_state 0.00000001 0.00001546 layer.4.conv_state 0.00029570 0.34640476 layer.4.output 0.00000206 0.00083147 ------------------------------------------------------------------------------------- TOTAL 0.00002569 0.04013209 (elements=1,499,136) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1499136 Total Bytes 227884 BPFP 1.2161 bits/point EBPFP 1.9249 equivalent bits/point MSE 0.040132 ---------------------- -------------------------------------------------------- Time: 4.842s Load: 0.009s, Pack+Encode: 2.542s, Decode+Unpack: 2.291s ---------------------- -------------------------------------------------------- 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 0.0401 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample130-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,968B, BPFP=1.3408 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,804B, BPFP=0.4763 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,884B, BPFP=2.6572 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,876B, BPFP=1.2742 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,220B, BPFP=2.9834 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,300B, BPFP=0.9338 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,848B, BPFP=2.8926 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,804B, BPFP=0.7205 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,276B, BPFP=2.5088 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,844B, BPFP=1.0393 ⌛️ [2/4] FRONTEND: Frontend time: 2.538s (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: 2.284s [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 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000027 0.00013110 layer.1.conv_state 0.00049604 0.84527093 layer.2.ssm_state 0.00000001 0.00000876 layer.2.conv_state 0.00012832 0.17106450 layer.3.ssm_state 0.00000001 0.00001011 layer.3.conv_state 0.00011179 0.15975235 layer.4.ssm_state 0.00000001 0.00001493 layer.4.conv_state 0.00022786 0.34004879 layer.4.output 0.00000216 0.00075139 ------------------------------------------------------------------------------------- TOTAL 0.00002431 0.03828315 (elements=1,564,672) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1564672 Total Bytes 229592 BPFP 1.1739 bits/point EBPFP 1.8526 equivalent bits/point MSE 0.038283 ---------------------- -------------------------------------------------------- Time: 4.832s Load: 0.010s, Pack+Encode: 2.538s, Decode+Unpack: 2.284s ---------------------- -------------------------------------------------------- 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 0.0383 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample134-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 21,980B, BPFP=1.3416 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,784B, BPFP=0.4751 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,872B, BPFP=2.6543 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,884B, BPFP=1.2747 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,216B, BPFP=2.9824 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,528B, BPFP=0.9478 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,888B, BPFP=2.9023 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,376B, BPFP=0.6943 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,224B, BPFP=2.4961 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 97,752B, BPFP=1.0726 ⌛️ [2/4] FRONTEND: Frontend time: 2.532s (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: 2.295s [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 0.00000558 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00012972 layer.1.conv_state 0.00048610 0.84663892 layer.2.ssm_state 0.00000001 0.00000878 layer.2.conv_state 0.00014962 0.17230305 layer.3.ssm_state 0.00000001 0.00001051 layer.3.conv_state 0.00007021 0.16048883 layer.4.ssm_state 0.00000001 0.00001519 layer.4.conv_state 0.00031005 0.34629935 layer.4.output 0.00000198 0.00074606 ------------------------------------------------------------------------------------- TOTAL 0.00002555 0.03888078 (elements=1,548,288) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1548288 Total Bytes 230272 BPFP 1.1898 bits/point EBPFP 1.8745 equivalent bits/point MSE 0.038881 ---------------------- -------------------------------------------------------- Time: 4.836s Load: 0.010s, Pack+Encode: 2.532s, Decode+Unpack: 2.295s ---------------------- -------------------------------------------------------- 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 0.0389 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample135-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,912B, BPFP=1.3374 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,824B, BPFP=0.4775 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,852B, BPFP=1.2727 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,272B, BPFP=2.9961 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,384B, BPFP=0.9390 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,936B, BPFP=2.9141 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,248B, BPFP=0.7476 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,304B, BPFP=2.5156 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 93,840B, BPFP=1.1041 ⌛️ [2/4] FRONTEND: Frontend time: 2.545s (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: 2.297s [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 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00012943 layer.1.conv_state 0.00050515 0.84441239 layer.2.ssm_state 0.00000001 0.00000874 layer.2.conv_state 0.00013851 0.17279394 layer.3.ssm_state 0.00000001 0.00001044 layer.3.conv_state 0.00011516 0.16110776 layer.4.ssm_state 0.00000001 0.00001456 layer.4.conv_state 0.00022993 0.33119372 layer.4.output 0.00000208 0.00080347 ------------------------------------------------------------------------------------- TOTAL 0.00002578 0.03980246 (elements=1,499,136) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1499136 Total Bytes 227192 BPFP 1.2124 bits/point EBPFP 1.9240 equivalent bits/point MSE 0.039802 ---------------------- -------------------------------------------------------- Time: 4.853s Load: 0.011s, Pack+Encode: 2.545s, Decode+Unpack: 2.297s ---------------------- -------------------------------------------------------- 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 0.0398 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample142-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.012s ------------------------------------------------------------ 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: 21,788B, BPFP=1.3298 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,772B, BPFP=0.4744 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,648B, BPFP=1.2603 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,216B, BPFP=2.9824 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,392B, BPFP=0.9395 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,928B, BPFP=2.9121 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,680B, BPFP=0.7129 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,360B, BPFP=2.5293 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 92,356B, BPFP=1.1713 ⌛️ [2/4] FRONTEND: Frontend time: 2.535s (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: 2.295s [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 0.00000558 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00012945 layer.1.conv_state 0.00051446 0.83890623 layer.2.ssm_state 0.00000001 0.00000884 layer.2.conv_state 0.00012903 0.17145069 layer.3.ssm_state 0.00000001 0.00001041 layer.3.conv_state 0.00007035 0.16060635 layer.4.ssm_state 0.00000004 0.00001474 layer.4.conv_state 0.00022206 0.33701959 layer.4.output 0.00000218 0.00082996 ------------------------------------------------------------------------------------- TOTAL 0.00002543 0.04110155 (elements=1,449,984) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1449984 Total Bytes 224760 BPFP 1.2401 bits/point EBPFP 1.9706 equivalent bits/point MSE 0.041102 ---------------------- -------------------------------------------------------- Time: 4.841s Load: 0.012s, Pack+Encode: 2.535s, Decode+Unpack: 2.295s ---------------------- -------------------------------------------------------- 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 0.0411 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample143-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 21,920B, BPFP=1.3379 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,772B, BPFP=0.4744 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,824B, BPFP=2.6426 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,112B, BPFP=1.2886 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,192B, BPFP=2.9766 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,620B, BPFP=0.9534 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,872B, BPFP=2.8984 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,468B, BPFP=0.7000 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,300B, BPFP=2.5146 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,480B, BPFP=1.1116 ⌛️ [2/4] FRONTEND: Frontend time: 2.532s (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: 2.289s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00012960 layer.1.conv_state 0.00048448 0.84198701 layer.2.ssm_state 0.00000001 0.00000876 layer.2.conv_state 0.00013520 0.17060325 layer.3.ssm_state 0.00000001 0.00001037 layer.3.conv_state 0.00007505 0.15972179 layer.4.ssm_state 0.00000002 0.00001562 layer.4.conv_state 0.00034342 0.35386622 layer.4.output 0.00000207 0.00081646 ------------------------------------------------------------------------------------- TOTAL 0.00002685 0.04017284 (elements=1,499,136) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1499136 Total Bytes 227328 BPFP 1.2131 bits/point EBPFP 1.9220 equivalent bits/point MSE 0.040173 ---------------------- -------------------------------------------------------- Time: 4.832s Load: 0.011s, Pack+Encode: 2.532s, Decode+Unpack: 2.289s ---------------------- -------------------------------------------------------- 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 0.0402 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample144-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,972B, BPFP=1.3411 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,768B, BPFP=0.4741 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,844B, BPFP=2.6475 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,596B, BPFP=1.2571 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,276B, BPFP=2.9971 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,400B, BPFP=0.9399 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,900B, BPFP=2.9053 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,980B, BPFP=0.7312 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,276B, BPFP=2.5088 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 129,896B, BPFP=1.2878 ⌛️ [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, 197, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.350s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000027 0.00012885 layer.1.conv_state 0.00049519 0.84264940 layer.2.ssm_state 0.00000001 0.00000875 layer.2.conv_state 0.00014288 0.17391546 layer.3.ssm_state 0.00000001 0.00001035 layer.3.conv_state 0.00007032 0.16157213 layer.4.ssm_state 0.00000002 0.00001483 layer.4.conv_state 0.00023237 0.34108609 layer.4.output 0.00000183 0.00077626 ------------------------------------------------------------------------------------- TOTAL 0.00002283 0.03693944 (elements=1,626,112) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1626112 Total Bytes 262676 BPFP 1.2923 bits/point EBPFP 1.9455 equivalent bits/point MSE 0.036939 ---------------------- -------------------------------------------------------- Time: 4.952s Load: 0.013s, Pack+Encode: 2.589s, Decode+Unpack: 2.350s ---------------------- -------------------------------------------------------- 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 0.0369 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample148-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,984B, BPFP=1.3418 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,772B, BPFP=0.4744 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,852B, BPFP=1.2727 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,228B, BPFP=2.9854 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,196B, BPFP=0.9275 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,952B, BPFP=2.9180 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,628B, BPFP=0.7097 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,276B, BPFP=2.5088 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 97,548B, BPFP=1.1341 ⌛️ [2/4] FRONTEND: Frontend time: 2.529s (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: 2.292s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00013053 layer.1.conv_state 0.00048592 0.84427762 layer.2.ssm_state 0.00000001 0.00000879 layer.2.conv_state 0.00011654 0.17063631 layer.3.ssm_state 0.00000001 0.00001030 layer.3.conv_state 0.00011837 0.15945055 layer.4.ssm_state 0.00000001 0.00001475 layer.4.conv_state 0.00020779 0.33622789 layer.4.output 0.00000221 0.00080973 ------------------------------------------------------------------------------------- TOTAL 0.00002440 0.03961705 (elements=1,507,328) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1507328 Total Bytes 230056 BPFP 1.2210 bits/point EBPFP 1.9243 equivalent bits/point MSE 0.039617 ---------------------- -------------------------------------------------------- Time: 4.830s Load: 0.009s, Pack+Encode: 2.529s, Decode+Unpack: 2.292s ---------------------- -------------------------------------------------------- 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 0.0396 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample149-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 21,928B, BPFP=1.3384 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,828B, BPFP=0.4778 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,876B, BPFP=2.6553 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,852B, BPFP=1.2727 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,284B, BPFP=2.9990 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,460B, BPFP=0.9436 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,940B, BPFP=2.9150 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,156B, BPFP=0.7419 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,324B, BPFP=2.5205 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 91,628B, BPFP=1.0912 ⌛️ [2/4] FRONTEND: Frontend time: 2.532s (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: 2.293s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000029 0.00012991 layer.1.conv_state 0.00049823 0.84623671 layer.2.ssm_state 0.00000001 0.00000874 layer.2.conv_state 0.00014568 0.17306842 layer.3.ssm_state 0.00000001 0.00001050 layer.3.conv_state 0.00011339 0.16142140 layer.4.ssm_state 0.00000002 0.00001476 layer.4.conv_state 0.00023177 0.33488122 layer.4.output 0.00000211 0.00080990 ------------------------------------------------------------------------------------- TOTAL 0.00002593 0.04015377 (elements=1,490,944) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1490944 Total Bytes 225044 BPFP 1.2075 bits/point EBPFP 1.9234 equivalent bits/point MSE 0.040154 ---------------------- -------------------------------------------------------- Time: 4.833s Load: 0.009s, Pack+Encode: 2.532s, Decode+Unpack: 2.293s ---------------------- -------------------------------------------------------- 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 0.0402 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample153-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.012s ------------------------------------------------------------ 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: 22,004B, BPFP=1.3430 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,796B, BPFP=0.4758 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,844B, BPFP=2.6475 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,908B, BPFP=1.2761 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,220B, BPFP=2.9834 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,572B, BPFP=0.9504 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,904B, BPFP=2.9062 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,672B, BPFP=0.7124 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,260B, BPFP=2.5049 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 83,848B, BPFP=1.0047 ⌛️ [2/4] FRONTEND: Frontend time: 2.540s (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: 2.292s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000043 0.00012868 layer.1.conv_state 0.00049103 0.84332746 layer.2.ssm_state 0.00000001 0.00000884 layer.2.conv_state 0.00014094 0.17110451 layer.3.ssm_state 0.00000001 0.00001058 layer.3.conv_state 0.00012075 0.15956336 layer.4.ssm_state 0.00000001 0.00001511 layer.4.conv_state 0.00020319 0.34651250 layer.4.output 0.00000221 0.00077655 ------------------------------------------------------------------------------------- TOTAL 0.00002532 0.04035511 (elements=1,486,848) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1486848 Total Bytes 216796 BPFP 1.1665 bits/point EBPFP 1.8818 equivalent bits/point MSE 0.040355 ---------------------- -------------------------------------------------------- Time: 4.845s Load: 0.012s, Pack+Encode: 2.540s, Decode+Unpack: 2.292s ---------------------- -------------------------------------------------------- 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 0.0404 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample156-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 22,148B, BPFP=1.3518 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,804B, BPFP=0.4763 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,828B, BPFP=2.6436 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,056B, BPFP=1.2852 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,236B, BPFP=2.9873 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,656B, BPFP=0.9556 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,888B, BPFP=2.9023 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,892B, BPFP=0.7258 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,344B, BPFP=2.5254 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,312B, BPFP=1.1612 ⌛️ [2/4] FRONTEND: Frontend time: 2.562s (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: 2.294s [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 0.00000567 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00013055 layer.1.conv_state 0.00050753 0.84385276 layer.2.ssm_state 0.00000001 0.00000885 layer.2.conv_state 0.00012648 0.17163682 layer.3.ssm_state 0.00000001 0.00001073 layer.3.conv_state 0.00011960 0.16146693 layer.4.ssm_state 0.00000001 0.00001530 layer.4.conv_state 0.00022111 0.34093377 layer.4.output 0.00000221 0.00078269 ------------------------------------------------------------------------------------- TOTAL 0.00002578 0.04040954 (elements=1,482,752) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1482752 Total Bytes 229932 BPFP 1.2406 bits/point EBPFP 1.9615 equivalent bits/point MSE 0.040410 ---------------------- -------------------------------------------------------- Time: 4.867s Load: 0.010s, Pack+Encode: 2.562s, Decode+Unpack: 2.294s ---------------------- -------------------------------------------------------- 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 0.0404 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample158-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.012s ------------------------------------------------------------ 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: 21,984B, BPFP=1.3418 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,816B, BPFP=0.4771 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,824B, BPFP=2.6426 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,020B, BPFP=1.2830 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,208B, BPFP=2.9805 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,148B, BPFP=0.9246 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,872B, BPFP=2.8984 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,856B, BPFP=0.7236 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,292B, BPFP=2.5127 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 90,148B, BPFP=1.0869 ⌛️ [2/4] FRONTEND: Frontend time: 2.535s (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: 2.294s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00013086 layer.1.conv_state 0.00049392 0.84219044 layer.2.ssm_state 0.00000001 0.00000885 layer.2.conv_state 0.00013642 0.17037173 layer.3.ssm_state 0.00000001 0.00001009 layer.3.conv_state 0.00011766 0.15867242 layer.4.ssm_state 0.00000003 0.00001473 layer.4.conv_state 0.00021047 0.33732313 layer.4.output 0.00000223 0.00081561 ------------------------------------------------------------------------------------- TOTAL 0.00002544 0.04021795 (elements=1,482,752) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1482752 Total Bytes 222936 BPFP 1.2028 bits/point EBPFP 1.9193 equivalent bits/point MSE 0.040218 ---------------------- -------------------------------------------------------- Time: 4.841s Load: 0.012s, Pack+Encode: 2.535s, Decode+Unpack: 2.294s ---------------------- -------------------------------------------------------- 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 0.0402 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample159-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 21,980B, BPFP=1.3416 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,764B, BPFP=0.4739 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,812B, BPFP=2.6396 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,908B, BPFP=1.2761 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,232B, BPFP=2.9863 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,556B, BPFP=0.9495 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,932B, BPFP=2.9131 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,716B, BPFP=0.7151 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,260B, BPFP=2.5049 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,120B, BPFP=1.2433 ⌛️ [2/4] FRONTEND: Frontend time: 2.545s (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: 2.303s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000012 0.00013000 layer.1.conv_state 0.00048080 0.84292483 layer.2.ssm_state 0.00000001 0.00000885 layer.2.conv_state 0.00015695 0.17141579 layer.3.ssm_state 0.00000001 0.00001053 layer.3.conv_state 0.00007366 0.16024612 layer.4.ssm_state 0.00000001 0.00001516 layer.4.conv_state 0.00025051 0.34688032 layer.4.output 0.00000211 0.00086027 ------------------------------------------------------------------------------------- TOTAL 0.00002617 0.04176622 (elements=1,437,696) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1437696 Total Bytes 229048 BPFP 1.2745 bits/point EBPFP 2.0142 equivalent bits/point MSE 0.041766 ---------------------- -------------------------------------------------------- Time: 4.857s Load: 0.009s, Pack+Encode: 2.545s, Decode+Unpack: 2.303s ---------------------- -------------------------------------------------------- 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 0.0418 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample162-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,004B, BPFP=1.3430 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,788B, BPFP=0.4753 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,936B, BPFP=2.6699 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,864B, BPFP=1.2734 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,312B, BPFP=3.0059 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,580B, BPFP=0.9509 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,980B, BPFP=2.9248 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,308B, BPFP=0.6902 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,284B, BPFP=2.5107 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 93,152B, BPFP=0.9678 ⌛️ [2/4] FRONTEND: Frontend time: 2.548s (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: 2.292s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00012671 layer.1.conv_state 0.00047809 0.85267264 layer.2.ssm_state 0.00000001 0.00000873 layer.2.conv_state 0.00013752 0.17493302 layer.3.ssm_state 0.00000001 0.00001045 layer.3.conv_state 0.00007009 0.16284947 layer.4.ssm_state 0.00000002 0.00001546 layer.4.conv_state 0.00055818 0.36314467 layer.4.output 0.00000209 0.00079330 ------------------------------------------------------------------------------------- TOTAL 0.00002970 0.03849521 (elements=1,589,248) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1589248 Total Bytes 225976 BPFP 1.1375 bits/point EBPFP 1.8061 equivalent bits/point MSE 0.038495 ---------------------- -------------------------------------------------------- Time: 4.851s Load: 0.012s, Pack+Encode: 2.548s, Decode+Unpack: 2.292s ---------------------- -------------------------------------------------------- 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 0.0385 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample166-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,916B, BPFP=1.3376 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,792B, BPFP=0.4756 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,840B, BPFP=2.6465 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,700B, BPFP=1.2634 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,212B, BPFP=2.9814 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,056B, BPFP=0.9189 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,928B, BPFP=2.9121 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,600B, BPFP=0.7080 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,408B, BPFP=2.5410 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,996B, BPFP=1.0725 ⌛️ [2/4] FRONTEND: Frontend time: 2.538s (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: 2.295s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00013146 layer.1.conv_state 0.00049089 0.83895308 layer.2.ssm_state 0.00000001 0.00000875 layer.2.conv_state 0.00013154 0.16970760 layer.3.ssm_state 0.00000001 0.00001039 layer.3.conv_state 0.00011542 0.15892412 layer.4.ssm_state 0.00000001 0.00001498 layer.4.conv_state 0.00026976 0.34456664 layer.4.output 0.00000213 0.00077455 ------------------------------------------------------------------------------------- TOTAL 0.00002576 0.03911408 (elements=1,527,808) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1527808 Total Bytes 227216 BPFP 1.1898 bits/point EBPFP 1.8821 equivalent bits/point MSE 0.039114 ---------------------- -------------------------------------------------------- Time: 4.844s Load: 0.011s, Pack+Encode: 2.538s, Decode+Unpack: 2.295s ---------------------- -------------------------------------------------------- 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 0.0391 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample167-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 21,996B, BPFP=1.3425 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,836B, BPFP=0.4783 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,820B, BPFP=2.6416 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,032B, BPFP=1.2837 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,200B, BPFP=2.9785 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,520B, BPFP=0.9473 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,852B, BPFP=2.8936 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,060B, BPFP=0.6750 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,304B, BPFP=2.5156 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 93,924B, BPFP=1.0135 ⌛️ [2/4] FRONTEND: Frontend time: 2.540s (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: 2.292s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000011 0.00012772 layer.1.conv_state 0.00048935 0.84607047 layer.2.ssm_state 0.00000001 0.00000881 layer.2.conv_state 0.00014754 0.17092940 layer.3.ssm_state 0.00000001 0.00001044 layer.3.conv_state 0.00011438 0.15865819 layer.4.ssm_state 0.00000005 0.00001610 layer.4.conv_state 0.00050179 0.35722226 layer.4.output 0.00000190 0.00071138 ------------------------------------------------------------------------------------- TOTAL 0.00003030 0.03871407 (elements=1,560,576) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1560576 Total Bytes 226312 BPFP 1.1601 bits/point EBPFP 1.8388 equivalent bits/point MSE 0.038714 ---------------------- -------------------------------------------------------- Time: 4.842s Load: 0.010s, Pack+Encode: 2.540s, Decode+Unpack: 2.292s ---------------------- -------------------------------------------------------- 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 0.0387 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample173-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 21,884B, BPFP=1.3357 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,788B, BPFP=0.4753 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,892B, BPFP=2.6592 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,808B, BPFP=1.2700 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,268B, BPFP=2.9951 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,360B, BPFP=0.9375 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,860B, BPFP=2.8955 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,720B, BPFP=0.7153 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,176B, BPFP=2.4844 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 119,960B, BPFP=1.1210 ⌛️ [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, 209, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.349s [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 0.00000557 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000014 0.00012954 layer.1.conv_state 0.00048877 0.84613496 layer.2.ssm_state 0.00000001 0.00000888 layer.2.conv_state 0.00012277 0.17274185 layer.3.ssm_state 0.00000001 0.00001025 layer.3.conv_state 0.00012099 0.16094342 layer.4.ssm_state 0.00000001 0.00001483 layer.4.conv_state 0.00025427 0.34345657 layer.4.output 0.00000161 0.00069778 ------------------------------------------------------------------------------------- TOTAL 0.00002299 0.03591766 (elements=1,675,264) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1675264 Total Bytes 252484 BPFP 1.2057 bits/point EBPFP 1.8386 equivalent bits/point MSE 0.035918 ---------------------- -------------------------------------------------------- Time: 4.961s Load: 0.011s, Pack+Encode: 2.601s, Decode+Unpack: 2.349s ---------------------- -------------------------------------------------------- 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 0.0359 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample175-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 21,940B, BPFP=1.3391 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,780B, BPFP=0.4749 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,860B, BPFP=2.6514 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,848B, BPFP=1.2725 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,256B, BPFP=2.9922 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,412B, BPFP=0.9407 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,888B, BPFP=2.9023 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,004B, BPFP=0.7327 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,224B, BPFP=2.4961 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 93,084B, BPFP=1.0389 ⌛️ [2/4] FRONTEND: Frontend time: 2.548s (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: 2.298s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00012823 layer.1.conv_state 0.00049563 0.83918303 layer.2.ssm_state 0.00000001 0.00000874 layer.2.conv_state 0.00014857 0.17227918 layer.3.ssm_state 0.00000001 0.00001060 layer.3.conv_state 0.00007001 0.16050263 layer.4.ssm_state 0.00000002 0.00001468 layer.4.conv_state 0.00020707 0.33621737 layer.4.output 0.00000204 0.00074529 ------------------------------------------------------------------------------------- TOTAL 0.00002375 0.03881098 (elements=1,536,000) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1536000 Total Bytes 226064 BPFP 1.1774 bits/point EBPFP 1.8700 equivalent bits/point MSE 0.038811 ---------------------- -------------------------------------------------------- Time: 4.858s Load: 0.011s, Pack+Encode: 2.548s, Decode+Unpack: 2.298s ---------------------- -------------------------------------------------------- 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 0.0388 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample179-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 22,032B, BPFP=1.3447 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,780B, BPFP=0.4749 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,824B, BPFP=2.6426 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,856B, BPFP=1.2729 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,284B, BPFP=2.9990 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,100B, BPFP=0.9216 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,856B, BPFP=2.8945 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,944B, BPFP=0.7290 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,312B, BPFP=2.5176 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,472B, BPFP=1.0955 ⌛️ [2/4] FRONTEND: Frontend time: 2.558s (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: 2.294s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00013080 layer.1.conv_state 0.00050694 0.84539694 layer.2.ssm_state 0.00000001 0.00000884 layer.2.conv_state 0.00012346 0.17192259 layer.3.ssm_state 0.00000001 0.00001044 layer.3.conv_state 0.00011101 0.16048442 layer.4.ssm_state 0.00000002 0.00001469 layer.4.conv_state 0.00022697 0.33456230 layer.4.output 0.00000216 0.00080256 ------------------------------------------------------------------------------------- TOTAL 0.00002499 0.03923464 (elements=1,523,712) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1523712 Total Bytes 229228 BPFP 1.2035 bits/point EBPFP 1.9005 equivalent bits/point MSE 0.039235 ---------------------- -------------------------------------------------------- Time: 4.861s Load: 0.009s, Pack+Encode: 2.558s, Decode+Unpack: 2.294s ---------------------- -------------------------------------------------------- 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 0.0392 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample186-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,032B, BPFP=1.3447 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,836B, BPFP=0.4783 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,916B, BPFP=1.2766 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,220B, BPFP=2.9834 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,464B, BPFP=0.9438 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,960B, BPFP=2.9199 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,388B, BPFP=0.6951 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,420B, BPFP=2.5439 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 125,408B, BPFP=1.2691 ⌛️ [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, 193, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.360s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000031 0.00012862 layer.1.conv_state 0.00048110 0.84665805 layer.2.ssm_state 0.00000001 0.00000872 layer.2.conv_state 0.00012031 0.17119509 layer.3.ssm_state 0.00000001 0.00001045 layer.3.conv_state 0.00007804 0.16008005 layer.4.ssm_state 0.00000001 0.00001537 layer.4.conv_state 0.00027310 0.34332547 layer.4.output 0.00000181 0.00075056 ------------------------------------------------------------------------------------- TOTAL 0.00002328 0.03733636 (elements=1,609,728) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1609728 Total Bytes 258264 BPFP 1.2835 bits/point EBPFP 1.9438 equivalent bits/point MSE 0.037336 ---------------------- -------------------------------------------------------- Time: 4.971s Load: 0.010s, Pack+Encode: 2.601s, Decode+Unpack: 2.360s ---------------------- -------------------------------------------------------- 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 0.0373 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample19-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,924B, BPFP=1.3381 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,800B, BPFP=0.4761 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,848B, BPFP=2.6484 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,912B, BPFP=1.2764 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,276B, BPFP=2.9971 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,532B, BPFP=0.9480 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,884B, BPFP=2.9014 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,808B, BPFP=0.7207 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,348B, BPFP=2.5264 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 95,992B, BPFP=1.1791 ⌛️ [2/4] FRONTEND: Frontend time: 2.553s (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: 2.306s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000038 0.00012884 layer.1.conv_state 0.00048847 0.84321153 layer.2.ssm_state 0.00000001 0.00000884 layer.2.conv_state 0.00014206 0.17253722 layer.3.ssm_state 0.00000001 0.00001052 layer.3.conv_state 0.00011550 0.16058524 layer.4.ssm_state 0.00000002 0.00001483 layer.4.conv_state 0.00021972 0.33863175 layer.4.output 0.00000203 0.00084704 ------------------------------------------------------------------------------------- TOTAL 0.00002571 0.04070380 (elements=1,470,464) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1470464 Total Bytes 229092 BPFP 1.2464 bits/point EBPFP 1.9705 equivalent bits/point MSE 0.040704 ---------------------- -------------------------------------------------------- Time: 4.868s Load: 0.009s, Pack+Encode: 2.553s, Decode+Unpack: 2.306s ---------------------- -------------------------------------------------------- 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 0.0407 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample190-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,852B, BPFP=1.3337 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,800B, BPFP=0.4761 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,860B, BPFP=2.6514 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,932B, BPFP=1.2776 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,232B, BPFP=2.9863 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,628B, BPFP=0.9539 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 12,036B, BPFP=2.9385 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,288B, BPFP=0.6890 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,428B, BPFP=2.5459 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 141,904B, BPFP=0.8941 ⌛️ [2/4] FRONTEND: Frontend time: 2.662s (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: 2.398s [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 0.00000566 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000012 0.00013080 layer.1.conv_state 0.00048398 0.84486341 layer.2.ssm_state 0.00000001 0.00000871 layer.2.conv_state 0.00013032 0.17257532 layer.3.ssm_state 0.00000001 0.00001046 layer.3.conv_state 0.00007314 0.16240801 layer.4.ssm_state 0.00000002 0.00001550 layer.4.conv_state 0.00041555 0.35530719 layer.4.output 0.00000121 0.00051824 ------------------------------------------------------------------------------------- TOTAL 0.00002034 0.02902005 (elements=2,088,960) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2088960 Total Bytes 274728 BPFP 1.0521 bits/point EBPFP 1.5608 equivalent bits/point MSE 0.029020 ---------------------- -------------------------------------------------------- Time: 5.073s Load: 0.013s, Pack+Encode: 2.662s, Decode+Unpack: 2.398s ---------------------- -------------------------------------------------------- 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 0.0290 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample2-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 21,844B, BPFP=1.3333 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,796B, BPFP=0.4758 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,848B, BPFP=2.6484 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,684B, BPFP=1.2625 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,192B, BPFP=2.9766 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,344B, BPFP=0.9365 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,884B, BPFP=2.9014 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,364B, BPFP=0.6936 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,376B, BPFP=2.5332 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 128,888B, BPFP=1.2976 ⌛️ [2/4] FRONTEND: Frontend time: 2.610s (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: 2.361s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00012839 layer.1.conv_state 0.00048732 0.84873664 layer.2.ssm_state 0.00000001 0.00000869 layer.2.conv_state 0.00011647 0.17146097 layer.3.ssm_state 0.00000001 0.00001043 layer.3.conv_state 0.00011908 0.15895545 layer.4.ssm_state 0.00000001 0.00001507 layer.4.conv_state 0.00022797 0.33992204 layer.4.output 0.00000187 0.00073892 ------------------------------------------------------------------------------------- TOTAL 0.00002322 0.03719339 (elements=1,613,824) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1613824 Total Bytes 260988 BPFP 1.2938 bits/point EBPFP 1.9486 equivalent bits/point MSE 0.037193 ---------------------- -------------------------------------------------------- Time: 4.980s Load: 0.009s, Pack+Encode: 2.610s, Decode+Unpack: 2.361s ---------------------- -------------------------------------------------------- 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 0.0372 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample20-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 21,936B, BPFP=1.3389 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,808B, BPFP=0.4766 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,828B, BPFP=2.6436 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,968B, BPFP=1.2798 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,244B, BPFP=2.9893 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,392B, BPFP=0.9395 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,952B, BPFP=2.9180 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,900B, BPFP=0.7263 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,364B, BPFP=2.5303 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,120B, BPFP=1.0915 ⌛️ [2/4] FRONTEND: Frontend time: 2.551s (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: 2.298s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00013128 layer.1.conv_state 0.00049938 0.83934385 layer.2.ssm_state 0.00000001 0.00000877 layer.2.conv_state 0.00013201 0.17125537 layer.3.ssm_state 0.00000001 0.00001054 layer.3.conv_state 0.00007471 0.16100265 layer.4.ssm_state 0.00000001 0.00001535 layer.4.conv_state 0.00021358 0.34614697 layer.4.output 0.00000218 0.00078297 ------------------------------------------------------------------------------------- TOTAL 0.00002396 0.03934144 (elements=1,523,712) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1523712 Total Bytes 229280 BPFP 1.2038 bits/point EBPFP 1.9029 equivalent bits/point MSE 0.039341 ---------------------- -------------------------------------------------------- Time: 4.858s Load: 0.009s, Pack+Encode: 2.551s, Decode+Unpack: 2.298s ---------------------- -------------------------------------------------------- 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 0.0393 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample204-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,956B, BPFP=1.3401 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,844B, BPFP=0.4788 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,872B, BPFP=2.6543 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,028B, BPFP=1.2834 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,316B, BPFP=3.0068 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,540B, BPFP=0.9485 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,936B, BPFP=2.9141 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,284B, BPFP=0.6887 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,340B, BPFP=2.5244 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 99,156B, BPFP=1.1528 ⌛️ [2/4] FRONTEND: Frontend time: 2.542s (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: 2.298s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000036 0.00013011 layer.1.conv_state 0.00050661 0.84627473 layer.2.ssm_state 0.00000001 0.00000890 layer.2.conv_state 0.00012774 0.17178239 layer.3.ssm_state 0.00000001 0.00001040 layer.3.conv_state 0.00007574 0.15991183 layer.4.ssm_state 0.00000011 0.00001636 layer.4.conv_state 0.00082926 0.36553425 layer.4.output 0.00000222 0.00082520 ------------------------------------------------------------------------------------- TOTAL 0.00003770 0.04033969 (elements=1,507,328) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1507328 Total Bytes 232040 BPFP 1.2315 bits/point EBPFP 1.9368 equivalent bits/point MSE 0.040340 ---------------------- -------------------------------------------------------- Time: 4.848s Load: 0.009s, Pack+Encode: 2.542s, Decode+Unpack: 2.298s ---------------------- -------------------------------------------------------- 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 0.0403 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample21-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,892B, BPFP=1.3362 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,824B, BPFP=0.4775 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,848B, BPFP=2.6484 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,892B, BPFP=1.2751 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,248B, BPFP=2.9902 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,572B, BPFP=0.9504 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,892B, BPFP=2.9033 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,544B, BPFP=0.7046 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,324B, BPFP=2.5205 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 93,152B, BPFP=1.1371 ⌛️ [2/4] FRONTEND: Frontend time: 2.550s (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: 2.289s [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 0.00000558 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000030 0.00012924 layer.1.conv_state 0.00048992 0.84710884 layer.2.ssm_state 0.00000001 0.00000871 layer.2.conv_state 0.00014897 0.17326421 layer.3.ssm_state 0.00000001 0.00001045 layer.3.conv_state 0.00007436 0.16073813 layer.4.ssm_state 0.00000001 0.00001543 layer.4.conv_state 0.00026192 0.34363788 layer.4.output 0.00000216 0.00083516 ------------------------------------------------------------------------------------- TOTAL 0.00002591 0.04080528 (elements=1,474,560) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1474560 Total Bytes 225956 BPFP 1.2259 bits/point EBPFP 1.9464 equivalent bits/point MSE 0.040805 ---------------------- -------------------------------------------------------- Time: 4.848s Load: 0.009s, Pack+Encode: 2.550s, Decode+Unpack: 2.289s ---------------------- -------------------------------------------------------- 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 0.0408 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample218-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 21,836B, BPFP=1.3328 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,828B, BPFP=0.4778 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,904B, BPFP=2.6621 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,812B, BPFP=1.2703 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,228B, BPFP=2.9854 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,484B, BPFP=0.9451 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,872B, BPFP=2.8984 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,272B, BPFP=0.6880 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,320B, BPFP=2.5195 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 95,820B, BPFP=1.0171 ⌛️ [2/4] FRONTEND: Frontend time: 2.542s (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: 2.292s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000010 0.00012970 layer.1.conv_state 0.00050597 0.84874773 layer.2.ssm_state 0.00000001 0.00000875 layer.2.conv_state 0.00012948 0.17206131 layer.3.ssm_state 0.00000001 0.00001022 layer.3.conv_state 0.00007506 0.15950210 layer.4.ssm_state 0.00000001 0.00001531 layer.4.conv_state 0.00035639 0.34947336 layer.4.output 0.00000208 0.00070895 ------------------------------------------------------------------------------------- TOTAL 0.00002628 0.03835159 (elements=1,572,864) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1572864 Total Bytes 228144 BPFP 1.1604 bits/point EBPFP 1.8334 equivalent bits/point MSE 0.038352 ---------------------- -------------------------------------------------------- Time: 4.844s Load: 0.009s, Pack+Encode: 2.542s, Decode+Unpack: 2.292s ---------------------- -------------------------------------------------------- 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 0.0384 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample22-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 21,984B, BPFP=1.3418 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,824B, BPFP=0.4775 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,864B, BPFP=2.6523 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,764B, BPFP=1.2673 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,260B, BPFP=2.9932 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,300B, BPFP=0.9338 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,936B, BPFP=2.9141 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,636B, BPFP=0.7102 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,236B, BPFP=2.4990 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,520B, BPFP=1.0301 ⌛️ [2/4] FRONTEND: Frontend time: 2.544s (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: 2.291s [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 0.00000558 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000010 0.00013061 layer.1.conv_state 0.00049328 0.84409952 layer.2.ssm_state 0.00000001 0.00000884 layer.2.conv_state 0.00013895 0.17278379 layer.3.ssm_state 0.00000001 0.00001009 layer.3.conv_state 0.00011983 0.15971518 layer.4.ssm_state 0.00000001 0.00001455 layer.4.conv_state 0.00020516 0.33866134 layer.4.output 0.00000202 0.00069879 ------------------------------------------------------------------------------------- TOTAL 0.00002402 0.03814164 (elements=1,568,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1568768 Total Bytes 229092 BPFP 1.1683 bits/point EBPFP 1.8443 equivalent bits/point MSE 0.038142 ---------------------- -------------------------------------------------------- Time: 4.844s Load: 0.009s, Pack+Encode: 2.544s, Decode+Unpack: 2.291s ---------------------- -------------------------------------------------------- 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 0.0381 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample23-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,040B, BPFP=1.3452 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,808B, BPFP=0.4766 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,860B, BPFP=2.6514 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,852B, BPFP=1.2727 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,232B, BPFP=2.9863 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,132B, BPFP=0.9236 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,956B, BPFP=2.9189 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,676B, BPFP=0.7126 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,360B, BPFP=2.5293 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 95,096B, BPFP=1.1325 ⌛️ [2/4] FRONTEND: Frontend time: 2.544s (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: 2.300s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00012982 layer.1.conv_state 0.00051090 0.84185892 layer.2.ssm_state 0.00000001 0.00000876 layer.2.conv_state 0.00012582 0.17194721 layer.3.ssm_state 0.00000001 0.00001044 layer.3.conv_state 0.00007785 0.16091298 layer.4.ssm_state 0.00000001 0.00001549 layer.4.conv_state 0.00021500 0.34894744 layer.4.output 0.00000225 0.00079574 ------------------------------------------------------------------------------------- TOTAL 0.00002468 0.04032456 (elements=1,490,944) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1490944 Total Bytes 227780 BPFP 1.2222 bits/point EBPFP 1.9342 equivalent bits/point MSE 0.040325 ---------------------- -------------------------------------------------------- Time: 4.854s Load: 0.009s, Pack+Encode: 2.544s, Decode+Unpack: 2.300s ---------------------- -------------------------------------------------------- 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 0.0403 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample234-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,000B, BPFP=1.3428 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,796B, BPFP=0.4758 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,848B, BPFP=2.6484 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,840B, BPFP=1.2720 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,196B, BPFP=2.9775 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,260B, BPFP=0.9314 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,884B, BPFP=2.9014 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,612B, BPFP=0.7087 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,296B, BPFP=2.5137 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 120,584B, BPFP=1.2016 ⌛️ [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, 196, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.362s [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 0.00000565 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00013006 layer.1.conv_state 0.00048758 0.84496796 layer.2.ssm_state 0.00000001 0.00000877 layer.2.conv_state 0.00011566 0.17115606 layer.3.ssm_state 0.00000001 0.00001028 layer.3.conv_state 0.00011820 0.15908396 layer.4.ssm_state 0.00000001 0.00001531 layer.4.conv_state 0.00025975 0.34994486 layer.4.output 0.00000193 0.00071943 ------------------------------------------------------------------------------------- TOTAL 0.00002375 0.03712256 (elements=1,622,016) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1622016 Total Bytes 253084 BPFP 1.2482 bits/point EBPFP 1.9018 equivalent bits/point MSE 0.037123 ---------------------- -------------------------------------------------------- Time: 4.976s Load: 0.010s, Pack+Encode: 2.604s, Decode+Unpack: 2.362s ---------------------- -------------------------------------------------------- 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 0.0371 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample24-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 21,852B, BPFP=1.3337 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,784B, BPFP=0.4751 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,840B, BPFP=2.6465 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,592B, BPFP=1.2568 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,208B, BPFP=2.9805 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,128B, BPFP=0.9233 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,896B, BPFP=2.9043 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,636B, BPFP=0.7102 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,296B, BPFP=2.5137 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 124,964B, BPFP=1.2083 ⌛️ [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, 202, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.351s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000011 0.00012898 layer.1.conv_state 0.00050439 0.83870178 layer.2.ssm_state 0.00000001 0.00000870 layer.2.conv_state 0.00014180 0.16960312 layer.3.ssm_state 0.00000001 0.00001025 layer.3.conv_state 0.00006834 0.15724429 layer.4.ssm_state 0.00000001 0.00001464 layer.4.conv_state 0.00020995 0.33512604 layer.4.output 0.00000183 0.00071410 ------------------------------------------------------------------------------------- TOTAL 0.00002224 0.03608929 (elements=1,646,592) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1646592 Total Bytes 256964 BPFP 1.2485 bits/point EBPFP 1.8898 equivalent bits/point MSE 0.036089 ---------------------- -------------------------------------------------------- Time: 4.961s Load: 0.011s, Pack+Encode: 2.600s, Decode+Unpack: 2.351s ---------------------- -------------------------------------------------------- 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 0.0361 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample25-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,064B, BPFP=1.3467 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,816B, BPFP=0.4771 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,868B, BPFP=2.6533 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,940B, BPFP=1.2781 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,216B, BPFP=2.9824 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,676B, BPFP=0.9568 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,884B, BPFP=2.9014 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,592B, BPFP=0.7075 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,328B, BPFP=2.5215 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 130,540B, BPFP=1.3008 ⌛️ [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, 196, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.354s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00012992 layer.1.conv_state 0.00047314 0.84331483 layer.2.ssm_state 0.00000001 0.00000879 layer.2.conv_state 0.00012900 0.17163226 layer.3.ssm_state 0.00000001 0.00001038 layer.3.conv_state 0.00011898 0.15970081 layer.4.ssm_state 0.00000002 0.00001513 layer.4.conv_state 0.00022845 0.34075135 layer.4.output 0.00000185 0.00072636 ------------------------------------------------------------------------------------- TOTAL 0.00002308 0.03692893 (elements=1,622,016) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1622016 Total Bytes 263692 BPFP 1.3006 bits/point EBPFP 1.9573 equivalent bits/point MSE 0.036929 ---------------------- -------------------------------------------------------- Time: 4.964s Load: 0.010s, Pack+Encode: 2.601s, Decode+Unpack: 2.354s ---------------------- -------------------------------------------------------- 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 0.0369 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample26-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,008B, BPFP=1.3433 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,748B, BPFP=0.4729 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,840B, BPFP=2.6465 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,844B, BPFP=1.2722 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,224B, BPFP=2.9844 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,156B, BPFP=0.9250 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,920B, BPFP=2.9102 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,208B, BPFP=0.7451 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,376B, BPFP=2.5332 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,720B, BPFP=1.1783 ⌛️ [2/4] FRONTEND: Frontend time: 2.548s (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: 2.306s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000030 0.00013120 layer.1.conv_state 0.00049487 0.84128451 layer.2.ssm_state 0.00000001 0.00000871 layer.2.conv_state 0.00012267 0.17191899 layer.3.ssm_state 0.00000001 0.00001035 layer.3.conv_state 0.00011738 0.16042624 layer.4.ssm_state 0.00000002 0.00001501 layer.4.conv_state 0.00024516 0.34167796 layer.4.output 0.00000212 0.00080648 ------------------------------------------------------------------------------------- TOTAL 0.00002620 0.04091711 (elements=1,462,272) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1462272 Total Bytes 227812 BPFP 1.2463 bits/point EBPFP 1.9745 equivalent bits/point MSE 0.040917 ---------------------- -------------------------------------------------------- Time: 4.862s Load: 0.009s, Pack+Encode: 2.548s, Decode+Unpack: 2.306s ---------------------- -------------------------------------------------------- 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 0.0409 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample266-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 22,040B, BPFP=1.3452 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,788B, BPFP=0.4753 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,816B, BPFP=1.2705 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,244B, BPFP=2.9893 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,300B, BPFP=0.9338 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,912B, BPFP=2.9082 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,260B, BPFP=0.6873 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,300B, BPFP=2.5146 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 127,792B, BPFP=1.2800 ⌛️ [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, 195, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.358s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000009 0.00013144 layer.1.conv_state 0.00049297 0.84304500 layer.2.ssm_state 0.00000001 0.00000874 layer.2.conv_state 0.00013859 0.17163257 layer.3.ssm_state 0.00000001 0.00001041 layer.3.conv_state 0.00007636 0.15996298 layer.4.ssm_state 0.00000001 0.00001501 layer.4.conv_state 0.00030466 0.34577698 layer.4.output 0.00000195 0.00073240 ------------------------------------------------------------------------------------- TOTAL 0.00002444 0.03712531 (elements=1,617,920) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1617920 Total Bytes 260072 BPFP 1.2860 bits/point EBPFP 1.9400 equivalent bits/point MSE 0.037125 ---------------------- -------------------------------------------------------- Time: 4.974s Load: 0.011s, Pack+Encode: 2.605s, Decode+Unpack: 2.358s ---------------------- -------------------------------------------------------- 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 0.0371 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample27-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 21,956B, BPFP=1.3401 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,788B, BPFP=0.4753 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,860B, BPFP=2.6514 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,840B, BPFP=1.2720 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,256B, BPFP=2.9922 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,464B, BPFP=0.9438 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,908B, BPFP=2.9072 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,000B, BPFP=0.7324 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,324B, BPFP=2.5205 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 76,476B, BPFP=0.9699 ⌛️ [2/4] FRONTEND: Frontend time: 2.546s (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: 2.296s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000039 0.00012794 layer.1.conv_state 0.00049616 0.84405005 layer.2.ssm_state 0.00000001 0.00000885 layer.2.conv_state 0.00014030 0.17248456 layer.3.ssm_state 0.00000001 0.00001066 layer.3.conv_state 0.00011494 0.16090943 layer.4.ssm_state 0.00000003 0.00001478 layer.4.conv_state 0.00023159 0.33622885 layer.4.output 0.00000204 0.00081493 ------------------------------------------------------------------------------------- TOTAL 0.00002644 0.04122349 (elements=1,449,984) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1449984 Total Bytes 209640 BPFP 1.1566 bits/point EBPFP 1.8914 equivalent bits/point MSE 0.041223 ---------------------- -------------------------------------------------------- Time: 4.853s Load: 0.011s, Pack+Encode: 2.546s, Decode+Unpack: 2.296s ---------------------- -------------------------------------------------------- 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 0.0412 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample283-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.013s ------------------------------------------------------------ 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: 21,916B, BPFP=1.3376 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,792B, BPFP=0.4756 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,884B, BPFP=2.6572 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,996B, BPFP=1.2815 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,220B, BPFP=2.9834 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,748B, BPFP=0.9612 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,864B, BPFP=2.8965 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,896B, BPFP=0.7261 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,388B, BPFP=2.5361 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 92,384B, BPFP=0.9701 ⌛️ [2/4] FRONTEND: Frontend time: 2.535s (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: 2.290s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00013075 layer.1.conv_state 0.00049578 0.84730458 layer.2.ssm_state 0.00000001 0.00000883 layer.2.conv_state 0.00013264 0.17239460 layer.3.ssm_state 0.00000001 0.00001034 layer.3.conv_state 0.00012070 0.16118379 layer.4.ssm_state 0.00000002 0.00001468 layer.4.conv_state 0.00022222 0.33498052 layer.4.output 0.00000189 0.00071975 ------------------------------------------------------------------------------------- TOTAL 0.00002409 0.03787329 (elements=1,581,056) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1581056 Total Bytes 225856 BPFP 1.1428 bits/point EBPFP 1.8182 equivalent bits/point MSE 0.037873 ---------------------- -------------------------------------------------------- Time: 4.839s Load: 0.013s, Pack+Encode: 2.535s, Decode+Unpack: 2.290s ---------------------- -------------------------------------------------------- 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 0.0379 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample29-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.013s ------------------------------------------------------------ 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: 21,984B, BPFP=1.3418 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,812B, BPFP=0.4768 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,868B, BPFP=2.6533 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,900B, BPFP=1.2756 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,212B, BPFP=2.9814 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,556B, BPFP=0.9495 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,896B, BPFP=2.9043 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,272B, BPFP=0.6880 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,364B, BPFP=2.5303 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 112,800B, BPFP=0.7982 ⌛️ [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, 276, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.402s [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 0.00000565 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000027 0.00012942 layer.1.conv_state 0.00048777 0.84477597 layer.2.ssm_state 0.00000001 0.00000879 layer.2.conv_state 0.00011906 0.16935772 layer.3.ssm_state 0.00000001 0.00001034 layer.3.conv_state 0.00011276 0.15854403 layer.4.ssm_state 0.00000006 0.00001595 layer.4.conv_state 0.00059394 0.36294240 layer.4.output 0.00000125 0.00058490 ------------------------------------------------------------------------------------- TOTAL 0.00002528 0.03110231 (elements=1,949,696) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1949696 Total Bytes 245432 BPFP 1.0071 bits/point EBPFP 1.5513 equivalent bits/point MSE 0.031102 ---------------------- -------------------------------------------------------- Time: 5.069s Load: 0.013s, Pack+Encode: 2.654s, Decode+Unpack: 2.402s ---------------------- -------------------------------------------------------- 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 0.0311 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample3-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,972B, BPFP=1.3411 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,784B, BPFP=0.4751 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,824B, BPFP=2.6426 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,704B, BPFP=1.2637 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,208B, BPFP=2.9805 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,388B, BPFP=0.9392 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,872B, BPFP=2.8984 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,452B, BPFP=0.7600 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,248B, BPFP=2.5020 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 154,140B, BPFP=1.5205 ⌛️ [2/4] FRONTEND: Frontend time: 2.597s (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: 2.350s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00012872 layer.1.conv_state 0.00048081 0.84585744 layer.2.ssm_state 0.00000001 0.00000882 layer.2.conv_state 0.00011915 0.17158660 layer.3.ssm_state 0.00000001 0.00001044 layer.3.conv_state 0.00011359 0.16060548 layer.4.ssm_state 0.00000002 0.00001413 layer.4.conv_state 0.00023406 0.33056027 layer.4.output 0.00000200 0.00078845 ------------------------------------------------------------------------------------- TOTAL 0.00002301 0.03664126 (elements=1,630,208) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1630208 Total Bytes 287360 BPFP 1.4102 bits/point EBPFP 2.0639 equivalent bits/point MSE 0.036641 ---------------------- -------------------------------------------------------- Time: 4.957s Load: 0.010s, Pack+Encode: 2.597s, Decode+Unpack: 2.350s ---------------------- -------------------------------------------------------- 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 0.0366 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample30-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,948B, BPFP=1.3396 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,816B, BPFP=0.4771 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,884B, BPFP=2.6572 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,672B, BPFP=1.2617 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,308B, BPFP=3.0049 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,228B, BPFP=0.9294 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,976B, BPFP=2.9238 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,512B, BPFP=0.7026 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,364B, BPFP=2.5303 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,536B, BPFP=1.0774 ⌛️ [2/4] FRONTEND: Frontend time: 2.552s (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: 2.292s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00013123 layer.1.conv_state 0.00049445 0.84436053 layer.2.ssm_state 0.00000001 0.00000882 layer.2.conv_state 0.00015634 0.17226338 layer.3.ssm_state 0.00000001 0.00001022 layer.3.conv_state 0.00006751 0.16010496 layer.4.ssm_state 0.00000001 0.00001498 layer.4.conv_state 0.00021972 0.34114933 layer.4.output 0.00000206 0.00078105 ------------------------------------------------------------------------------------- TOTAL 0.00002411 0.03903477 (elements=1,536,000) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1536000 Total Bytes 229012 BPFP 1.1928 bits/point EBPFP 1.8827 equivalent bits/point MSE 0.039035 ---------------------- -------------------------------------------------------- Time: 4.855s Load: 0.011s, Pack+Encode: 2.552s, Decode+Unpack: 2.292s ---------------------- -------------------------------------------------------- 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 0.0390 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample31-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.014s ------------------------------------------------------------ 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: 22,044B, BPFP=1.3455 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,776B, BPFP=0.4746 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,824B, BPFP=2.6426 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,924B, BPFP=1.2771 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,208B, BPFP=2.9805 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,540B, BPFP=0.9485 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,848B, BPFP=2.8926 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,888B, BPFP=0.7256 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,272B, BPFP=2.5078 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 122,672B, BPFP=1.0649 ⌛️ [2/4] FRONTEND: Frontend time: 2.595s (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: 2.350s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000027 0.00012907 layer.1.conv_state 0.00049817 0.84286201 layer.2.ssm_state 0.00000001 0.00000889 layer.2.conv_state 0.00012605 0.17126004 layer.3.ssm_state 0.00000001 0.00001058 layer.3.conv_state 0.00007245 0.16117409 layer.4.ssm_state 0.00000001 0.00001505 layer.4.conv_state 0.00021821 0.34290862 layer.4.output 0.00000156 0.00066881 ------------------------------------------------------------------------------------- TOTAL 0.00002082 0.03448093 (elements=1,740,800) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1740800 Total Bytes 255764 BPFP 1.1754 bits/point EBPFP 1.7870 equivalent bits/point MSE 0.034481 ---------------------- -------------------------------------------------------- Time: 4.958s Load: 0.014s, Pack+Encode: 2.595s, Decode+Unpack: 2.350s ---------------------- -------------------------------------------------------- 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 0.0345 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample32-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.012s ------------------------------------------------------------ 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: 21,884B, BPFP=1.3357 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,780B, BPFP=0.4749 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,824B, BPFP=2.6426 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,936B, BPFP=1.2778 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,216B, BPFP=2.9824 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,328B, BPFP=0.9355 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,916B, BPFP=2.9092 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,812B, BPFP=0.7209 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,304B, BPFP=2.5156 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 97,208B, BPFP=1.2491 ⌛️ [2/4] FRONTEND: Frontend time: 2.545s (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: 2.291s [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 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00013016 layer.1.conv_state 0.00048212 0.84465116 layer.2.ssm_state 0.00000001 0.00000884 layer.2.conv_state 0.00013258 0.17162699 layer.3.ssm_state 0.00000001 0.00001049 layer.3.conv_state 0.00007465 0.16034773 layer.4.ssm_state 0.00000001 0.00001494 layer.4.conv_state 0.00020440 0.33977291 layer.4.output 0.00000217 0.00083552 ------------------------------------------------------------------------------------- TOTAL 0.00002460 0.04152413 (elements=1,441,792) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1441792 Total Bytes 229976 BPFP 1.2761 bits/point EBPFP 2.0127 equivalent bits/point MSE 0.041524 ---------------------- -------------------------------------------------------- Time: 4.847s Load: 0.012s, Pack+Encode: 2.545s, Decode+Unpack: 2.291s ---------------------- -------------------------------------------------------- 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 0.0415 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample323-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.014s ------------------------------------------------------------ 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: 22,016B, BPFP=1.3438 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,804B, BPFP=0.4763 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,860B, BPFP=2.6514 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,904B, BPFP=1.2759 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,208B, BPFP=2.9805 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,364B, BPFP=0.9377 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,868B, BPFP=2.8975 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,056B, BPFP=0.7358 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,324B, BPFP=2.5205 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 112,864B, BPFP=1.0753 ⌛️ [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, 205, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.353s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000024 0.00012875 layer.1.conv_state 0.00049079 0.83725077 layer.2.ssm_state 0.00000001 0.00000876 layer.2.conv_state 0.00014720 0.16981900 layer.3.ssm_state 0.00000001 0.00001038 layer.3.conv_state 0.00011227 0.15795705 layer.4.ssm_state 0.00000001 0.00001507 layer.4.conv_state 0.00021844 0.34800154 layer.4.output 0.00000170 0.00067279 ------------------------------------------------------------------------------------- TOTAL 0.00002290 0.03605038 (elements=1,658,880) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1658880 Total Bytes 246036 BPFP 1.1865 bits/point EBPFP 1.8287 equivalent bits/point MSE 0.036050 ---------------------- -------------------------------------------------------- Time: 4.957s Load: 0.014s, Pack+Encode: 2.591s, Decode+Unpack: 2.353s ---------------------- -------------------------------------------------------- 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 0.0361 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample33-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 21,968B, BPFP=1.3408 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,772B, BPFP=0.4744 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,836B, BPFP=2.6455 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,852B, BPFP=1.2727 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,240B, BPFP=2.9883 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,544B, BPFP=0.9487 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,928B, BPFP=2.9121 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,284B, BPFP=0.6887 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,448B, BPFP=2.5508 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 115,948B, BPFP=1.0155 ⌛️ [2/4] FRONTEND: Frontend time: 2.597s (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: 2.348s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00012859 layer.1.conv_state 0.00048351 0.84390962 layer.2.ssm_state 0.00000001 0.00000866 layer.2.conv_state 0.00011966 0.17011400 layer.3.ssm_state 0.00000001 0.00001043 layer.3.conv_state 0.00006977 0.16009250 layer.4.ssm_state 0.00000002 0.00001577 layer.4.conv_state 0.00033313 0.35627204 layer.4.output 0.00000142 0.00065211 ------------------------------------------------------------------------------------- TOTAL 0.00002256 0.03486241 (elements=1,732,608) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1732608 Total Bytes 248588 BPFP 1.1478 bits/point EBPFP 1.7603 equivalent bits/point MSE 0.034862 ---------------------- -------------------------------------------------------- Time: 4.956s Load: 0.011s, Pack+Encode: 2.597s, Decode+Unpack: 2.348s ---------------------- -------------------------------------------------------- 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 0.0349 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample34-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.012s ------------------------------------------------------------ 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: 22,036B, BPFP=1.3450 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,808B, BPFP=0.4766 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,828B, BPFP=2.6436 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,140B, BPFP=1.2903 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,248B, BPFP=2.9902 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,460B, BPFP=0.9436 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,928B, BPFP=2.9121 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,680B, BPFP=0.7129 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,340B, BPFP=2.5244 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 88,552B, BPFP=0.9451 ⌛️ [2/4] FRONTEND: Frontend time: 2.538s (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: 2.291s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000021 0.00012824 layer.1.conv_state 0.00049749 0.84385228 layer.2.ssm_state 0.00000001 0.00000876 layer.2.conv_state 0.00012539 0.17098008 layer.3.ssm_state 0.00000001 0.00001042 layer.3.conv_state 0.00007456 0.15999733 layer.4.ssm_state 0.00000001 0.00001570 layer.4.conv_state 0.00032004 0.35146722 layer.4.output 0.00000203 0.00074861 ------------------------------------------------------------------------------------- TOTAL 0.00002530 0.03839591 (elements=1,568,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1568768 Total Bytes 221788 BPFP 1.1310 bits/point EBPFP 1.8105 equivalent bits/point MSE 0.038396 ---------------------- -------------------------------------------------------- Time: 4.841s Load: 0.012s, Pack+Encode: 2.538s, Decode+Unpack: 2.291s ---------------------- -------------------------------------------------------- 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 0.0384 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample35-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 22,024B, BPFP=1.3442 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,868B, BPFP=0.4802 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,828B, BPFP=2.6436 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,888B, BPFP=1.2749 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,244B, BPFP=2.9893 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,600B, BPFP=0.9521 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,952B, BPFP=2.9180 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,540B, BPFP=0.7043 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,376B, BPFP=2.5332 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,712B, BPFP=1.0436 ⌛️ [2/4] FRONTEND: Frontend time: 2.531s (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: 2.294s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000013 0.00013073 layer.1.conv_state 0.00048788 0.84586918 layer.2.ssm_state 0.00000001 0.00000880 layer.2.conv_state 0.00012276 0.17204589 layer.3.ssm_state 0.00000001 0.00001065 layer.3.conv_state 0.00007521 0.16177970 layer.4.ssm_state 0.00000001 0.00001545 layer.4.conv_state 0.00025615 0.35018080 layer.4.output 0.00000200 0.00071375 ------------------------------------------------------------------------------------- TOTAL 0.00002381 0.03865232 (elements=1,560,576) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1560576 Total Bytes 229800 BPFP 1.1780 bits/point EBPFP 1.8603 equivalent bits/point MSE 0.038652 ---------------------- -------------------------------------------------------- Time: 4.834s Load: 0.010s, Pack+Encode: 2.531s, Decode+Unpack: 2.294s ---------------------- -------------------------------------------------------- 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 0.0387 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample36-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 22,012B, BPFP=1.3435 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,808B, BPFP=0.4766 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,848B, BPFP=2.6484 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,932B, BPFP=1.2776 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,204B, BPFP=2.9795 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,560B, BPFP=0.9497 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,868B, BPFP=2.8975 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,208B, BPFP=0.6841 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,236B, BPFP=2.4990 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 134,028B, BPFP=1.2832 ⌛️ [2/4] FRONTEND: Frontend time: 2.593s (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: 2.355s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00012882 layer.1.conv_state 0.00048573 0.84659791 layer.2.ssm_state 0.00000001 0.00000883 layer.2.conv_state 0.00014000 0.17184198 layer.3.ssm_state 0.00000001 0.00001058 layer.3.conv_state 0.00007101 0.16031089 layer.4.ssm_state 0.00000002 0.00001568 layer.4.conv_state 0.00054851 0.35493839 layer.4.output 0.00000173 0.00073993 ------------------------------------------------------------------------------------- TOTAL 0.00002845 0.03658106 (elements=1,654,784) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1654784 Total Bytes 266472 BPFP 1.2883 bits/point EBPFP 1.9285 equivalent bits/point MSE 0.036581 ---------------------- -------------------------------------------------------- Time: 4.959s Load: 0.011s, Pack+Encode: 2.593s, Decode+Unpack: 2.355s ---------------------- -------------------------------------------------------- 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 0.0366 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample37-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 22,120B, BPFP=1.3501 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,756B, BPFP=0.4734 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,888B, BPFP=1.2749 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,196B, BPFP=2.9775 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,596B, BPFP=0.9519 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,908B, BPFP=2.9072 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,692B, BPFP=0.7136 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,280B, BPFP=2.5098 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 88,776B, BPFP=0.9174 ⌛️ [2/4] FRONTEND: Frontend time: 2.534s (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: 2.285s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000013 0.00013016 layer.1.conv_state 0.00048943 0.84482002 layer.2.ssm_state 0.00000001 0.00000874 layer.2.conv_state 0.00011993 0.17165354 layer.3.ssm_state 0.00000001 0.00001048 layer.3.conv_state 0.00011634 0.15927786 layer.4.ssm_state 0.00000001 0.00001498 layer.4.conv_state 0.00025516 0.34417903 layer.4.output 0.00000196 0.00075210 ------------------------------------------------------------------------------------- TOTAL 0.00002414 0.03768610 (elements=1,593,344) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1593344 Total Bytes 221832 BPFP 1.1138 bits/point EBPFP 1.7819 equivalent bits/point MSE 0.037686 ---------------------- -------------------------------------------------------- Time: 4.828s Load: 0.010s, Pack+Encode: 2.534s, Decode+Unpack: 2.285s ---------------------- -------------------------------------------------------- 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 0.0377 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample38-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.012s ------------------------------------------------------------ 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: 22,028B, BPFP=1.3445 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,832B, BPFP=0.4780 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,732B, BPFP=1.2654 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,248B, BPFP=2.9902 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,420B, BPFP=0.9412 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,916B, BPFP=2.9092 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,128B, BPFP=0.7402 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,188B, BPFP=2.4873 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 93,464B, BPFP=0.9710 ⌛️ [2/4] FRONTEND: Frontend time: 2.533s (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: 2.296s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000027 0.00012983 layer.1.conv_state 0.00048897 0.85100818 layer.2.ssm_state 0.00000001 0.00000877 layer.2.conv_state 0.00011975 0.17127629 layer.3.ssm_state 0.00000001 0.00001030 layer.3.conv_state 0.00006970 0.16040142 layer.4.ssm_state 0.00000001 0.00001436 layer.4.conv_state 0.00027957 0.33502698 layer.4.output 0.00000196 0.00074327 ------------------------------------------------------------------------------------- TOTAL 0.00002374 0.03773119 (elements=1,589,248) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1589248 Total Bytes 226576 BPFP 1.1405 bits/point EBPFP 1.8106 equivalent bits/point MSE 0.037731 ---------------------- -------------------------------------------------------- Time: 4.842s Load: 0.012s, Pack+Encode: 2.533s, Decode+Unpack: 2.296s ---------------------- -------------------------------------------------------- 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 0.0377 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample39-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.015s ------------------------------------------------------------ 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: 22,052B, BPFP=1.3459 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,816B, BPFP=0.4771 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,880B, BPFP=2.6562 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,868B, BPFP=1.2737 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,220B, BPFP=2.9834 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,788B, BPFP=0.9636 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,840B, BPFP=2.8906 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,788B, BPFP=0.7195 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,232B, BPFP=2.4980 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 152,512B, BPFP=1.0992 ⌛️ [2/4] FRONTEND: Frontend time: 2.650s (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: 2.397s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00012971 layer.1.conv_state 0.00049489 0.84921634 layer.2.ssm_state 0.00000001 0.00000872 layer.2.conv_state 0.00013525 0.17114787 layer.3.ssm_state 0.00000001 0.00001048 layer.3.conv_state 0.00011705 0.15976165 layer.4.ssm_state 0.00000001 0.00001495 layer.4.conv_state 0.00022610 0.34288204 layer.4.output 0.00000138 0.00058453 ------------------------------------------------------------------------------------- TOTAL 0.00001983 0.03121180 (elements=1,929,216) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1929216 Total Bytes 285764 BPFP 1.1850 bits/point EBPFP 1.7376 equivalent bits/point MSE 0.031212 ---------------------- -------------------------------------------------------- Time: 5.062s Load: 0.015s, Pack+Encode: 2.650s, Decode+Unpack: 2.397s ---------------------- -------------------------------------------------------- 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 0.0312 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample4-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,920B, BPFP=1.3379 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,864B, BPFP=0.4800 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,872B, BPFP=2.6543 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,072B, BPFP=1.2861 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,256B, BPFP=2.9922 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,696B, BPFP=0.9580 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,868B, BPFP=2.8975 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,012B, BPFP=0.7332 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,328B, BPFP=2.5215 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 106,268B, BPFP=1.2354 ⌛️ [2/4] FRONTEND: Frontend time: 2.540s (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: 2.297s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000047 0.00013220 layer.1.conv_state 0.00050445 0.84243727 layer.2.ssm_state 0.00000001 0.00000887 layer.2.conv_state 0.00012625 0.17296535 layer.3.ssm_state 0.00000001 0.00001027 layer.3.conv_state 0.00011525 0.16071501 layer.4.ssm_state 0.00000001 0.00001486 layer.4.conv_state 0.00022432 0.33723181 layer.4.output 0.00000216 0.00084385 ------------------------------------------------------------------------------------- TOTAL 0.00002530 0.03969273 (elements=1,507,328) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1507328 Total Bytes 239924 BPFP 1.2734 bits/point EBPFP 1.9827 equivalent bits/point MSE 0.039693 ---------------------- -------------------------------------------------------- Time: 4.846s Load: 0.009s, Pack+Encode: 2.540s, Decode+Unpack: 2.297s ---------------------- -------------------------------------------------------- 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 0.0397 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample40-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,976B, BPFP=1.3413 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,780B, BPFP=0.4749 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,852B, BPFP=1.2727 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,236B, BPFP=2.9873 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,500B, BPFP=0.9460 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,912B, BPFP=2.9082 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,252B, BPFP=0.6868 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,312B, BPFP=2.5176 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,664B, BPFP=1.0687 ⌛️ [2/4] FRONTEND: Frontend time: 2.534s (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: 2.291s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000013 0.00013065 layer.1.conv_state 0.00048463 0.84386396 layer.2.ssm_state 0.00000001 0.00000885 layer.2.conv_state 0.00012632 0.17144029 layer.3.ssm_state 0.00000001 0.00001069 layer.3.conv_state 0.00007282 0.15971273 layer.4.ssm_state 0.00000003 0.00001585 layer.4.conv_state 0.00058008 0.36492091 layer.4.output 0.00000205 0.00078005 ------------------------------------------------------------------------------------- TOTAL 0.00003121 0.03971262 (elements=1,527,808) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1527808 Total Bytes 227104 BPFP 1.1892 bits/point EBPFP 1.8827 equivalent bits/point MSE 0.039713 ---------------------- -------------------------------------------------------- Time: 4.835s Load: 0.009s, Pack+Encode: 2.534s, Decode+Unpack: 2.291s ---------------------- -------------------------------------------------------- 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 0.0397 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample403-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.013s ------------------------------------------------------------ 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: 21,860B, BPFP=1.3342 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,796B, BPFP=0.4758 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,848B, BPFP=2.6484 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,844B, BPFP=1.2722 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,228B, BPFP=2.9854 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,196B, BPFP=0.9275 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 12,020B, BPFP=2.9346 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,616B, BPFP=0.7090 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,356B, BPFP=2.5283 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 133,488B, BPFP=1.3234 ⌛️ [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, 197, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.355s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00013090 layer.1.conv_state 0.00050234 0.84087092 layer.2.ssm_state 0.00000001 0.00000873 layer.2.conv_state 0.00013726 0.17183389 layer.3.ssm_state 0.00000001 0.00001038 layer.3.conv_state 0.00007731 0.16084789 layer.4.ssm_state 0.00000001 0.00001549 layer.4.conv_state 0.00032633 0.35256419 layer.4.output 0.00000190 0.00073062 ------------------------------------------------------------------------------------- TOTAL 0.00002493 0.03705593 (elements=1,626,112) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1626112 Total Bytes 266020 BPFP 1.3087 bits/point EBPFP 1.9608 equivalent bits/point MSE 0.037056 ---------------------- -------------------------------------------------------- Time: 4.957s Load: 0.013s, Pack+Encode: 2.590s, Decode+Unpack: 2.355s ---------------------- -------------------------------------------------------- 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 0.0371 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample41-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,020B, BPFP=1.3440 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,780B, BPFP=0.4749 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,888B, BPFP=2.6582 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,012B, BPFP=1.2825 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,220B, BPFP=2.9834 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,472B, BPFP=0.9443 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,860B, BPFP=2.8955 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,580B, BPFP=0.7068 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,316B, BPFP=2.5186 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 95,992B, BPFP=1.0134 ⌛️ [2/4] FRONTEND: Frontend time: 2.538s (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: 2.288s [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 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000010 0.00013013 layer.1.conv_state 0.00049834 0.84390104 layer.2.ssm_state 0.00000001 0.00000878 layer.2.conv_state 0.00012928 0.17256075 layer.3.ssm_state 0.00000001 0.00001028 layer.3.conv_state 0.00007819 0.16119117 layer.4.ssm_state 0.00000001 0.00001522 layer.4.conv_state 0.00030927 0.34942269 layer.4.output 0.00000194 0.00070769 ------------------------------------------------------------------------------------- TOTAL 0.00002507 0.03819696 (elements=1,576,960) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1576960 Total Bytes 228908 BPFP 1.1613 bits/point EBPFP 1.8356 equivalent bits/point MSE 0.038197 ---------------------- -------------------------------------------------------- Time: 4.837s Load: 0.010s, Pack+Encode: 2.538s, Decode+Unpack: 2.288s ---------------------- -------------------------------------------------------- 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 0.0382 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample42-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,012B, BPFP=1.3435 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,800B, BPFP=0.4761 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,848B, BPFP=2.6484 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,880B, BPFP=1.2744 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,216B, BPFP=2.9824 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,296B, BPFP=0.9336 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,960B, BPFP=2.9199 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,024B, BPFP=0.7339 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,352B, BPFP=2.5273 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 128,128B, BPFP=1.1917 ⌛️ [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, 210, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.356s [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 0.00000565 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00013061 layer.1.conv_state 0.00049741 0.84395480 layer.2.ssm_state 0.00000001 0.00000872 layer.2.conv_state 0.00015034 0.17230816 layer.3.ssm_state 0.00000001 0.00001040 layer.3.conv_state 0.00007531 0.16112500 layer.4.ssm_state 0.00000004 0.00001521 layer.4.conv_state 0.00020482 0.34664807 layer.4.output 0.00000169 0.00069110 ------------------------------------------------------------------------------------- TOTAL 0.00002184 0.03584327 (elements=1,679,360) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1679360 Total Bytes 261284 BPFP 1.2447 bits/point EBPFP 1.8790 equivalent bits/point MSE 0.035843 ---------------------- -------------------------------------------------------- Time: 4.965s Load: 0.010s, Pack+Encode: 2.599s, Decode+Unpack: 2.356s ---------------------- -------------------------------------------------------- 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 0.0358 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample44-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,944B, BPFP=1.3394 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,796B, BPFP=0.4758 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,860B, BPFP=2.6514 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,692B, BPFP=1.2629 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,272B, BPFP=2.9961 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,412B, BPFP=0.9407 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,928B, BPFP=2.9121 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,676B, BPFP=0.7126 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,368B, BPFP=2.5312 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,304B, BPFP=1.0222 ⌛️ [2/4] FRONTEND: Frontend time: 2.555s (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: 2.315s [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, 184, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000012 0.00012921 layer.1.conv_state 0.00051907 0.83965838 layer.2.ssm_state 0.00000001 0.00000886 layer.2.conv_state 0.00017046 0.17237553 layer.3.ssm_state 0.00000001 0.00001065 layer.3.conv_state 0.00006998 0.16014510 layer.4.ssm_state 0.00000001 0.00001510 layer.4.conv_state 0.00024001 0.34284419 layer.4.output 0.00000196 0.00072485 ------------------------------------------------------------------------------------- TOTAL 0.00002482 0.03805167 (elements=1,572,864) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1572864 Total Bytes 229020 BPFP 1.1649 bits/point EBPFP 1.8399 equivalent bits/point MSE 0.038052 ---------------------- -------------------------------------------------------- Time: 4.880s Load: 0.009s, Pack+Encode: 2.555s, Decode+Unpack: 2.315s ---------------------- -------------------------------------------------------- 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 0.0381 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample45-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 21,776B, BPFP=1.3291 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,800B, BPFP=0.4761 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,920B, BPFP=2.6660 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,732B, BPFP=1.2654 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,260B, BPFP=2.9932 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,412B, BPFP=0.9407 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,892B, BPFP=2.9033 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,556B, BPFP=0.7053 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,172B, BPFP=2.4834 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 130,232B, BPFP=1.2978 ⌛️ [2/4] FRONTEND: Frontend time: 2.624s (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: 2.359s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00012865 layer.1.conv_state 0.00047987 0.84714592 layer.2.ssm_state 0.00000001 0.00000873 layer.2.conv_state 0.00014269 0.17056730 layer.3.ssm_state 0.00000001 0.00001020 layer.3.conv_state 0.00007044 0.16032796 layer.4.ssm_state 0.00000001 0.00001516 layer.4.conv_state 0.00031684 0.35163572 layer.4.output 0.00000179 0.00075170 ------------------------------------------------------------------------------------- TOTAL 0.00002426 0.03722979 (elements=1,622,016) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1622016 Total Bytes 262520 BPFP 1.2948 bits/point EBPFP 1.9472 equivalent bits/point MSE 0.037230 ---------------------- -------------------------------------------------------- Time: 4.993s Load: 0.010s, Pack+Encode: 2.624s, Decode+Unpack: 2.359s ---------------------- -------------------------------------------------------- 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 0.0372 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample46-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,816B, BPFP=1.3315 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,828B, BPFP=0.4778 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,736B, BPFP=1.2656 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,280B, BPFP=2.9980 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,404B, BPFP=0.9402 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,904B, BPFP=2.9062 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,188B, BPFP=0.7439 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,376B, BPFP=2.5332 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 127,832B, BPFP=1.2299 ⌛️ [2/4] FRONTEND: Frontend time: 2.611s (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: 2.368s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000043 0.00012832 layer.1.conv_state 0.00048147 0.84376442 layer.2.ssm_state 0.00000001 0.00000875 layer.2.conv_state 0.00012906 0.17283955 layer.3.ssm_state 0.00000001 0.00001053 layer.3.conv_state 0.00011463 0.16183850 layer.4.ssm_state 0.00000002 0.00001479 layer.4.conv_state 0.00022714 0.33756426 layer.4.output 0.00000188 0.00071865 ------------------------------------------------------------------------------------- TOTAL 0.00002279 0.03630814 (elements=1,650,688) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1650688 Total Bytes 260984 BPFP 1.2648 bits/point EBPFP 1.9102 equivalent bits/point MSE 0.036308 ---------------------- -------------------------------------------------------- Time: 4.989s Load: 0.010s, Pack+Encode: 2.611s, Decode+Unpack: 2.368s ---------------------- -------------------------------------------------------- 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 0.0363 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample47-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,948B, BPFP=1.3396 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,780B, BPFP=0.4749 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,860B, BPFP=2.6514 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,684B, BPFP=1.2625 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,304B, BPFP=3.0039 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,348B, BPFP=0.9368 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 12,000B, BPFP=2.9297 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,548B, BPFP=0.7048 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,416B, BPFP=2.5430 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,744B, BPFP=1.0057 ⌛️ [2/4] FRONTEND: Frontend time: 2.551s (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: 2.297s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00012937 layer.1.conv_state 0.00050109 0.84611291 layer.2.ssm_state 0.00000001 0.00000873 layer.2.conv_state 0.00013610 0.17272067 layer.3.ssm_state 0.00000001 0.00001048 layer.3.conv_state 0.00007718 0.16082443 layer.4.ssm_state 0.00000001 0.00001477 layer.4.conv_state 0.00025147 0.34398907 layer.4.output 0.00000198 0.00071098 ------------------------------------------------------------------------------------- TOTAL 0.00002414 0.03822465 (elements=1,572,864) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1572864 Total Bytes 227400 BPFP 1.1566 bits/point EBPFP 1.8313 equivalent bits/point MSE 0.038225 ---------------------- -------------------------------------------------------- Time: 4.857s Load: 0.009s, Pack+Encode: 2.551s, Decode+Unpack: 2.297s ---------------------- -------------------------------------------------------- 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 0.0382 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample48-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 22,096B, BPFP=1.3486 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,816B, BPFP=0.4771 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,872B, BPFP=2.6543 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,864B, BPFP=1.2734 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,260B, BPFP=2.9932 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,264B, BPFP=0.9316 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,864B, BPFP=2.8965 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,040B, BPFP=0.7349 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,244B, BPFP=2.5010 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 82,876B, BPFP=0.8431 ⌛️ [2/4] FRONTEND: Frontend time: 2.551s (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: 2.306s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000024 0.00013137 layer.1.conv_state 0.00048279 0.84447843 layer.2.ssm_state 0.00000001 0.00000884 layer.2.conv_state 0.00015517 0.17012501 layer.3.ssm_state 0.00000001 0.00001037 layer.3.conv_state 0.00011498 0.15821236 layer.4.ssm_state 0.00000001 0.00001425 layer.4.conv_state 0.00025157 0.32806662 layer.4.output 0.00000201 0.00055552 ------------------------------------------------------------------------------------- TOTAL 0.00002449 0.03691846 (elements=1,605,632) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1605632 Total Bytes 215964 BPFP 1.0760 bits/point EBPFP 1.7391 equivalent bits/point MSE 0.036918 ---------------------- -------------------------------------------------------- Time: 4.867s Load: 0.010s, Pack+Encode: 2.551s, Decode+Unpack: 2.306s ---------------------- -------------------------------------------------------- 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 0.0369 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample51-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 21,892B, BPFP=1.3362 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,828B, BPFP=0.4778 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,836B, BPFP=2.6455 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,996B, BPFP=1.2815 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,236B, BPFP=2.9873 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,384B, BPFP=0.9390 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,900B, BPFP=2.9053 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,748B, BPFP=0.7170 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,236B, BPFP=2.4990 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,292B, BPFP=1.1063 ⌛️ [2/4] FRONTEND: Frontend time: 2.545s (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: 2.314s [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, 170, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.ssm_state 0.00000001 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000011 0.00013020 layer.1.conv_state 0.00049192 0.84315205 layer.2.ssm_state 0.00000001 0.00000891 layer.2.conv_state 0.00013328 0.17154075 layer.3.ssm_state 0.00000001 0.00001053 layer.3.conv_state 0.00007075 0.16035001 layer.4.ssm_state 0.00000001 0.00001536 layer.4.conv_state 0.00027001 0.34556195 layer.4.output 0.00000204 0.00081457 ------------------------------------------------------------------------------------- TOTAL 0.00002499 0.03962604 (elements=1,515,520) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1515520 Total Bytes 229116 BPFP 1.2094 bits/point EBPFP 1.9106 equivalent bits/point MSE 0.039626 ---------------------- -------------------------------------------------------- Time: 4.870s Load: 0.011s, Pack+Encode: 2.545s, Decode+Unpack: 2.314s ---------------------- -------------------------------------------------------- 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 0.0396 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample52-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 21,904B, BPFP=1.3369 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,768B, BPFP=0.4741 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,848B, BPFP=2.6484 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,828B, BPFP=1.2712 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,248B, BPFP=2.9902 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,228B, BPFP=0.9294 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,868B, BPFP=2.8975 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,188B, BPFP=0.7439 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,324B, BPFP=2.5205 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 135,960B, BPFP=1.3277 ⌛️ [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, 200, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.362s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000043 0.00012800 layer.1.conv_state 0.00049042 0.83626693 layer.2.ssm_state 0.00000001 0.00000871 layer.2.conv_state 0.00012605 0.17014191 layer.3.ssm_state 0.00000001 0.00001031 layer.3.conv_state 0.00011250 0.15888722 layer.4.ssm_state 0.00000002 0.00001463 layer.4.conv_state 0.00024020 0.33493409 layer.4.output 0.00000182 0.00069415 ------------------------------------------------------------------------------------- TOTAL 0.00002325 0.03624722 (elements=1,638,400) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1638400 Total Bytes 268932 BPFP 1.3131 bits/point EBPFP 1.9624 equivalent bits/point MSE 0.036247 ---------------------- -------------------------------------------------------- Time: 4.975s Load: 0.010s, Pack+Encode: 2.603s, Decode+Unpack: 2.362s ---------------------- -------------------------------------------------------- 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 0.0362 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample53-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 21,832B, BPFP=1.3325 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,808B, BPFP=0.4766 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,872B, BPFP=2.6543 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,704B, BPFP=1.2637 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,272B, BPFP=2.9961 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,272B, BPFP=0.9321 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,944B, BPFP=2.9160 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,992B, BPFP=0.7319 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,360B, BPFP=2.5293 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 130,936B, BPFP=1.2851 ⌛️ [2/4] FRONTEND: Frontend time: 2.598s (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: 2.356s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000043 0.00012802 layer.1.conv_state 0.00048601 0.84081209 layer.2.ssm_state 0.00000001 0.00000870 layer.2.conv_state 0.00014442 0.17297217 layer.3.ssm_state 0.00000001 0.00001048 layer.3.conv_state 0.00011344 0.16224086 layer.4.ssm_state 0.00000006 0.00001491 layer.4.conv_state 0.00022676 0.34033462 layer.4.output 0.00000196 0.00071531 ------------------------------------------------------------------------------------- TOTAL 0.00002341 0.03667031 (elements=1,634,304) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1634304 Total Bytes 263760 BPFP 1.2911 bits/point EBPFP 1.9413 equivalent bits/point MSE 0.036670 ---------------------- -------------------------------------------------------- Time: 4.964s Load: 0.010s, Pack+Encode: 2.598s, Decode+Unpack: 2.356s ---------------------- -------------------------------------------------------- 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 0.0367 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample55-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 21,976B, BPFP=1.3413 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,764B, BPFP=0.4739 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,848B, BPFP=2.6484 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,652B, BPFP=1.2605 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,240B, BPFP=2.9883 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,164B, BPFP=0.9255 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,924B, BPFP=2.9111 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,700B, BPFP=0.7141 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,304B, BPFP=2.5156 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 117,176B, BPFP=1.0645 ⌛️ [2/4] FRONTEND: Frontend time: 2.597s (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: 2.366s [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 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000015 0.00013013 layer.1.conv_state 0.00047793 0.84300911 layer.2.ssm_state 0.00000001 0.00000875 layer.2.conv_state 0.00012432 0.17365006 layer.3.ssm_state 0.00000001 0.00001010 layer.3.conv_state 0.00007747 0.16342980 layer.4.ssm_state 0.00000001 0.00001508 layer.4.conv_state 0.00032616 0.35202450 layer.4.output 0.00000162 0.00065832 ------------------------------------------------------------------------------------- TOTAL 0.00002306 0.03555840 (elements=1,699,840) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1699840 Total Bytes 249516 BPFP 1.1743 bits/point EBPFP 1.7971 equivalent bits/point MSE 0.035558 ---------------------- -------------------------------------------------------- Time: 4.974s Load: 0.010s, Pack+Encode: 2.597s, Decode+Unpack: 2.366s ---------------------- -------------------------------------------------------- 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 0.0356 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample56-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,124B, BPFP=1.3503 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,828B, BPFP=0.4778 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,824B, BPFP=2.6426 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,988B, BPFP=1.2810 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,240B, BPFP=2.9883 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,476B, BPFP=0.9446 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,880B, BPFP=2.9004 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,852B, BPFP=0.7234 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,272B, BPFP=2.5078 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 99,324B, BPFP=1.1149 ⌛️ [2/4] FRONTEND: Frontend time: 2.543s (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: 2.294s [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 0.00000565 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00013048 layer.1.conv_state 0.00048286 0.84227973 layer.2.ssm_state 0.00000001 0.00000875 layer.2.conv_state 0.00013493 0.17045967 layer.3.ssm_state 0.00000001 0.00001030 layer.3.conv_state 0.00011535 0.15849391 layer.4.ssm_state 0.00000001 0.00001504 layer.4.conv_state 0.00022609 0.34132272 layer.4.output 0.00000231 0.00080835 ------------------------------------------------------------------------------------- TOTAL 0.00002474 0.03903586 (elements=1,531,904) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1531904 Total Bytes 232576 BPFP 1.2146 bits/point EBPFP 1.9104 equivalent bits/point MSE 0.039036 ---------------------- -------------------------------------------------------- Time: 4.846s Load: 0.009s, Pack+Encode: 2.543s, Decode+Unpack: 2.294s ---------------------- -------------------------------------------------------- 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 0.0390 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample58-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,800B, BPFP=1.3306 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,804B, BPFP=0.4763 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,876B, BPFP=2.6553 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,716B, BPFP=1.2644 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,248B, BPFP=2.9902 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,276B, BPFP=0.9324 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,888B, BPFP=2.9023 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,084B, BPFP=0.7375 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,316B, BPFP=2.5186 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,680B, BPFP=0.9942 ⌛️ [2/4] FRONTEND: Frontend time: 2.546s (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: 2.298s [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 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000027 0.00012851 layer.1.conv_state 0.00049416 0.84354496 layer.2.ssm_state 0.00000001 0.00000872 layer.2.conv_state 0.00014839 0.17216282 layer.3.ssm_state 0.00000001 0.00001039 layer.3.conv_state 0.00011359 0.16140117 layer.4.ssm_state 0.00000002 0.00001478 layer.4.conv_state 0.00020754 0.33455652 layer.4.output 0.00000201 0.00069599 ------------------------------------------------------------------------------------- TOTAL 0.00002400 0.03777465 (elements=1,581,056) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1581056 Total Bytes 227456 BPFP 1.1509 bits/point EBPFP 1.8227 equivalent bits/point MSE 0.037775 ---------------------- -------------------------------------------------------- Time: 4.853s Load: 0.009s, Pack+Encode: 2.546s, Decode+Unpack: 2.298s ---------------------- -------------------------------------------------------- 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 0.0378 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample63-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 22,068B, BPFP=1.3469 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,868B, BPFP=0.4802 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,836B, BPFP=2.6455 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,016B, BPFP=1.2827 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,220B, BPFP=2.9834 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,444B, BPFP=0.9426 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,812B, BPFP=2.8838 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,852B, BPFP=0.7234 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,248B, BPFP=2.5020 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,816B, BPFP=1.0010 ⌛️ [2/4] FRONTEND: Frontend time: 2.546s (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: 2.293s [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 0.00000565 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000024 0.00012965 layer.1.conv_state 0.00048011 0.84220392 layer.2.ssm_state 0.00000001 0.00000878 layer.2.conv_state 0.00013394 0.16984117 layer.3.ssm_state 0.00000001 0.00001045 layer.3.conv_state 0.00011242 0.15857251 layer.4.ssm_state 0.00000002 0.00001481 layer.4.conv_state 0.00023441 0.33675355 layer.4.output 0.00000189 0.00071246 ------------------------------------------------------------------------------------- TOTAL 0.00002393 0.03778975 (elements=1,576,960) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1576960 Total Bytes 227948 BPFP 1.1564 bits/point EBPFP 1.8318 equivalent bits/point MSE 0.037790 ---------------------- -------------------------------------------------------- Time: 4.851s Load: 0.011s, Pack+Encode: 2.546s, Decode+Unpack: 2.293s ---------------------- -------------------------------------------------------- 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 0.0378 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample64-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.013s ------------------------------------------------------------ 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: 22,004B, BPFP=1.3430 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,812B, BPFP=0.4768 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,864B, BPFP=2.6523 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,808B, BPFP=1.2700 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,244B, BPFP=2.9893 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,220B, BPFP=0.9290 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,864B, BPFP=2.8965 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,792B, BPFP=0.7197 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,288B, BPFP=2.5117 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 92,648B, BPFP=1.0109 ⌛️ [2/4] FRONTEND: Frontend time: 2.540s (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: 2.291s [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 0.00000558 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000030 0.00013062 layer.1.conv_state 0.00050588 0.84494650 layer.2.ssm_state 0.00000001 0.00000881 layer.2.conv_state 0.00011591 0.17137831 layer.3.ssm_state 0.00000001 0.00001004 layer.3.conv_state 0.00011742 0.15914799 layer.4.ssm_state 0.00000003 0.00001462 layer.4.conv_state 0.00023463 0.33167151 layer.4.output 0.00000192 0.00070862 ------------------------------------------------------------------------------------- TOTAL 0.00002458 0.03837016 (elements=1,552,384) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1552384 Total Bytes 225312 BPFP 1.1611 bits/point EBPFP 1.8448 equivalent bits/point MSE 0.038370 ---------------------- -------------------------------------------------------- Time: 4.844s Load: 0.013s, Pack+Encode: 2.540s, Decode+Unpack: 2.291s ---------------------- -------------------------------------------------------- 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 0.0384 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample67-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,916B, BPFP=1.3376 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,840B, BPFP=0.4785 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,820B, BPFP=2.6416 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,884B, BPFP=1.2747 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,244B, BPFP=2.9893 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,256B, BPFP=0.9312 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,852B, BPFP=2.8936 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,228B, BPFP=0.7463 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,368B, BPFP=2.5312 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 89,072B, BPFP=0.9156 ⌛️ [2/4] FRONTEND: Frontend time: 2.561s (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: 2.293s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000044 0.00012770 layer.1.conv_state 0.00050610 0.84066856 layer.2.ssm_state 0.00000001 0.00000874 layer.2.conv_state 0.00013374 0.17062750 layer.3.ssm_state 0.00000001 0.00001017 layer.3.conv_state 0.00010957 0.15911572 layer.4.ssm_state 0.00000005 0.00001481 layer.4.conv_state 0.00022355 0.33686566 layer.4.output 0.00000189 0.00075982 ------------------------------------------------------------------------------------- TOTAL 0.00002392 0.03733537 (elements=1,597,440) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1597440 Total Bytes 222248 BPFP 1.1130 bits/point EBPFP 1.7800 equivalent bits/point MSE 0.037335 ---------------------- -------------------------------------------------------- Time: 4.864s Load: 0.009s, Pack+Encode: 2.561s, Decode+Unpack: 2.293s ---------------------- -------------------------------------------------------- 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 0.0373 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample68-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,988B, BPFP=1.3420 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,828B, BPFP=0.4778 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,864B, BPFP=2.6523 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,792B, BPFP=1.2690 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,224B, BPFP=2.9844 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,204B, BPFP=0.9280 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,816B, BPFP=2.8848 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,696B, BPFP=0.7139 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,300B, BPFP=2.5146 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,904B, BPFP=1.0903 ⌛️ [2/4] FRONTEND: Frontend time: 2.537s (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: 2.290s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00012985 layer.1.conv_state 0.00047344 0.84506357 layer.2.ssm_state 0.00000001 0.00000871 layer.2.conv_state 0.00014015 0.16988406 layer.3.ssm_state 0.00000001 0.00001037 layer.3.conv_state 0.00011599 0.15786244 layer.4.ssm_state 0.00000001 0.00001423 layer.4.conv_state 0.00022458 0.32719022 layer.4.output 0.00000213 0.00080439 ------------------------------------------------------------------------------------- TOTAL 0.00002480 0.03917571 (elements=1,515,520) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1515520 Total Bytes 227384 BPFP 1.2003 bits/point EBPFP 1.8996 equivalent bits/point MSE 0.039176 ---------------------- -------------------------------------------------------- Time: 4.837s Load: 0.009s, Pack+Encode: 2.537s, Decode+Unpack: 2.290s ---------------------- -------------------------------------------------------- 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 0.0392 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample69-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.013s ------------------------------------------------------------ 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: 22,072B, BPFP=1.3472 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,728B, BPFP=0.4717 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,824B, BPFP=2.6426 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,700B, BPFP=1.2634 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,208B, BPFP=2.9805 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,428B, BPFP=0.9417 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,972B, BPFP=2.9229 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,724B, BPFP=0.7156 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,420B, BPFP=2.5439 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 117,932B, BPFP=0.9802 ⌛️ [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, 235, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.356s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000010 0.00012989 layer.1.conv_state 0.00050392 0.84019357 layer.2.ssm_state 0.00000001 0.00000878 layer.2.conv_state 0.00010721 0.17356059 layer.3.ssm_state 0.00000001 0.00001048 layer.3.conv_state 0.00007772 0.16251937 layer.4.ssm_state 0.00000001 0.00001533 layer.4.conv_state 0.00021031 0.34809658 layer.4.output 0.00000154 0.00067239 ------------------------------------------------------------------------------------- TOTAL 0.00002006 0.03381903 (elements=1,781,760) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1781760 Total Bytes 250776 BPFP 1.1260 bits/point EBPFP 1.7224 equivalent bits/point MSE 0.033819 ---------------------- -------------------------------------------------------- Time: 4.965s Load: 0.013s, Pack+Encode: 2.596s, Decode+Unpack: 2.356s ---------------------- -------------------------------------------------------- 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 0.0338 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample7-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,060B, BPFP=1.3464 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,796B, BPFP=0.4758 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,836B, BPFP=2.6455 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,784B, BPFP=1.2686 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,220B, BPFP=2.9834 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,192B, BPFP=0.9272 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,844B, BPFP=2.8916 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,504B, BPFP=0.7021 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,264B, BPFP=2.5059 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 93,204B, BPFP=1.0343 ⌛️ [2/4] FRONTEND: Frontend time: 2.542s (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: 2.295s [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 0.00000564 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00012756 layer.1.conv_state 0.00049605 0.84172243 layer.2.ssm_state 0.00000001 0.00000883 layer.2.conv_state 0.00014004 0.16957842 layer.3.ssm_state 0.00000001 0.00001011 layer.3.conv_state 0.00011040 0.15720515 layer.4.ssm_state 0.00000002 0.00001538 layer.4.conv_state 0.00070183 0.36017713 layer.4.output 0.00000198 0.00074284 ------------------------------------------------------------------------------------- TOTAL 0.00003488 0.03914475 (elements=1,540,096) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1540096 Total Bytes 225472 BPFP 1.1712 bits/point EBPFP 1.8583 equivalent bits/point MSE 0.039145 ---------------------- -------------------------------------------------------- Time: 4.846s Load: 0.009s, Pack+Encode: 2.542s, Decode+Unpack: 2.295s ---------------------- -------------------------------------------------------- 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 0.0391 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample72-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,012B, BPFP=1.3435 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,812B, BPFP=0.4768 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,836B, BPFP=2.6455 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,948B, BPFP=1.2786 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,240B, BPFP=2.9883 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,296B, BPFP=0.9336 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,952B, BPFP=2.9180 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,440B, BPFP=0.6982 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,300B, BPFP=2.5146 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,396B, BPFP=1.0946 ⌛️ [2/4] FRONTEND: Frontend time: 2.539s (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: 2.303s [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 0.00000561 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000010 0.00012936 layer.1.conv_state 0.00050574 0.84354055 layer.2.ssm_state 0.00000001 0.00000869 layer.2.conv_state 0.00012521 0.17089060 layer.3.ssm_state 0.00000001 0.00001039 layer.3.conv_state 0.00011811 0.15917625 layer.4.ssm_state 0.00000001 0.00001558 layer.4.conv_state 0.00032558 0.35253280 layer.4.output 0.00000215 0.00078727 ------------------------------------------------------------------------------------- TOTAL 0.00002726 0.03952371 (elements=1,523,712) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1523712 Total Bytes 229000 BPFP 1.2023 bits/point EBPFP 1.8985 equivalent bits/point MSE 0.039524 ---------------------- -------------------------------------------------------- Time: 4.851s Load: 0.009s, Pack+Encode: 2.539s, Decode+Unpack: 2.303s ---------------------- -------------------------------------------------------- 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 0.0395 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample74-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 21,908B, BPFP=1.3372 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,796B, BPFP=0.4758 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,872B, BPFP=2.6543 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,092B, BPFP=1.2874 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,260B, BPFP=2.9932 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,652B, BPFP=0.9553 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,932B, BPFP=2.9131 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,668B, BPFP=0.7122 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,372B, BPFP=2.5322 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,148B, BPFP=1.1451 ⌛️ [2/4] FRONTEND: Frontend time: 2.541s (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: 2.287s [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 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000025 0.00013024 layer.1.conv_state 0.00049170 0.84703743 layer.2.ssm_state 0.00000001 0.00000878 layer.2.conv_state 0.00014440 0.17277153 layer.3.ssm_state 0.00000001 0.00001028 layer.3.conv_state 0.00007111 0.16041687 layer.4.ssm_state 0.00000001 0.00001548 layer.4.conv_state 0.00032601 0.35413623 layer.4.output 0.00000215 0.00081994 ------------------------------------------------------------------------------------- TOTAL 0.00002691 0.04057055 (elements=1,490,944) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1490944 Total Bytes 229468 BPFP 1.2313 bits/point EBPFP 1.9466 equivalent bits/point MSE 0.040571 ---------------------- -------------------------------------------------------- Time: 4.837s Load: 0.009s, Pack+Encode: 2.541s, Decode+Unpack: 2.287s ---------------------- -------------------------------------------------------- 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 0.0406 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample75-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 21,964B, BPFP=1.3406 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,812B, BPFP=0.4768 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,864B, BPFP=2.6523 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,872B, BPFP=1.2739 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,232B, BPFP=2.9863 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,264B, BPFP=0.9316 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,900B, BPFP=2.9053 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,800B, BPFP=0.7202 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,376B, BPFP=2.5332 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 97,448B, BPFP=1.1466 ⌛️ [2/4] FRONTEND: Frontend time: 2.549s (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: 2.289s [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 0.00000558 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000031 0.00013107 layer.1.conv_state 0.00048773 0.84134728 layer.2.ssm_state 0.00000001 0.00000869 layer.2.conv_state 0.00012531 0.17121613 layer.3.ssm_state 0.00000001 0.00001029 layer.3.conv_state 0.00012205 0.16018231 layer.4.ssm_state 0.00000002 0.00001451 layer.4.conv_state 0.00022309 0.33208191 layer.4.output 0.00000212 0.00081986 ------------------------------------------------------------------------------------- TOTAL 0.00002513 0.03970772 (elements=1,499,136) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1499136 Total Bytes 230300 BPFP 1.2290 bits/point EBPFP 1.9379 equivalent bits/point MSE 0.039708 ---------------------- -------------------------------------------------------- Time: 4.847s Load: 0.009s, Pack+Encode: 2.549s, Decode+Unpack: 2.289s ---------------------- -------------------------------------------------------- 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 0.0397 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample76-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,792B, BPFP=1.3301 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,852B, BPFP=0.4792 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,868B, BPFP=2.6533 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,792B, BPFP=1.2690 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,292B, BPFP=3.0010 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,396B, BPFP=0.9397 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,936B, BPFP=2.9141 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,992B, BPFP=0.7319 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,316B, BPFP=2.5186 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 101,580B, BPFP=1.1022 ⌛️ [2/4] FRONTEND: Frontend time: 2.542s (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: 2.292s [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 0.00000557 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00013222 layer.1.conv_state 0.00049117 0.84802812 layer.2.ssm_state 0.00000001 0.00000883 layer.2.conv_state 0.00012889 0.17400147 layer.3.ssm_state 0.00000001 0.00001052 layer.3.conv_state 0.00006806 0.16212468 layer.4.ssm_state 0.00000002 0.00001460 layer.4.conv_state 0.00022608 0.33624014 layer.4.output 0.00000211 0.00079461 ------------------------------------------------------------------------------------- TOTAL 0.00002335 0.03859090 (elements=1,556,480) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1556480 Total Bytes 234584 BPFP 1.2057 bits/point EBPFP 1.8893 equivalent bits/point MSE 0.038591 ---------------------- -------------------------------------------------------- Time: 4.843s Load: 0.009s, Pack+Encode: 2.542s, Decode+Unpack: 2.292s ---------------------- -------------------------------------------------------- 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 0.0386 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample78-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.009s ------------------------------------------------------------ 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: 22,016B, BPFP=1.3438 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,808B, BPFP=0.4766 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,872B, BPFP=2.6543 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 21,020B, BPFP=1.2830 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,176B, BPFP=2.9727 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,332B, BPFP=0.9358 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,840B, BPFP=2.8906 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,972B, BPFP=0.7307 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,312B, BPFP=2.5176 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,580B, BPFP=1.0263 ⌛️ [2/4] FRONTEND: Frontend time: 2.544s (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: 2.286s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000056 0.00013213 layer.1.conv_state 0.00050875 0.84944493 layer.2.ssm_state 0.00000001 0.00000879 layer.2.conv_state 0.00009667 0.17067000 layer.3.ssm_state 0.00000001 0.00001027 layer.3.conv_state 0.00011205 0.15984486 layer.4.ssm_state 0.00000001 0.00001489 layer.4.conv_state 0.00021440 0.33766109 layer.4.output 0.00000203 0.00071938 ------------------------------------------------------------------------------------- TOTAL 0.00002371 0.03849687 (elements=1,556,480) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1556480 Total Bytes 227696 BPFP 1.1703 bits/point EBPFP 1.8545 equivalent bits/point MSE 0.038497 ---------------------- -------------------------------------------------------- Time: 4.839s Load: 0.009s, Pack+Encode: 2.544s, Decode+Unpack: 2.286s ---------------------- -------------------------------------------------------- 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 0.0385 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample80-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,848B, BPFP=1.3335 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,764B, BPFP=0.4739 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,872B, BPFP=2.6543 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,728B, BPFP=1.2651 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,308B, BPFP=3.0049 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,416B, BPFP=0.9409 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 12,012B, BPFP=2.9326 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,424B, BPFP=0.6973 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,444B, BPFP=2.5498 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 94,140B, BPFP=1.0567 ⌛️ [2/4] FRONTEND: Frontend time: 2.539s (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: 2.291s [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 0.00000557 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000011 0.00013030 layer.1.conv_state 0.00049114 0.84383452 layer.2.ssm_state 0.00000001 0.00000878 layer.2.conv_state 0.00011528 0.17241374 layer.3.ssm_state 0.00000001 0.00001057 layer.3.conv_state 0.00007296 0.16179095 layer.4.ssm_state 0.00000001 0.00001534 layer.4.conv_state 0.00026103 0.34874746 layer.4.output 0.00000214 0.00077896 ------------------------------------------------------------------------------------- TOTAL 0.00002425 0.03932661 (elements=1,531,904) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1531904 Total Bytes 226724 BPFP 1.1840 bits/point EBPFP 1.8764 equivalent bits/point MSE 0.039327 ---------------------- -------------------------------------------------------- Time: 4.839s Load: 0.009s, Pack+Encode: 2.539s, Decode+Unpack: 2.291s ---------------------- -------------------------------------------------------- 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 0.0393 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample82-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.010s ------------------------------------------------------------ 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: 21,864B, BPFP=1.3345 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,780B, BPFP=0.4749 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,868B, BPFP=2.6533 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,216B, BPFP=1.2339 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,284B, BPFP=2.9990 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 14,780B, BPFP=0.9021 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 12,020B, BPFP=2.9346 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,396B, BPFP=0.7566 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,372B, BPFP=2.5322 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 95,432B, BPFP=1.0355 ⌛️ [2/4] FRONTEND: Frontend time: 2.537s (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: 2.283s [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 0.00000557 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000037 0.00012904 layer.1.conv_state 0.00049028 0.84551954 layer.2.ssm_state 0.00000001 0.00000877 layer.2.conv_state 0.00010997 0.17281380 layer.3.ssm_state 0.00000001 0.00001041 layer.3.conv_state 0.00007961 0.16112271 layer.4.ssm_state 0.00000002 0.00001417 layer.4.conv_state 0.00024454 0.32959917 layer.4.output 0.00000217 0.00072759 ------------------------------------------------------------------------------------- TOTAL 0.00002360 0.03832012 (elements=1,556,480) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1556480 Total Bytes 227780 BPFP 1.1707 bits/point EBPFP 1.8510 equivalent bits/point MSE 0.038320 ---------------------- -------------------------------------------------------- Time: 4.830s Load: 0.010s, Pack+Encode: 2.537s, Decode+Unpack: 2.283s ---------------------- -------------------------------------------------------- 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 0.0383 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample85-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,916B, BPFP=1.3376 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,756B, BPFP=0.4734 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,832B, BPFP=2.6445 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,732B, BPFP=1.2654 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,228B, BPFP=2.9854 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,384B, BPFP=0.9390 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,912B, BPFP=2.9082 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,572B, BPFP=0.7063 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,392B, BPFP=2.5371 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 96,440B, BPFP=1.0523 ⌛️ [2/4] FRONTEND: Frontend time: 2.530s (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: 2.283s [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 0.00000557 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000028 0.00012805 layer.1.conv_state 0.00049519 0.84487426 layer.2.ssm_state 0.00000001 0.00000879 layer.2.conv_state 0.00011503 0.17137587 layer.3.ssm_state 0.00000001 0.00001046 layer.3.conv_state 0.00006908 0.16010658 layer.4.ssm_state 0.00000001 0.00001513 layer.4.conv_state 0.00022394 0.34252527 layer.4.output 0.00000194 0.00072862 ------------------------------------------------------------------------------------- TOTAL 0.00002309 0.03862722 (elements=1,552,384) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1552384 Total Bytes 228932 BPFP 1.1798 bits/point EBPFP 1.8625 equivalent bits/point MSE 0.038627 ---------------------- -------------------------------------------------------- Time: 4.822s Load: 0.009s, Pack+Encode: 2.530s, Decode+Unpack: 2.283s ---------------------- -------------------------------------------------------- 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 0.0386 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample87-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 21,944B, BPFP=1.3394 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,800B, BPFP=0.4761 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,872B, BPFP=2.6543 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,784B, BPFP=1.2686 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,296B, BPFP=3.0020 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,416B, BPFP=0.9409 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,924B, BPFP=2.9111 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,624B, BPFP=0.7095 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,240B, BPFP=2.5000 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 125,200B, BPFP=1.1928 ⌛️ [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, 205, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.347s [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 0.00000560 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000010 0.00013023 layer.1.conv_state 0.00050539 0.84335709 layer.2.ssm_state 0.00000001 0.00000885 layer.2.conv_state 0.00015098 0.17295060 layer.3.ssm_state 0.00000001 0.00001015 layer.3.conv_state 0.00007050 0.16060811 layer.4.ssm_state 0.00000001 0.00001477 layer.4.conv_state 0.00024328 0.34521455 layer.4.output 0.00000184 0.00072577 ------------------------------------------------------------------------------------- TOTAL 0.00002299 0.03625707 (elements=1,658,880) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1658880 Total Bytes 257868 BPFP 1.2436 bits/point EBPFP 1.8834 equivalent bits/point MSE 0.036257 ---------------------- -------------------------------------------------------- Time: 4.951s Load: 0.011s, Pack+Encode: 2.594s, Decode+Unpack: 2.347s ---------------------- -------------------------------------------------------- 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 0.0363 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample9-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,900B, BPFP=1.3367 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,784B, BPFP=0.4751 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,856B, BPFP=2.6504 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,776B, BPFP=1.2681 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,248B, BPFP=2.9902 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,456B, BPFP=0.9434 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,924B, BPFP=2.9111 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,492B, BPFP=0.7014 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,348B, BPFP=2.5264 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 95,684B, BPFP=1.0740 ⌛️ [2/4] FRONTEND: Frontend time: 2.537s (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: 2.297s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000012 0.00012853 layer.1.conv_state 0.00049634 0.84178400 layer.2.ssm_state 0.00000001 0.00000882 layer.2.conv_state 0.00012038 0.17239180 layer.3.ssm_state 0.00000001 0.00001059 layer.3.conv_state 0.00007294 0.16100042 layer.4.ssm_state 0.00000001 0.00001538 layer.4.conv_state 0.00026211 0.34277272 layer.4.output 0.00000210 0.00077857 ------------------------------------------------------------------------------------- TOTAL 0.00002447 0.03913725 (elements=1,531,904) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1531904 Total Bytes 228236 BPFP 1.1919 bits/point EBPFP 1.8841 equivalent bits/point MSE 0.039137 ---------------------- -------------------------------------------------------- Time: 4.844s Load: 0.009s, Pack+Encode: 2.537s, Decode+Unpack: 2.297s ---------------------- -------------------------------------------------------- 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 0.0391 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample93-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 0.011s ------------------------------------------------------------ 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: 21,980B, BPFP=1.3416 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,772B, BPFP=0.4744 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,844B, BPFP=2.6475 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,864B, BPFP=1.2734 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,192B, BPFP=2.9766 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 14,960B, BPFP=0.9131 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,788B, BPFP=2.8779 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,316B, BPFP=0.7517 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,248B, BPFP=2.5020 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 90,892B, BPFP=0.9343 ⌛️ [2/4] FRONTEND: Frontend time: 2.543s (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: 2.296s [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 0.00000562 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000026 0.00013037 layer.1.conv_state 0.00050538 0.84379798 layer.2.ssm_state 0.00000001 0.00000871 layer.2.conv_state 0.00014003 0.16893710 layer.3.ssm_state 0.00000001 0.00001020 layer.3.conv_state 0.00010410 0.15571785 layer.4.ssm_state 0.00000002 0.00001409 layer.4.conv_state 0.00021663 0.32494000 layer.4.output 0.00000194 0.00079303 ------------------------------------------------------------------------------------- TOTAL 0.00002379 0.03706689 (elements=1,597,440) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1597440 Total Bytes 223624 BPFP 1.1199 bits/point EBPFP 1.7846 equivalent bits/point MSE 0.037067 ---------------------- -------------------------------------------------------- Time: 4.850s Load: 0.011s, Pack+Encode: 2.543s, Decode+Unpack: 2.296s ---------------------- -------------------------------------------------------- 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 0.0371 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample94-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 21,832B, BPFP=1.3325 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,768B, BPFP=0.4741 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,852B, BPFP=2.6494 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,720B, BPFP=1.2646 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,240B, BPFP=2.9883 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,308B, BPFP=0.9343 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,960B, BPFP=2.9199 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 11,692B, BPFP=0.7136 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,360B, BPFP=2.5293 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 91,944B, BPFP=0.9813 ⌛️ [2/4] FRONTEND: Frontend time: 2.551s (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: 2.288s [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 0.00000559 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000041 0.00012744 layer.1.conv_state 0.00050614 0.84429324 layer.2.ssm_state 0.00000001 0.00000882 layer.2.conv_state 0.00012733 0.17243189 layer.3.ssm_state 0.00000001 0.00001051 layer.3.conv_state 0.00007038 0.16017963 layer.4.ssm_state 0.00000001 0.00001508 layer.4.conv_state 0.00022687 0.33930776 layer.4.output 0.00000194 0.00071237 ------------------------------------------------------------------------------------- TOTAL 0.00002346 0.03816784 (elements=1,568,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1568768 Total Bytes 224444 BPFP 1.1446 bits/point EBPFP 1.8203 equivalent bits/point MSE 0.038168 ---------------------- -------------------------------------------------------- Time: 4.848s Load: 0.009s, Pack+Encode: 2.551s, Decode+Unpack: 2.288s ---------------------- -------------------------------------------------------- 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 0.0382 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample95-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/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: 22,024B, BPFP=1.3442 Using per-key quantization points (layer.0.conv_state: torch.Size([256])) for layer.0.conv_state layer.0.conv_state: 9,768B, BPFP=2.3848 Using per-key quantization points (layer.1.ssm_state: torch.Size([256])) for layer.1.ssm_state layer.1.ssm_state: 7,820B, BPFP=0.4773 Using per-key quantization points (layer.1.conv_state: torch.Size([256])) for layer.1.conv_state layer.1.conv_state: 10,840B, BPFP=2.6465 Using per-key quantization points (layer.2.ssm_state: torch.Size([256])) for layer.2.ssm_state layer.2.ssm_state: 20,796B, BPFP=1.2693 Using per-key quantization points (layer.2.conv_state: torch.Size([256])) for layer.2.conv_state layer.2.conv_state: 12,228B, BPFP=2.9854 Using per-key quantization points (layer.3.ssm_state: torch.Size([256])) for layer.3.ssm_state layer.3.ssm_state: 15,260B, BPFP=0.9314 Using per-key quantization points (layer.3.conv_state: torch.Size([256])) for layer.3.conv_state layer.3.conv_state: 11,856B, BPFP=2.8945 Using per-key quantization points (layer.4.ssm_state: torch.Size([256])) for layer.4.ssm_state layer.4.ssm_state: 12,172B, BPFP=0.7429 Using per-key quantization points (layer.4.conv_state: torch.Size([256])) for layer.4.conv_state layer.4.conv_state: 10,340B, BPFP=2.5244 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 88,184B, BPFP=0.9018 ⌛️ [2/4] FRONTEND: Frontend time: 2.542s (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: 2.289s [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 0.00000563 layer.0.conv_state 0.00014612 0.29410776 layer.1.ssm_state 0.00000040 0.00012848 layer.1.conv_state 0.00049989 0.83670658 layer.2.ssm_state 0.00000001 0.00000873 layer.2.conv_state 0.00013469 0.17016456 layer.3.ssm_state 0.00000001 0.00001028 layer.3.conv_state 0.00011145 0.15899943 layer.4.ssm_state 0.00000005 0.00001477 layer.4.conv_state 0.00022056 0.33589101 layer.4.output 0.00000184 0.00068195 ------------------------------------------------------------------------------------- TOTAL 0.00002370 0.03709099 (elements=1,601,536) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler falconmamba Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 1601536 Total Bytes 221288 BPFP 1.1054 bits/point EBPFP 1.7703 equivalent bits/point MSE 0.037091 ---------------------- -------------------------------------------------------- Time: 4.841s Load: 0.010s, Pack+Encode: 2.542s, Decode+Unpack: 2.289s ---------------------- -------------------------------------------------------- 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 0.0371 MSE: from ../datasets/FalconMamba-500features-L5wCache/tiiuae/falcon-mamba-7b-instruct/fc_truthfulqa_mc1/sample99-layer4-item1.zst to output-fixed/falconmamba/lambda0.004/elic-featurecoding-8bit-individual/fc_truthfulqa_mc1/sample99-layer4-item1.zst ------------------------ ---------------------------- TOTAL PROCESSING SUMMARY ------------------------ ---------------------------- Total files 100 Avg BPFP 1.1950 bits/point Avg EBPFP 1.8666 equivalent bits/point Avg MSE 0.037927 Avg Time 4.897s ------------------------ ----------------------------