Experiment: dtufc_elic-featurecoding_qwen_individual Log file: output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/dtufc_elic-featurecoding_qwen_individual.log DTUFCCodecConfig: arch: elic-featurecoding handler: qwen checkpoint: codec_weights/elic_hybrid/elic2022-official_lambda0.001_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_lambda0.001_epochs600_lr0.0001_bs60_patch256-256_checkpoint_best.pth.tar Checkpoint epoch: 286 Loaded elic-featurecoding (1-channel) on cuda:0 Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_k.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.0.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_k.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_v.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.0.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_0_v.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_k.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.1.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_k.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_v.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.1.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_1_v.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_k.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.2.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_k.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_v.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.2.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_2_v.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_k.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.3.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_k.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_v.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.3.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_3_v.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_k.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.4.k_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_k.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_v.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.4.v_cache' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_layer_4_v.json Loaded quantization points from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_feature.json: torch.Size([256]) Loaded per-key quantization points for key 'layer.4.output' from featurecoding_utils/transform_mapping/kmeans10samples-8bits/qwen3/arc_fewshot-8bit_feature.json Loaded per-key mappings: model=qwen Keys: ['layer.0.k_cache', 'layer.0.v_cache', 'layer.1.k_cache', 'layer.1.v_cache', 'layer.2.k_cache', 'layer.2.v_cache', 'layer.3.k_cache', 'layer.3.v_cache', 'layer.4.k_cache', 'layer.4.v_cache', 'layer.4.output'] ---------------- -------------------------------------------------------------------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding Checkpoint codec_weights/elic_hybrid/elic2022-official_lambda0.001_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/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge Output output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge ---------------- -------------------------------------------------------------------------------------------------------------------- Files found: 100 ---------------------------------------------------------------------- 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample0-layer4-item1.zst (1/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample0-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 243, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 243, 128) Output shape: (1, 243, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 243, 128]) -> torch.Size([1, 1, 243, 1024]) layer.0.v_cache: torch.Size([1, 8, 243, 128]) -> torch.Size([1, 1, 243, 1024]) layer.1.k_cache: torch.Size([1, 8, 243, 128]) -> torch.Size([1, 1, 243, 1024]) layer.1.v_cache: torch.Size([1, 8, 243, 128]) -> torch.Size([1, 1, 243, 1024]) layer.2.k_cache: torch.Size([1, 8, 243, 128]) -> torch.Size([1, 1, 243, 1024]) layer.2.v_cache: torch.Size([1, 8, 243, 128]) -> torch.Size([1, 1, 243, 1024]) layer.3.k_cache: torch.Size([1, 8, 243, 128]) -> torch.Size([1, 1, 243, 1024]) layer.3.v_cache: torch.Size([1, 8, 243, 128]) -> torch.Size([1, 1, 243, 1024]) layer.4.k_cache: torch.Size([1, 8, 243, 128]) -> torch.Size([1, 1, 243, 1024]) layer.4.v_cache: torch.Size([1, 8, 243, 128]) -> torch.Size([1, 1, 243, 1024]) layer.4.output: torch.Size([1, 243, 4096]) -> torch.Size([1, 1, 243, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,356B, BPFP=0.0436 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,232B, BPFP=0.2325 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,708B, BPFP=0.2157 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 8,792B, BPFP=0.2827 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,560B, BPFP=0.2431 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 10,016B, BPFP=0.3220 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,736B, BPFP=0.2809 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 9,864B, BPFP=0.3171 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,348B, BPFP=0.2041 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 10,188B, BPFP=0.3275 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,224B, BPFP=0.0420 ⌛️ [2/4] FRONTEND: Frontend time: 2.741s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 243, 128]) layer.0.v_cache: torch.Size([1, 8, 243, 128]) layer.1.k_cache: torch.Size([1, 8, 243, 128]) layer.1.v_cache: torch.Size([1, 8, 243, 128]) layer.2.k_cache: torch.Size([1, 8, 243, 128]) layer.2.v_cache: torch.Size([1, 8, 243, 128]) layer.3.k_cache: torch.Size([1, 8, 243, 128]) layer.3.v_cache: torch.Size([1, 8, 243, 128]) layer.4.k_cache: torch.Size([1, 8, 243, 128]) layer.4.v_cache: torch.Size([1, 8, 243, 128]) layer.4.output: torch.Size([1, 243, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.855s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 243, 128]) layer.0.v_cache: torch.Size([1, 8, 243, 128]) layer.1.k_cache: torch.Size([1, 8, 243, 128]) layer.1.v_cache: torch.Size([1, 8, 243, 128]) layer.2.k_cache: torch.Size([1, 8, 243, 128]) layer.2.v_cache: torch.Size([1, 8, 243, 128]) layer.3.k_cache: torch.Size([1, 8, 243, 128]) layer.3.v_cache: torch.Size([1, 8, 243, 128]) layer.4.k_cache: torch.Size([1, 8, 243, 128]) layer.4.v_cache: torch.Size([1, 8, 243, 128]) layer.4.output: torch.Size([1, 243, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02602252 24.42101096 layer.0.v_cache 0.00000026 0.00060069 layer.1.k_cache 0.00298528 3.44367987 layer.1.v_cache 0.00000081 0.00250262 layer.2.k_cache 0.00118445 1.51988076 layer.2.v_cache 0.00000112 0.00377478 layer.3.k_cache 0.00134199 1.81668405 layer.3.v_cache 0.00000210 0.00625695 layer.4.k_cache 0.00355146 3.61795961 layer.4.v_cache 0.00000307 0.01062691 layer.4.output 0.00016770 0.19078861 ------------------------------------------------------------------------------------- TOTAL 0.00255456 2.54329512 (elements=3,483,648) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 3483648 Total Bytes 82024 BPFP 0.1884 bits/point EBPFP 0.3767 equivalent bits/point MSE 2.543295 ---------------------- -------------------------------------------------------- Time: 4.608s Load: 0.012s, Pack+Encode: 2.741s, Decode+Unpack: 1.855s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 243, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 243, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5433 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample0-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample0-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample1-layer4-item1.zst (2/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample1-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 265, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.014s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 265, 128) Output shape: (1, 265, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 265, 128]) -> torch.Size([1, 1, 265, 1024]) layer.0.v_cache: torch.Size([1, 8, 265, 128]) -> torch.Size([1, 1, 265, 1024]) layer.1.k_cache: torch.Size([1, 8, 265, 128]) -> torch.Size([1, 1, 265, 1024]) layer.1.v_cache: torch.Size([1, 8, 265, 128]) -> torch.Size([1, 1, 265, 1024]) layer.2.k_cache: torch.Size([1, 8, 265, 128]) -> torch.Size([1, 1, 265, 1024]) layer.2.v_cache: torch.Size([1, 8, 265, 128]) -> torch.Size([1, 1, 265, 1024]) layer.3.k_cache: torch.Size([1, 8, 265, 128]) -> torch.Size([1, 1, 265, 1024]) layer.3.v_cache: torch.Size([1, 8, 265, 128]) -> torch.Size([1, 1, 265, 1024]) layer.4.k_cache: torch.Size([1, 8, 265, 128]) -> torch.Size([1, 1, 265, 1024]) layer.4.v_cache: torch.Size([1, 8, 265, 128]) -> torch.Size([1, 1, 265, 1024]) layer.4.output: torch.Size([1, 265, 4096]) -> torch.Size([1, 1, 265, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,500B, BPFP=0.0442 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 8,048B, BPFP=0.2373 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 7,876B, BPFP=0.2322 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 9,664B, BPFP=0.2849 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 9,256B, BPFP=0.2729 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 10,568B, BPFP=0.3116 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 10,916B, BPFP=0.3218 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 11,304B, BPFP=0.3333 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 7,288B, BPFP=0.2149 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 11,128B, BPFP=0.3281 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 6,376B, BPFP=0.0470 ⌛️ [2/4] FRONTEND: Frontend time: 2.571s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 265, 128]) layer.0.v_cache: torch.Size([1, 8, 265, 128]) layer.1.k_cache: torch.Size([1, 8, 265, 128]) layer.1.v_cache: torch.Size([1, 8, 265, 128]) layer.2.k_cache: torch.Size([1, 8, 265, 128]) layer.2.v_cache: torch.Size([1, 8, 265, 128]) layer.3.k_cache: torch.Size([1, 8, 265, 128]) layer.3.v_cache: torch.Size([1, 8, 265, 128]) layer.4.k_cache: torch.Size([1, 8, 265, 128]) layer.4.v_cache: torch.Size([1, 8, 265, 128]) layer.4.output: torch.Size([1, 265, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.896s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 265, 128]) layer.0.v_cache: torch.Size([1, 8, 265, 128]) layer.1.k_cache: torch.Size([1, 8, 265, 128]) layer.1.v_cache: torch.Size([1, 8, 265, 128]) layer.2.k_cache: torch.Size([1, 8, 265, 128]) layer.2.v_cache: torch.Size([1, 8, 265, 128]) layer.3.k_cache: torch.Size([1, 8, 265, 128]) layer.3.v_cache: torch.Size([1, 8, 265, 128]) layer.4.k_cache: torch.Size([1, 8, 265, 128]) layer.4.v_cache: torch.Size([1, 8, 265, 128]) layer.4.output: torch.Size([1, 265, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02444536 28.37499631 layer.0.v_cache 0.00000028 0.00062244 layer.1.k_cache 0.00298152 3.21005836 layer.1.v_cache 0.00000087 0.00260601 layer.2.k_cache 0.00117974 1.46261205 layer.2.v_cache 0.00000116 0.00380549 layer.3.k_cache 0.00134702 1.71873388 layer.3.v_cache 0.00000215 0.00634335 layer.4.k_cache 0.00352357 3.27761184 layer.4.v_cache 0.00000310 0.01043210 layer.4.output 0.00016704 0.19995680 ------------------------------------------------------------------------------------- TOTAL 0.00243950 2.77626065 (elements=3,799,040) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 3799040 Total Bytes 93924 BPFP 0.1978 bits/point EBPFP 0.3956 equivalent bits/point MSE 2.776261 ---------------------- -------------------------------------------------------- Time: 4.482s Load: 0.014s, Pack+Encode: 2.571s, Decode+Unpack: 1.896s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 265, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 265, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7763 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample1-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample1-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample10-layer4-item1.zst (3/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample10-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 213, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 213, 128) Output shape: (1, 213, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.0.v_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.1.k_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.1.v_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.2.k_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.2.v_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.3.k_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.3.v_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.4.k_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.4.v_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.4.output: torch.Size([1, 213, 4096]) -> torch.Size([1, 1, 213, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,200B, BPFP=0.0440 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 8,232B, BPFP=0.3019 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,748B, BPFP=0.2475 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 8,912B, BPFP=0.3269 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,720B, BPFP=0.2832 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 9,856B, BPFP=0.3615 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,648B, BPFP=0.3172 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 9,828B, BPFP=0.3605 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,240B, BPFP=0.2289 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 10,016B, BPFP=0.3674 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 8,332B, BPFP=0.0764 ⌛️ [2/4] FRONTEND: Frontend time: 2.290s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 213, 128]) layer.0.v_cache: torch.Size([1, 8, 213, 128]) layer.1.k_cache: torch.Size([1, 8, 213, 128]) layer.1.v_cache: torch.Size([1, 8, 213, 128]) layer.2.k_cache: torch.Size([1, 8, 213, 128]) layer.2.v_cache: torch.Size([1, 8, 213, 128]) layer.3.k_cache: torch.Size([1, 8, 213, 128]) layer.3.v_cache: torch.Size([1, 8, 213, 128]) layer.4.k_cache: torch.Size([1, 8, 213, 128]) layer.4.v_cache: torch.Size([1, 8, 213, 128]) layer.4.output: torch.Size([1, 213, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.031s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 213, 128]) layer.0.v_cache: torch.Size([1, 8, 213, 128]) layer.1.k_cache: torch.Size([1, 8, 213, 128]) layer.1.v_cache: torch.Size([1, 8, 213, 128]) layer.2.k_cache: torch.Size([1, 8, 213, 128]) layer.2.v_cache: torch.Size([1, 8, 213, 128]) layer.3.k_cache: torch.Size([1, 8, 213, 128]) layer.3.v_cache: torch.Size([1, 8, 213, 128]) layer.4.k_cache: torch.Size([1, 8, 213, 128]) layer.4.v_cache: torch.Size([1, 8, 213, 128]) layer.4.output: torch.Size([1, 213, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02614268 28.29701484 layer.0.v_cache 0.00000026 0.00061958 layer.1.k_cache 0.00299160 3.51319993 layer.1.v_cache 0.00000086 0.00258477 layer.2.k_cache 0.00117415 1.48504954 layer.2.v_cache 0.00000138 0.00387374 layer.3.k_cache 0.00130742 1.74944653 layer.3.v_cache 0.00000224 0.00647355 layer.4.k_cache 0.00347054 3.45500792 layer.4.v_cache 0.00000320 0.01060623 layer.4.output 0.00018153 0.20950799 ------------------------------------------------------------------------------------- TOTAL 0.00255861 2.81156490 (elements=3,053,568) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 3053568 Total Bytes 85732 BPFP 0.2246 bits/point EBPFP 0.4492 equivalent bits/point MSE 2.811565 ---------------------- -------------------------------------------------------- Time: 4.332s Load: 0.011s, Pack+Encode: 2.290s, Decode+Unpack: 2.031s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 213, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8116 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample10-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample10-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample100-layer4-item1.zst (4/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample100-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 173, 128) Output shape: (1, 173, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.output: torch.Size([1, 173, 4096]) -> torch.Size([1, 1, 173, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,040B, BPFP=0.0470 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,120B, BPFP=0.2764 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,400B, BPFP=0.2439 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,812B, BPFP=0.2625 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,604B, BPFP=0.2531 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,200B, BPFP=0.3251 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,772B, BPFP=0.3058 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,284B, BPFP=0.3289 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,144B, BPFP=0.2323 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,384B, BPFP=0.3335 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,288B, BPFP=0.0597 ⌛️ [2/4] FRONTEND: Frontend time: 2.100s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) layer.4.output: torch.Size([1, 173, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.639s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) layer.4.output: torch.Size([1, 173, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02711869 28.44845105 layer.0.v_cache 0.00000027 0.00063748 layer.1.k_cache 0.00316613 3.50602061 layer.1.v_cache 0.00000085 0.00262185 layer.2.k_cache 0.00116533 1.47713228 layer.2.v_cache 0.00000134 0.00373254 layer.3.k_cache 0.00131265 1.78364166 layer.3.v_cache 0.00000224 0.00629945 layer.4.k_cache 0.00350563 3.62487969 layer.4.v_cache 0.00000306 0.01036704 layer.4.output 0.00018814 0.19193905 ------------------------------------------------------------------------------------- TOTAL 0.00264491 2.83082428 (elements=2,480,128) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2480128 Total Bytes 63048 BPFP 0.2034 bits/point EBPFP 0.4067 equivalent bits/point MSE 2.830824 ---------------------- -------------------------------------------------------- Time: 3.748s Load: 0.009s, Pack+Encode: 2.100s, Decode+Unpack: 1.639s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8308 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample100-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample100-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample101-layer4-item1.zst (5/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample101-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 162, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 162, 128) Output shape: (1, 162, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.0.v_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.1.k_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.1.v_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.2.k_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.2.v_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.3.k_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.3.v_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.4.k_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.4.v_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.4.output: torch.Size([1, 162, 4096]) -> torch.Size([1, 1, 162, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 992B, BPFP=0.0478 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,684B, BPFP=0.2741 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,028B, BPFP=0.2425 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,208B, BPFP=0.2994 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,880B, BPFP=0.2836 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,004B, BPFP=0.3378 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,732B, BPFP=0.3247 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,212B, BPFP=0.3478 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,132B, BPFP=0.2475 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,236B, BPFP=0.3490 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,984B, BPFP=0.0480 ⌛️ [2/4] FRONTEND: Frontend time: 1.953s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 162, 128]) layer.0.v_cache: torch.Size([1, 8, 162, 128]) layer.1.k_cache: torch.Size([1, 8, 162, 128]) layer.1.v_cache: torch.Size([1, 8, 162, 128]) layer.2.k_cache: torch.Size([1, 8, 162, 128]) layer.2.v_cache: torch.Size([1, 8, 162, 128]) layer.3.k_cache: torch.Size([1, 8, 162, 128]) layer.3.v_cache: torch.Size([1, 8, 162, 128]) layer.4.k_cache: torch.Size([1, 8, 162, 128]) layer.4.v_cache: torch.Size([1, 8, 162, 128]) layer.4.output: torch.Size([1, 162, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.412s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 162, 128]) layer.0.v_cache: torch.Size([1, 8, 162, 128]) layer.1.k_cache: torch.Size([1, 8, 162, 128]) layer.1.v_cache: torch.Size([1, 8, 162, 128]) layer.2.k_cache: torch.Size([1, 8, 162, 128]) layer.2.v_cache: torch.Size([1, 8, 162, 128]) layer.3.k_cache: torch.Size([1, 8, 162, 128]) layer.3.v_cache: torch.Size([1, 8, 162, 128]) layer.4.k_cache: torch.Size([1, 8, 162, 128]) layer.4.v_cache: torch.Size([1, 8, 162, 128]) layer.4.output: torch.Size([1, 162, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02797630 27.57303422 layer.0.v_cache 0.00000026 0.00061626 layer.1.k_cache 0.00305816 3.24558060 layer.1.v_cache 0.00000079 0.00252558 layer.2.k_cache 0.00117271 1.50031177 layer.2.v_cache 0.00000112 0.00382285 layer.3.k_cache 0.00136252 1.72721806 layer.3.v_cache 0.00000203 0.00616286 layer.4.k_cache 0.00349277 3.43176194 layer.4.v_cache 0.00000321 0.01056087 layer.4.output 0.00018139 0.18818272 ------------------------------------------------------------------------------------- TOTAL 0.00269967 2.73245185 (elements=2,322,432) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2322432 Total Bytes 61092 BPFP 0.2104 bits/point EBPFP 0.4209 equivalent bits/point MSE 2.732452 ---------------------- -------------------------------------------------------- Time: 3.375s Load: 0.010s, Pack+Encode: 1.953s, Decode+Unpack: 1.412s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 162, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7325 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample101-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample101-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample102-layer4-item1.zst (6/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample102-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 156, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.008s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 156, 128) Output shape: (1, 156, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.0.v_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.1.k_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.1.v_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.2.k_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.2.v_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.3.k_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.3.v_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.4.k_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.4.v_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.4.output: torch.Size([1, 156, 4096]) -> torch.Size([1, 1, 156, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,008B, BPFP=0.0505 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,948B, BPFP=0.2979 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,764B, BPFP=0.2386 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,864B, BPFP=0.2937 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,408B, BPFP=0.2708 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,900B, BPFP=0.3456 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,612B, BPFP=0.3311 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,140B, BPFP=0.3576 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,864B, BPFP=0.2436 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,920B, BPFP=0.3466 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,840B, BPFP=0.0481 ⌛️ [2/4] FRONTEND: Frontend time: 2.001s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 156, 128]) layer.0.v_cache: torch.Size([1, 8, 156, 128]) layer.1.k_cache: torch.Size([1, 8, 156, 128]) layer.1.v_cache: torch.Size([1, 8, 156, 128]) layer.2.k_cache: torch.Size([1, 8, 156, 128]) layer.2.v_cache: torch.Size([1, 8, 156, 128]) layer.3.k_cache: torch.Size([1, 8, 156, 128]) layer.3.v_cache: torch.Size([1, 8, 156, 128]) layer.4.k_cache: torch.Size([1, 8, 156, 128]) layer.4.v_cache: torch.Size([1, 8, 156, 128]) layer.4.output: torch.Size([1, 156, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.498s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 156, 128]) layer.0.v_cache: torch.Size([1, 8, 156, 128]) layer.1.k_cache: torch.Size([1, 8, 156, 128]) layer.1.v_cache: torch.Size([1, 8, 156, 128]) layer.2.k_cache: torch.Size([1, 8, 156, 128]) layer.2.v_cache: torch.Size([1, 8, 156, 128]) layer.3.k_cache: torch.Size([1, 8, 156, 128]) layer.3.v_cache: torch.Size([1, 8, 156, 128]) layer.4.k_cache: torch.Size([1, 8, 156, 128]) layer.4.v_cache: torch.Size([1, 8, 156, 128]) layer.4.output: torch.Size([1, 156, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02683717 30.02873973 layer.0.v_cache 0.00000027 0.00060221 layer.1.k_cache 0.00303383 3.37394362 layer.1.v_cache 0.00000085 0.00253447 layer.2.k_cache 0.00113250 1.46058753 layer.2.v_cache 0.00000110 0.00359857 layer.3.k_cache 0.00132873 1.77781442 layer.3.v_cache 0.00000207 0.00594472 layer.4.k_cache 0.00336583 3.44646200 layer.4.v_cache 0.00000309 0.01039850 layer.4.output 0.00014117 0.18354770 ------------------------------------------------------------------------------------- TOTAL 0.00259072 2.91748690 (elements=2,236,416) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2236416 Total Bytes 59268 BPFP 0.2120 bits/point EBPFP 0.4240 equivalent bits/point MSE 2.917487 ---------------------- -------------------------------------------------------- Time: 3.508s Load: 0.008s, Pack+Encode: 2.001s, Decode+Unpack: 1.498s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 156, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.9175 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample102-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample102-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample103-layer4-item1.zst (7/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample103-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 158, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 158, 128) Output shape: (1, 158, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 158, 128]) -> torch.Size([1, 1, 158, 1024]) layer.0.v_cache: torch.Size([1, 8, 158, 128]) -> torch.Size([1, 1, 158, 1024]) layer.1.k_cache: torch.Size([1, 8, 158, 128]) -> torch.Size([1, 1, 158, 1024]) layer.1.v_cache: torch.Size([1, 8, 158, 128]) -> torch.Size([1, 1, 158, 1024]) layer.2.k_cache: torch.Size([1, 8, 158, 128]) -> torch.Size([1, 1, 158, 1024]) layer.2.v_cache: torch.Size([1, 8, 158, 128]) -> torch.Size([1, 1, 158, 1024]) layer.3.k_cache: torch.Size([1, 8, 158, 128]) -> torch.Size([1, 1, 158, 1024]) layer.3.v_cache: torch.Size([1, 8, 158, 128]) -> torch.Size([1, 1, 158, 1024]) layer.4.k_cache: torch.Size([1, 8, 158, 128]) -> torch.Size([1, 1, 158, 1024]) layer.4.v_cache: torch.Size([1, 8, 158, 128]) -> torch.Size([1, 1, 158, 1024]) layer.4.output: torch.Size([1, 158, 4096]) -> torch.Size([1, 1, 158, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,008B, BPFP=0.0498 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,684B, BPFP=0.2811 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,588B, BPFP=0.2269 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,600B, BPFP=0.2769 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,292B, BPFP=0.2617 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,064B, BPFP=0.3493 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,276B, BPFP=0.3103 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,484B, BPFP=0.3206 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,284B, BPFP=0.2118 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,600B, BPFP=0.3263 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,404B, BPFP=0.0421 ⌛️ [2/4] FRONTEND: Frontend time: 1.863s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 158, 128]) layer.0.v_cache: torch.Size([1, 8, 158, 128]) layer.1.k_cache: torch.Size([1, 8, 158, 128]) layer.1.v_cache: torch.Size([1, 8, 158, 128]) layer.2.k_cache: torch.Size([1, 8, 158, 128]) layer.2.v_cache: torch.Size([1, 8, 158, 128]) layer.3.k_cache: torch.Size([1, 8, 158, 128]) layer.3.v_cache: torch.Size([1, 8, 158, 128]) layer.4.k_cache: torch.Size([1, 8, 158, 128]) layer.4.v_cache: torch.Size([1, 8, 158, 128]) layer.4.output: torch.Size([1, 158, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.628s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 158, 128]) layer.0.v_cache: torch.Size([1, 8, 158, 128]) layer.1.k_cache: torch.Size([1, 8, 158, 128]) layer.1.v_cache: torch.Size([1, 8, 158, 128]) layer.2.k_cache: torch.Size([1, 8, 158, 128]) layer.2.v_cache: torch.Size([1, 8, 158, 128]) layer.3.k_cache: torch.Size([1, 8, 158, 128]) layer.3.v_cache: torch.Size([1, 8, 158, 128]) layer.4.k_cache: torch.Size([1, 8, 158, 128]) layer.4.v_cache: torch.Size([1, 8, 158, 128]) layer.4.output: torch.Size([1, 158, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02787131 29.93170243 layer.0.v_cache 0.00000026 0.00059988 layer.1.k_cache 0.00312961 3.28179352 layer.1.v_cache 0.00000083 0.00249705 layer.2.k_cache 0.00117781 1.49675374 layer.2.v_cache 0.00000107 0.00365315 layer.3.k_cache 0.00134038 1.75388423 layer.3.v_cache 0.00000203 0.00595493 layer.4.k_cache 0.00346136 3.27884491 layer.4.v_cache 0.00000301 0.01025821 layer.4.output 0.00016064 0.17442229 ------------------------------------------------------------------------------------- TOTAL 0.00268788 2.89025937 (elements=2,265,088) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2265088 Total Bytes 56284 BPFP 0.1988 bits/point EBPFP 0.3976 equivalent bits/point MSE 2.890259 ---------------------- -------------------------------------------------------- Time: 3.500s Load: 0.009s, Pack+Encode: 1.863s, Decode+Unpack: 1.628s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 158, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 158, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8903 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample103-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample103-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample104-layer4-item1.zst (8/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample104-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 182, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 182, 128) Output shape: (1, 182, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.0.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.1.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.1.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.2.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.2.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.3.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.3.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.4.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.4.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.4.output: torch.Size([1, 182, 4096]) -> torch.Size([1, 1, 182, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,148B, BPFP=0.0493 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,496B, BPFP=0.2359 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,104B, BPFP=0.2191 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,860B, BPFP=0.2945 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,592B, BPFP=0.2400 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,764B, BPFP=0.3333 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,764B, BPFP=0.2904 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,224B, BPFP=0.3530 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,744B, BPFP=0.2036 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,688B, BPFP=0.3300 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,760B, BPFP=0.0511 ⌛️ [2/4] FRONTEND: Frontend time: 1.901s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 182, 128]) layer.0.v_cache: torch.Size([1, 8, 182, 128]) layer.1.k_cache: torch.Size([1, 8, 182, 128]) layer.1.v_cache: torch.Size([1, 8, 182, 128]) layer.2.k_cache: torch.Size([1, 8, 182, 128]) layer.2.v_cache: torch.Size([1, 8, 182, 128]) layer.3.k_cache: torch.Size([1, 8, 182, 128]) layer.3.v_cache: torch.Size([1, 8, 182, 128]) layer.4.k_cache: torch.Size([1, 8, 182, 128]) layer.4.v_cache: torch.Size([1, 8, 182, 128]) layer.4.output: torch.Size([1, 182, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.497s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 182, 128]) layer.0.v_cache: torch.Size([1, 8, 182, 128]) layer.1.k_cache: torch.Size([1, 8, 182, 128]) layer.1.v_cache: torch.Size([1, 8, 182, 128]) layer.2.k_cache: torch.Size([1, 8, 182, 128]) layer.2.v_cache: torch.Size([1, 8, 182, 128]) layer.3.k_cache: torch.Size([1, 8, 182, 128]) layer.3.v_cache: torch.Size([1, 8, 182, 128]) layer.4.k_cache: torch.Size([1, 8, 182, 128]) layer.4.v_cache: torch.Size([1, 8, 182, 128]) layer.4.output: torch.Size([1, 182, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02710123 28.39188863 layer.0.v_cache 0.00000028 0.00060666 layer.1.k_cache 0.00304590 3.57703878 layer.1.v_cache 0.00000075 0.00244286 layer.2.k_cache 0.00114207 1.50058721 layer.2.v_cache 0.00000117 0.00366041 layer.3.k_cache 0.00132402 1.88634952 layer.3.v_cache 0.00000202 0.00598628 layer.4.k_cache 0.00339800 3.69150350 layer.4.v_cache 0.00000283 0.00977277 layer.4.output 0.00018478 0.19399102 ------------------------------------------------------------------------------------- TOTAL 0.00262553 2.84612862 (elements=2,609,152) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2609152 Total Bytes 64144 BPFP 0.1967 bits/point EBPFP 0.3933 equivalent bits/point MSE 2.846129 ---------------------- -------------------------------------------------------- Time: 3.407s Load: 0.009s, Pack+Encode: 1.901s, Decode+Unpack: 1.497s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 182, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8461 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample104-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample104-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample105-layer4-item1.zst (9/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample105-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 178, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 178, 128) Output shape: (1, 178, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.0.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.1.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.1.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.2.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.2.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.3.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.3.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.4.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.4.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.4.output: torch.Size([1, 178, 4096]) -> torch.Size([1, 1, 178, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,056B, BPFP=0.0463 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,032B, BPFP=0.2647 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,908B, BPFP=0.2593 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,388B, BPFP=0.2804 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,848B, BPFP=0.2567 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,684B, BPFP=0.3373 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,612B, BPFP=0.2902 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,108B, BPFP=0.3559 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,260B, BPFP=0.2309 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,680B, BPFP=0.3371 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,860B, BPFP=0.0424 ⌛️ [2/4] FRONTEND: Frontend time: 1.965s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 178, 128]) layer.0.v_cache: torch.Size([1, 8, 178, 128]) layer.1.k_cache: torch.Size([1, 8, 178, 128]) layer.1.v_cache: torch.Size([1, 8, 178, 128]) layer.2.k_cache: torch.Size([1, 8, 178, 128]) layer.2.v_cache: torch.Size([1, 8, 178, 128]) layer.3.k_cache: torch.Size([1, 8, 178, 128]) layer.3.v_cache: torch.Size([1, 8, 178, 128]) layer.4.k_cache: torch.Size([1, 8, 178, 128]) layer.4.v_cache: torch.Size([1, 8, 178, 128]) layer.4.output: torch.Size([1, 178, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.444s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 178, 128]) layer.0.v_cache: torch.Size([1, 8, 178, 128]) layer.1.k_cache: torch.Size([1, 8, 178, 128]) layer.1.v_cache: torch.Size([1, 8, 178, 128]) layer.2.k_cache: torch.Size([1, 8, 178, 128]) layer.2.v_cache: torch.Size([1, 8, 178, 128]) layer.3.k_cache: torch.Size([1, 8, 178, 128]) layer.3.v_cache: torch.Size([1, 8, 178, 128]) layer.4.k_cache: torch.Size([1, 8, 178, 128]) layer.4.v_cache: torch.Size([1, 8, 178, 128]) layer.4.output: torch.Size([1, 178, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02621927 25.27223051 layer.0.v_cache 0.00000027 0.00061290 layer.1.k_cache 0.00300007 3.87350635 layer.1.v_cache 0.00000077 0.00249385 layer.2.k_cache 0.00117703 1.49874192 layer.2.v_cache 0.00000113 0.00366159 layer.3.k_cache 0.00133116 1.77957702 layer.3.v_cache 0.00000210 0.00638096 layer.4.k_cache 0.00351662 3.68505722 layer.4.v_cache 0.00000301 0.01013641 layer.4.output 0.00016984 0.16948149 ------------------------------------------------------------------------------------- TOTAL 0.00256649 2.62930891 (elements=2,551,808) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2551808 Total Bytes 64436 BPFP 0.2020 bits/point EBPFP 0.4040 equivalent bits/point MSE 2.629309 ---------------------- -------------------------------------------------------- Time: 3.419s Load: 0.011s, Pack+Encode: 1.965s, Decode+Unpack: 1.444s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 178, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6293 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample105-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample105-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample106-layer4-item1.zst (10/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample106-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 201, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 201, 128) Output shape: (1, 201, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 201, 128]) -> torch.Size([1, 1, 201, 1024]) layer.0.v_cache: torch.Size([1, 8, 201, 128]) -> torch.Size([1, 1, 201, 1024]) layer.1.k_cache: torch.Size([1, 8, 201, 128]) -> torch.Size([1, 1, 201, 1024]) layer.1.v_cache: torch.Size([1, 8, 201, 128]) -> torch.Size([1, 1, 201, 1024]) layer.2.k_cache: torch.Size([1, 8, 201, 128]) -> torch.Size([1, 1, 201, 1024]) layer.2.v_cache: torch.Size([1, 8, 201, 128]) -> torch.Size([1, 1, 201, 1024]) layer.3.k_cache: torch.Size([1, 8, 201, 128]) -> torch.Size([1, 1, 201, 1024]) layer.3.v_cache: torch.Size([1, 8, 201, 128]) -> torch.Size([1, 1, 201, 1024]) layer.4.k_cache: torch.Size([1, 8, 201, 128]) -> torch.Size([1, 1, 201, 1024]) layer.4.v_cache: torch.Size([1, 8, 201, 128]) -> torch.Size([1, 1, 201, 1024]) layer.4.output: torch.Size([1, 201, 4096]) -> torch.Size([1, 1, 201, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,196B, BPFP=0.0465 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,224B, BPFP=0.2808 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,228B, BPFP=0.2421 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 7,676B, BPFP=0.2984 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,924B, BPFP=0.3080 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 8,380B, BPFP=0.3257 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,800B, BPFP=0.3420 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,720B, BPFP=0.3389 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,780B, BPFP=0.2247 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 10,080B, BPFP=0.3918 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,888B, BPFP=0.0378 ⌛️ [2/4] FRONTEND: Frontend time: 2.325s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 201, 128]) layer.0.v_cache: torch.Size([1, 8, 201, 128]) layer.1.k_cache: torch.Size([1, 8, 201, 128]) layer.1.v_cache: torch.Size([1, 8, 201, 128]) layer.2.k_cache: torch.Size([1, 8, 201, 128]) layer.2.v_cache: torch.Size([1, 8, 201, 128]) layer.3.k_cache: torch.Size([1, 8, 201, 128]) layer.3.v_cache: torch.Size([1, 8, 201, 128]) layer.4.k_cache: torch.Size([1, 8, 201, 128]) layer.4.v_cache: torch.Size([1, 8, 201, 128]) layer.4.output: torch.Size([1, 201, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.003s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 201, 128]) layer.0.v_cache: torch.Size([1, 8, 201, 128]) layer.1.k_cache: torch.Size([1, 8, 201, 128]) layer.1.v_cache: torch.Size([1, 8, 201, 128]) layer.2.k_cache: torch.Size([1, 8, 201, 128]) layer.2.v_cache: torch.Size([1, 8, 201, 128]) layer.3.k_cache: torch.Size([1, 8, 201, 128]) layer.3.v_cache: torch.Size([1, 8, 201, 128]) layer.4.k_cache: torch.Size([1, 8, 201, 128]) layer.4.v_cache: torch.Size([1, 8, 201, 128]) layer.4.output: torch.Size([1, 201, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02721510 32.88864515 layer.0.v_cache 0.00000027 0.00063039 layer.1.k_cache 0.00313268 2.95495059 layer.1.v_cache 0.00000087 0.00233401 layer.2.k_cache 0.00111785 1.40817109 layer.2.v_cache 0.00000101 0.00316146 layer.3.k_cache 0.00129873 1.68430051 layer.3.v_cache 0.00000193 0.00537204 layer.4.k_cache 0.00357359 3.60182607 layer.4.v_cache 0.00000284 0.00888631 layer.4.output 0.00012728 0.14701252 ------------------------------------------------------------------------------------- TOTAL 0.00263243 3.08188055 (elements=2,881,536) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2881536 Total Bytes 75896 BPFP 0.2107 bits/point EBPFP 0.4214 equivalent bits/point MSE 3.081881 ---------------------- -------------------------------------------------------- Time: 4.338s Load: 0.010s, Pack+Encode: 2.325s, Decode+Unpack: 2.003s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 201, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 201, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 3.0819 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample106-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample106-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample107-layer4-item1.zst (11/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample107-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 163, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 163, 128) Output shape: (1, 163, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.output: torch.Size([1, 163, 4096]) -> torch.Size([1, 1, 163, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,008B, BPFP=0.0483 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,536B, BPFP=0.2653 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,032B, BPFP=0.2412 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,760B, BPFP=0.2761 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,576B, BPFP=0.2673 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,812B, BPFP=0.3265 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,080B, BPFP=0.3393 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,976B, BPFP=0.3344 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,920B, BPFP=0.2358 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,996B, BPFP=0.3353 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,332B, BPFP=0.0399 ⌛️ [2/4] FRONTEND: Frontend time: 1.872s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) layer.4.output: torch.Size([1, 163, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.642s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) layer.4.output: torch.Size([1, 163, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02712815 25.78828365 layer.0.v_cache 0.00000026 0.00059614 layer.1.k_cache 0.00315731 3.22957419 layer.1.v_cache 0.00000075 0.00234751 layer.2.k_cache 0.00115274 1.47658618 layer.2.v_cache 0.00000104 0.00341979 layer.3.k_cache 0.00137819 1.76197206 layer.3.v_cache 0.00000196 0.00572379 layer.4.k_cache 0.00344312 3.41109799 layer.4.v_cache 0.00000284 0.00954739 layer.4.output 0.00020076 0.17218847 ------------------------------------------------------------------------------------- TOTAL 0.00264781 2.59842161 (elements=2,336,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2336768 Total Bytes 59028 BPFP 0.2021 bits/point EBPFP 0.4042 equivalent bits/point MSE 2.598422 ---------------------- -------------------------------------------------------- Time: 3.523s Load: 0.009s, Pack+Encode: 1.872s, Decode+Unpack: 1.642s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 163, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5984 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample107-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample107-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample108-layer4-item1.zst (12/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample108-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 163, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 163, 128) Output shape: (1, 163, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.output: torch.Size([1, 163, 4096]) -> torch.Size([1, 1, 163, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,020B, BPFP=0.0489 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,952B, BPFP=0.2853 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,012B, BPFP=0.2402 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,608B, BPFP=0.2688 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,608B, BPFP=0.2688 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,828B, BPFP=0.3273 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,152B, BPFP=0.3428 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,568B, BPFP=0.3148 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,900B, BPFP=0.2349 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,748B, BPFP=0.3234 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,476B, BPFP=0.0417 ⌛️ [2/4] FRONTEND: Frontend time: 1.914s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) layer.4.output: torch.Size([1, 163, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.425s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) layer.4.output: torch.Size([1, 163, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02704066 24.07520430 layer.0.v_cache 0.00000026 0.00062030 layer.1.k_cache 0.00309474 3.20499799 layer.1.v_cache 0.00000079 0.00248156 layer.2.k_cache 0.00116731 1.47063647 layer.2.v_cache 0.00000107 0.00359397 layer.3.k_cache 0.00138539 1.77601427 layer.3.v_cache 0.00000193 0.00582528 layer.4.k_cache 0.00353297 3.59151454 layer.4.v_cache 0.00000296 0.00996505 layer.4.output 0.00020208 0.18261812 ------------------------------------------------------------------------------------- TOTAL 0.00264546 2.49080902 (elements=2,336,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2336768 Total Bytes 58872 BPFP 0.2016 bits/point EBPFP 0.4031 equivalent bits/point MSE 2.490809 ---------------------- -------------------------------------------------------- Time: 3.350s Load: 0.011s, Pack+Encode: 1.914s, Decode+Unpack: 1.425s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 163, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.4908 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample108-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample108-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample109-layer4-item1.zst (13/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample109-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 178, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 178, 128) Output shape: (1, 178, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.0.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.1.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.1.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.2.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.2.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.3.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.3.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.4.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.4.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.4.output: torch.Size([1, 178, 4096]) -> torch.Size([1, 1, 178, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,068B, BPFP=0.0469 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,476B, BPFP=0.2403 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,704B, BPFP=0.2504 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,164B, BPFP=0.2705 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,748B, BPFP=0.2523 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,844B, BPFP=0.3443 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,560B, BPFP=0.2879 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,228B, BPFP=0.3172 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,200B, BPFP=0.2282 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,460B, BPFP=0.3274 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,192B, BPFP=0.0460 ⌛️ [2/4] FRONTEND: Frontend time: 1.917s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 178, 128]) layer.0.v_cache: torch.Size([1, 8, 178, 128]) layer.1.k_cache: torch.Size([1, 8, 178, 128]) layer.1.v_cache: torch.Size([1, 8, 178, 128]) layer.2.k_cache: torch.Size([1, 8, 178, 128]) layer.2.v_cache: torch.Size([1, 8, 178, 128]) layer.3.k_cache: torch.Size([1, 8, 178, 128]) layer.3.v_cache: torch.Size([1, 8, 178, 128]) layer.4.k_cache: torch.Size([1, 8, 178, 128]) layer.4.v_cache: torch.Size([1, 8, 178, 128]) layer.4.output: torch.Size([1, 178, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.493s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 178, 128]) layer.0.v_cache: torch.Size([1, 8, 178, 128]) layer.1.k_cache: torch.Size([1, 8, 178, 128]) layer.1.v_cache: torch.Size([1, 8, 178, 128]) layer.2.k_cache: torch.Size([1, 8, 178, 128]) layer.2.v_cache: torch.Size([1, 8, 178, 128]) layer.3.k_cache: torch.Size([1, 8, 178, 128]) layer.3.v_cache: torch.Size([1, 8, 178, 128]) layer.4.k_cache: torch.Size([1, 8, 178, 128]) layer.4.v_cache: torch.Size([1, 8, 178, 128]) layer.4.output: torch.Size([1, 178, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02815543 25.43658927 layer.0.v_cache 0.00000027 0.00061824 layer.1.k_cache 0.00311575 3.59574770 layer.1.v_cache 0.00000080 0.00259021 layer.2.k_cache 0.00115654 1.51638040 layer.2.v_cache 0.00000128 0.00380887 layer.3.k_cache 0.00138975 1.74840674 layer.3.v_cache 0.00000206 0.00616442 layer.4.k_cache 0.00349998 3.71795757 layer.4.v_cache 0.00000302 0.01039891 layer.4.output 0.00017961 0.20164363 ------------------------------------------------------------------------------------- TOTAL 0.00271738 2.63180263 (elements=2,551,808) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2551808 Total Bytes 62644 BPFP 0.1964 bits/point EBPFP 0.3928 equivalent bits/point MSE 2.631803 ---------------------- -------------------------------------------------------- Time: 3.419s Load: 0.009s, Pack+Encode: 1.917s, Decode+Unpack: 1.493s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 178, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6318 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample109-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample109-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample11-layer4-item1.zst (14/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample11-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 190, 128) Output shape: (1, 190, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.0.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.1.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.1.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.2.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.2.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.3.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.3.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.4.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.4.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.4.output: torch.Size([1, 190, 4096]) -> torch.Size([1, 1, 190, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,208B, BPFP=0.0497 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,816B, BPFP=0.2391 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,136B, BPFP=0.1701 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,048B, BPFP=0.2487 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 4,968B, BPFP=0.2043 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,236B, BPFP=0.2975 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 5,588B, BPFP=0.2298 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,356B, BPFP=0.3025 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 3,944B, BPFP=0.1622 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,468B, BPFP=0.3071 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,208B, BPFP=0.0535 ⌛️ [2/4] FRONTEND: Frontend time: 1.867s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 190, 128]) layer.0.v_cache: torch.Size([1, 8, 190, 128]) layer.1.k_cache: torch.Size([1, 8, 190, 128]) layer.1.v_cache: torch.Size([1, 8, 190, 128]) layer.2.k_cache: torch.Size([1, 8, 190, 128]) layer.2.v_cache: torch.Size([1, 8, 190, 128]) layer.3.k_cache: torch.Size([1, 8, 190, 128]) layer.3.v_cache: torch.Size([1, 8, 190, 128]) layer.4.k_cache: torch.Size([1, 8, 190, 128]) layer.4.v_cache: torch.Size([1, 8, 190, 128]) layer.4.output: torch.Size([1, 190, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.654s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 190, 128]) layer.0.v_cache: torch.Size([1, 8, 190, 128]) layer.1.k_cache: torch.Size([1, 8, 190, 128]) layer.1.v_cache: torch.Size([1, 8, 190, 128]) layer.2.k_cache: torch.Size([1, 8, 190, 128]) layer.2.v_cache: torch.Size([1, 8, 190, 128]) layer.3.k_cache: torch.Size([1, 8, 190, 128]) layer.3.v_cache: torch.Size([1, 8, 190, 128]) layer.4.k_cache: torch.Size([1, 8, 190, 128]) layer.4.v_cache: torch.Size([1, 8, 190, 128]) layer.4.output: torch.Size([1, 190, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02605901 27.11864463 layer.0.v_cache 0.00000026 0.00063212 layer.1.k_cache 0.00295386 3.09120162 layer.1.v_cache 0.00000080 0.00263695 layer.2.k_cache 0.00114499 1.54576641 layer.2.v_cache 0.00000122 0.00386014 layer.3.k_cache 0.00131775 1.77652508 layer.3.v_cache 0.00000219 0.00652439 layer.4.k_cache 0.00343004 3.48126189 layer.4.v_cache 0.00000314 0.01090597 layer.4.output 0.00018832 0.20949314 ------------------------------------------------------------------------------------- TOTAL 0.00254761 2.70542370 (elements=2,723,840) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2723840 Total Bytes 58976 BPFP 0.1732 bits/point EBPFP 0.3464 equivalent bits/point MSE 2.705424 ---------------------- -------------------------------------------------------- Time: 3.533s Load: 0.012s, Pack+Encode: 1.867s, Decode+Unpack: 1.654s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7054 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample11-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample11-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample111-layer4-item1.zst (15/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample111-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 167, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 167, 128) Output shape: (1, 167, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.0.v_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.1.k_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.1.v_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.2.k_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.2.v_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.3.k_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.3.v_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.4.k_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.4.v_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.4.output: torch.Size([1, 167, 4096]) -> torch.Size([1, 1, 167, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,020B, BPFP=0.0477 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,892B, BPFP=0.2756 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,472B, BPFP=0.2560 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,508B, BPFP=0.3045 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,580B, BPFP=0.2610 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,528B, BPFP=0.3522 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,932B, BPFP=0.3243 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,248B, BPFP=0.3391 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,208B, BPFP=0.2436 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,084B, BPFP=0.3314 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,804B, BPFP=0.0445 ⌛️ [2/4] FRONTEND: Frontend time: 1.927s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 167, 128]) layer.0.v_cache: torch.Size([1, 8, 167, 128]) layer.1.k_cache: torch.Size([1, 8, 167, 128]) layer.1.v_cache: torch.Size([1, 8, 167, 128]) layer.2.k_cache: torch.Size([1, 8, 167, 128]) layer.2.v_cache: torch.Size([1, 8, 167, 128]) layer.3.k_cache: torch.Size([1, 8, 167, 128]) layer.3.v_cache: torch.Size([1, 8, 167, 128]) layer.4.k_cache: torch.Size([1, 8, 167, 128]) layer.4.v_cache: torch.Size([1, 8, 167, 128]) layer.4.output: torch.Size([1, 167, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.423s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 167, 128]) layer.0.v_cache: torch.Size([1, 8, 167, 128]) layer.1.k_cache: torch.Size([1, 8, 167, 128]) layer.1.v_cache: torch.Size([1, 8, 167, 128]) layer.2.k_cache: torch.Size([1, 8, 167, 128]) layer.2.v_cache: torch.Size([1, 8, 167, 128]) layer.3.k_cache: torch.Size([1, 8, 167, 128]) layer.3.v_cache: torch.Size([1, 8, 167, 128]) layer.4.k_cache: torch.Size([1, 8, 167, 128]) layer.4.v_cache: torch.Size([1, 8, 167, 128]) layer.4.output: torch.Size([1, 167, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02739838 25.38935021 layer.0.v_cache 0.00000026 0.00061714 layer.1.k_cache 0.00313824 3.68575045 layer.1.v_cache 0.00000083 0.00252354 layer.2.k_cache 0.00117185 1.46945793 layer.2.v_cache 0.00000105 0.00354467 layer.3.k_cache 0.00137006 1.77347606 layer.3.v_cache 0.00000199 0.00579769 layer.4.k_cache 0.00340675 3.56957131 layer.4.v_cache 0.00000299 0.01010053 layer.4.output 0.00021989 0.19990199 ------------------------------------------------------------------------------------- TOTAL 0.00266943 2.62212839 (elements=2,394,112) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2394112 Total Bytes 62276 BPFP 0.2081 bits/point EBPFP 0.4162 equivalent bits/point MSE 2.622128 ---------------------- -------------------------------------------------------- Time: 3.359s Load: 0.010s, Pack+Encode: 1.927s, Decode+Unpack: 1.423s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 167, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6221 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample111-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample111-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample112-layer4-item1.zst (16/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample112-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 164, 128) Output shape: (1, 164, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.output: torch.Size([1, 164, 4096]) -> torch.Size([1, 1, 164, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,028B, BPFP=0.0490 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,580B, BPFP=0.2658 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,412B, BPFP=0.2578 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,144B, BPFP=0.2927 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,840B, BPFP=0.2782 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,088B, BPFP=0.3377 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,892B, BPFP=0.3283 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,156B, BPFP=0.3409 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,032B, BPFP=0.2397 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,760B, BPFP=0.3220 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,384B, BPFP=0.0403 ⌛️ [2/4] FRONTEND: Frontend time: 1.958s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) layer.4.output: torch.Size([1, 164, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.526s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) layer.4.output: torch.Size([1, 164, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02765335 26.51921565 layer.0.v_cache 0.00000026 0.00059028 layer.1.k_cache 0.00306024 3.36572675 layer.1.v_cache 0.00000075 0.00238625 layer.2.k_cache 0.00113567 1.48715461 layer.2.v_cache 0.00000102 0.00340224 layer.3.k_cache 0.00140214 1.70929420 layer.3.v_cache 0.00000194 0.00568453 layer.4.k_cache 0.00347389 3.45669109 layer.4.v_cache 0.00000286 0.00965610 layer.4.output 0.00016775 0.15874716 ------------------------------------------------------------------------------------- TOTAL 0.00267165 2.65677074 (elements=2,351,104) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2351104 Total Bytes 60316 BPFP 0.2052 bits/point EBPFP 0.4105 equivalent bits/point MSE 2.656771 ---------------------- -------------------------------------------------------- Time: 3.493s Load: 0.009s, Pack+Encode: 1.958s, Decode+Unpack: 1.526s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6568 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample112-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample112-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample12-layer4-item1.zst (17/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample12-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 204, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 204, 128) Output shape: (1, 204, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.0.v_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.1.k_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.1.v_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.2.k_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.2.v_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.3.k_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.3.v_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.4.k_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.4.v_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.4.output: torch.Size([1, 204, 4096]) -> torch.Size([1, 1, 204, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,212B, BPFP=0.0464 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,100B, BPFP=0.2719 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,604B, BPFP=0.2529 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 7,804B, BPFP=0.2989 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,776B, BPFP=0.2978 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 9,436B, BPFP=0.3614 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,584B, BPFP=0.3287 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 9,196B, BPFP=0.3522 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,056B, BPFP=0.2319 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 9,544B, BPFP=0.3655 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,344B, BPFP=0.0416 ⌛️ [2/4] FRONTEND: Frontend time: 2.224s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 204, 128]) layer.0.v_cache: torch.Size([1, 8, 204, 128]) layer.1.k_cache: torch.Size([1, 8, 204, 128]) layer.1.v_cache: torch.Size([1, 8, 204, 128]) layer.2.k_cache: torch.Size([1, 8, 204, 128]) layer.2.v_cache: torch.Size([1, 8, 204, 128]) layer.3.k_cache: torch.Size([1, 8, 204, 128]) layer.3.v_cache: torch.Size([1, 8, 204, 128]) layer.4.k_cache: torch.Size([1, 8, 204, 128]) layer.4.v_cache: torch.Size([1, 8, 204, 128]) layer.4.output: torch.Size([1, 204, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.112s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 204, 128]) layer.0.v_cache: torch.Size([1, 8, 204, 128]) layer.1.k_cache: torch.Size([1, 8, 204, 128]) layer.1.v_cache: torch.Size([1, 8, 204, 128]) layer.2.k_cache: torch.Size([1, 8, 204, 128]) layer.2.v_cache: torch.Size([1, 8, 204, 128]) layer.3.k_cache: torch.Size([1, 8, 204, 128]) layer.3.v_cache: torch.Size([1, 8, 204, 128]) layer.4.k_cache: torch.Size([1, 8, 204, 128]) layer.4.v_cache: torch.Size([1, 8, 204, 128]) layer.4.output: torch.Size([1, 204, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02692792 32.90141506 layer.0.v_cache 0.00000026 0.00059969 layer.1.k_cache 0.00296634 3.18932387 layer.1.v_cache 0.00000076 0.00245946 layer.2.k_cache 0.00113428 1.46317441 layer.2.v_cache 0.00000108 0.00353318 layer.3.k_cache 0.00134079 1.66904614 layer.3.v_cache 0.00000199 0.00587136 layer.4.k_cache 0.00357197 3.46924606 layer.4.v_cache 0.00000302 0.00989768 layer.4.output 0.00017435 0.19108660 ------------------------------------------------------------------------------------- TOTAL 0.00261756 3.10563667 (elements=2,924,544) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2924544 Total Bytes 77656 BPFP 0.2124 bits/point EBPFP 0.4249 equivalent bits/point MSE 3.105637 ---------------------- -------------------------------------------------------- Time: 4.347s Load: 0.011s, Pack+Encode: 2.224s, Decode+Unpack: 2.112s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 204, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 3.1056 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample12-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample12-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample13-layer4-item1.zst (18/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample13-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 207, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 207, 128) Output shape: (1, 207, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.0.v_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.1.k_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.1.v_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.2.k_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.2.v_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.3.k_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.3.v_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.4.k_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.4.v_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.4.output: torch.Size([1, 207, 4096]) -> torch.Size([1, 1, 207, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,228B, BPFP=0.0463 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,344B, BPFP=0.2772 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,744B, BPFP=0.2545 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 8,756B, BPFP=0.3305 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,788B, BPFP=0.2939 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 9,908B, BPFP=0.3739 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,632B, BPFP=0.3258 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 10,204B, BPFP=0.3851 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,508B, BPFP=0.2456 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 10,464B, BPFP=0.3949 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,624B, BPFP=0.0436 ⌛️ [2/4] FRONTEND: Frontend time: 2.195s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 207, 128]) layer.0.v_cache: torch.Size([1, 8, 207, 128]) layer.1.k_cache: torch.Size([1, 8, 207, 128]) layer.1.v_cache: torch.Size([1, 8, 207, 128]) layer.2.k_cache: torch.Size([1, 8, 207, 128]) layer.2.v_cache: torch.Size([1, 8, 207, 128]) layer.3.k_cache: torch.Size([1, 8, 207, 128]) layer.3.v_cache: torch.Size([1, 8, 207, 128]) layer.4.k_cache: torch.Size([1, 8, 207, 128]) layer.4.v_cache: torch.Size([1, 8, 207, 128]) layer.4.output: torch.Size([1, 207, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.964s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 207, 128]) layer.0.v_cache: torch.Size([1, 8, 207, 128]) layer.1.k_cache: torch.Size([1, 8, 207, 128]) layer.1.v_cache: torch.Size([1, 8, 207, 128]) layer.2.k_cache: torch.Size([1, 8, 207, 128]) layer.2.v_cache: torch.Size([1, 8, 207, 128]) layer.3.k_cache: torch.Size([1, 8, 207, 128]) layer.3.v_cache: torch.Size([1, 8, 207, 128]) layer.4.k_cache: torch.Size([1, 8, 207, 128]) layer.4.v_cache: torch.Size([1, 8, 207, 128]) layer.4.output: torch.Size([1, 207, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02640885 28.69415431 layer.0.v_cache 0.00000027 0.00061132 layer.1.k_cache 0.00296321 3.42251808 layer.1.v_cache 0.00000080 0.00251093 layer.2.k_cache 0.00115200 1.46201225 layer.2.v_cache 0.00000114 0.00368907 layer.3.k_cache 0.00135178 1.72303809 layer.3.v_cache 0.00000205 0.00615633 layer.4.k_cache 0.00347904 3.44527564 layer.4.v_cache 0.00000310 0.01038468 layer.4.output 0.00016949 0.18275219 ------------------------------------------------------------------------------------- TOTAL 0.00257430 2.82152568 (elements=2,967,552) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2967552 Total Bytes 82200 BPFP 0.2216 bits/point EBPFP 0.4432 equivalent bits/point MSE 2.821526 ---------------------- -------------------------------------------------------- Time: 4.170s Load: 0.012s, Pack+Encode: 2.195s, Decode+Unpack: 1.964s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 207, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8215 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample13-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample13-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample14-layer4-item1.zst (19/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample14-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 192, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 192, 128) Output shape: (1, 192, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 192, 128]) -> torch.Size([1, 1, 192, 1024]) layer.0.v_cache: torch.Size([1, 8, 192, 128]) -> torch.Size([1, 1, 192, 1024]) layer.1.k_cache: torch.Size([1, 8, 192, 128]) -> torch.Size([1, 1, 192, 1024]) layer.1.v_cache: torch.Size([1, 8, 192, 128]) -> torch.Size([1, 1, 192, 1024]) layer.2.k_cache: torch.Size([1, 8, 192, 128]) -> torch.Size([1, 1, 192, 1024]) layer.2.v_cache: torch.Size([1, 8, 192, 128]) -> torch.Size([1, 1, 192, 1024]) layer.3.k_cache: torch.Size([1, 8, 192, 128]) -> torch.Size([1, 1, 192, 1024]) layer.3.v_cache: torch.Size([1, 8, 192, 128]) -> torch.Size([1, 1, 192, 1024]) layer.4.k_cache: torch.Size([1, 8, 192, 128]) -> torch.Size([1, 1, 192, 1024]) layer.4.v_cache: torch.Size([1, 8, 192, 128]) -> torch.Size([1, 1, 192, 1024]) layer.4.output: torch.Size([1, 192, 4096]) -> torch.Size([1, 1, 192, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 880B, BPFP=0.0358 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 4,256B, BPFP=0.1732 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 3,596B, BPFP=0.1463 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,708B, BPFP=0.2323 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 4,280B, BPFP=0.1742 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,212B, BPFP=0.2528 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 4,804B, BPFP=0.1955 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,692B, BPFP=0.2723 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 3,512B, BPFP=0.1429 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,680B, BPFP=0.2718 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,012B, BPFP=0.0306 ⌛️ [2/4] FRONTEND: Frontend time: 1.919s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 192, 128]) layer.0.v_cache: torch.Size([1, 8, 192, 128]) layer.1.k_cache: torch.Size([1, 8, 192, 128]) layer.1.v_cache: torch.Size([1, 8, 192, 128]) layer.2.k_cache: torch.Size([1, 8, 192, 128]) layer.2.v_cache: torch.Size([1, 8, 192, 128]) layer.3.k_cache: torch.Size([1, 8, 192, 128]) layer.3.v_cache: torch.Size([1, 8, 192, 128]) layer.4.k_cache: torch.Size([1, 8, 192, 128]) layer.4.v_cache: torch.Size([1, 8, 192, 128]) layer.4.output: torch.Size([1, 192, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.423s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 192, 128]) layer.0.v_cache: torch.Size([1, 8, 192, 128]) layer.1.k_cache: torch.Size([1, 8, 192, 128]) layer.1.v_cache: torch.Size([1, 8, 192, 128]) layer.2.k_cache: torch.Size([1, 8, 192, 128]) layer.2.v_cache: torch.Size([1, 8, 192, 128]) layer.3.k_cache: torch.Size([1, 8, 192, 128]) layer.3.v_cache: torch.Size([1, 8, 192, 128]) layer.4.k_cache: torch.Size([1, 8, 192, 128]) layer.4.v_cache: torch.Size([1, 8, 192, 128]) layer.4.output: torch.Size([1, 192, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02717149 29.21301524 layer.0.v_cache 0.00000027 0.00059245 layer.1.k_cache 0.00306874 2.60159572 layer.1.v_cache 0.00000081 0.00262915 layer.2.k_cache 0.00115207 1.36086702 layer.2.v_cache 0.00000118 0.00363913 layer.3.k_cache 0.00133924 1.59003099 layer.3.v_cache 0.00000214 0.00622380 layer.4.k_cache 0.00342495 2.91896693 layer.4.v_cache 0.00000316 0.01049546 layer.4.output 0.00018805 0.18585181 ------------------------------------------------------------------------------------- TOTAL 0.00263687 2.74653308 (elements=2,752,512) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2752512 Total Bytes 49632 BPFP 0.1443 bits/point EBPFP 0.2885 equivalent bits/point MSE 2.746533 ---------------------- -------------------------------------------------------- Time: 3.352s Load: 0.010s, Pack+Encode: 1.919s, Decode+Unpack: 1.423s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 192, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 192, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7465 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample14-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample14-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample16-layer4-item1.zst (20/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample16-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 207, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.013s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 207, 128) Output shape: (1, 207, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.0.v_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.1.k_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.1.v_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.2.k_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.2.v_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.3.k_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.3.v_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.4.k_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.4.v_cache: torch.Size([1, 8, 207, 128]) -> torch.Size([1, 1, 207, 1024]) layer.4.output: torch.Size([1, 207, 4096]) -> torch.Size([1, 1, 207, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,224B, BPFP=0.0462 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,412B, BPFP=0.2797 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,616B, BPFP=0.2497 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 8,532B, BPFP=0.3220 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,904B, BPFP=0.2983 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 10,120B, BPFP=0.3819 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,376B, BPFP=0.3161 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 10,012B, BPFP=0.3779 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,244B, BPFP=0.2357 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 10,440B, BPFP=0.3940 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,724B, BPFP=0.0540 ⌛️ [2/4] FRONTEND: Frontend time: 2.357s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 207, 128]) layer.0.v_cache: torch.Size([1, 8, 207, 128]) layer.1.k_cache: torch.Size([1, 8, 207, 128]) layer.1.v_cache: torch.Size([1, 8, 207, 128]) layer.2.k_cache: torch.Size([1, 8, 207, 128]) layer.2.v_cache: torch.Size([1, 8, 207, 128]) layer.3.k_cache: torch.Size([1, 8, 207, 128]) layer.3.v_cache: torch.Size([1, 8, 207, 128]) layer.4.k_cache: torch.Size([1, 8, 207, 128]) layer.4.v_cache: torch.Size([1, 8, 207, 128]) layer.4.output: torch.Size([1, 207, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.847s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 207, 128]) layer.0.v_cache: torch.Size([1, 8, 207, 128]) layer.1.k_cache: torch.Size([1, 8, 207, 128]) layer.1.v_cache: torch.Size([1, 8, 207, 128]) layer.2.k_cache: torch.Size([1, 8, 207, 128]) layer.2.v_cache: torch.Size([1, 8, 207, 128]) layer.3.k_cache: torch.Size([1, 8, 207, 128]) layer.3.v_cache: torch.Size([1, 8, 207, 128]) layer.4.k_cache: torch.Size([1, 8, 207, 128]) layer.4.v_cache: torch.Size([1, 8, 207, 128]) layer.4.output: torch.Size([1, 207, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02737031 30.00682178 layer.0.v_cache 0.00000026 0.00063986 layer.1.k_cache 0.00295586 3.34284184 layer.1.v_cache 0.00000087 0.00258704 layer.2.k_cache 0.00115404 1.54978405 layer.2.v_cache 0.00000118 0.00377942 layer.3.k_cache 0.00130266 1.71052824 layer.3.v_cache 0.00000222 0.00641868 layer.4.k_cache 0.00345388 3.70897406 layer.4.v_cache 0.00000327 0.01084382 layer.4.output 0.00018665 0.19416848 ------------------------------------------------------------------------------------- TOTAL 0.00264223 2.93713519 (elements=2,967,552) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2967552 Total Bytes 82604 BPFP 0.2227 bits/point EBPFP 0.4454 equivalent bits/point MSE 2.937135 ---------------------- -------------------------------------------------------- Time: 4.216s Load: 0.013s, Pack+Encode: 2.357s, Decode+Unpack: 1.847s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 207, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 207, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.9371 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample16-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample16-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample17-layer4-item1.zst (21/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample17-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 213, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 213, 128) Output shape: (1, 213, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.0.v_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.1.k_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.1.v_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.2.k_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.2.v_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.3.k_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.3.v_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.4.k_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.4.v_cache: torch.Size([1, 8, 213, 128]) -> torch.Size([1, 1, 213, 1024]) layer.4.output: torch.Size([1, 213, 4096]) -> torch.Size([1, 1, 213, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,204B, BPFP=0.0442 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,212B, BPFP=0.2645 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,536B, BPFP=0.2397 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 8,680B, BPFP=0.3184 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,724B, BPFP=0.2833 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 9,332B, BPFP=0.3423 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,504B, BPFP=0.3119 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 9,604B, BPFP=0.3523 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,576B, BPFP=0.2412 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 9,680B, BPFP=0.3550 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,480B, BPFP=0.0411 ⌛️ [2/4] FRONTEND: Frontend time: 2.297s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 213, 128]) layer.0.v_cache: torch.Size([1, 8, 213, 128]) layer.1.k_cache: torch.Size([1, 8, 213, 128]) layer.1.v_cache: torch.Size([1, 8, 213, 128]) layer.2.k_cache: torch.Size([1, 8, 213, 128]) layer.2.v_cache: torch.Size([1, 8, 213, 128]) layer.3.k_cache: torch.Size([1, 8, 213, 128]) layer.3.v_cache: torch.Size([1, 8, 213, 128]) layer.4.k_cache: torch.Size([1, 8, 213, 128]) layer.4.v_cache: torch.Size([1, 8, 213, 128]) layer.4.output: torch.Size([1, 213, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.987s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 213, 128]) layer.0.v_cache: torch.Size([1, 8, 213, 128]) layer.1.k_cache: torch.Size([1, 8, 213, 128]) layer.1.v_cache: torch.Size([1, 8, 213, 128]) layer.2.k_cache: torch.Size([1, 8, 213, 128]) layer.2.v_cache: torch.Size([1, 8, 213, 128]) layer.3.k_cache: torch.Size([1, 8, 213, 128]) layer.3.v_cache: torch.Size([1, 8, 213, 128]) layer.4.k_cache: torch.Size([1, 8, 213, 128]) layer.4.v_cache: torch.Size([1, 8, 213, 128]) layer.4.output: torch.Size([1, 213, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02567197 29.66439261 layer.0.v_cache 0.00000028 0.00062127 layer.1.k_cache 0.00297252 3.29295442 layer.1.v_cache 0.00000081 0.00252556 layer.2.k_cache 0.00119374 1.44891830 layer.2.v_cache 0.00000110 0.00361248 layer.3.k_cache 0.00130971 1.69372530 layer.3.v_cache 0.00000208 0.00599021 layer.4.k_cache 0.00366947 3.43863302 layer.4.v_cache 0.00000305 0.01029903 layer.4.output 0.00013440 0.18797820 ------------------------------------------------------------------------------------- TOTAL 0.00252588 2.87954179 (elements=3,053,568) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 3053568 Total Bytes 79532 BPFP 0.2084 bits/point EBPFP 0.4167 equivalent bits/point MSE 2.879542 ---------------------- -------------------------------------------------------- Time: 4.294s Load: 0.011s, Pack+Encode: 2.297s, Decode+Unpack: 1.987s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 213, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 213, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8795 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample17-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample17-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample18-layer4-item1.zst (22/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample18-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 191, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 191, 128) Output shape: (1, 191, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.0.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.1.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.1.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.2.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.2.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.3.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.3.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.4.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.4.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.4.output: torch.Size([1, 191, 4096]) -> torch.Size([1, 1, 191, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,064B, BPFP=0.0435 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,676B, BPFP=0.2322 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 3,904B, BPFP=0.1597 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,964B, BPFP=0.2439 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 4,476B, BPFP=0.1831 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,200B, BPFP=0.2945 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 5,136B, BPFP=0.2101 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,420B, BPFP=0.3035 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 3,724B, BPFP=0.1523 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,192B, BPFP=0.2942 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,652B, BPFP=0.0373 ⌛️ [2/4] FRONTEND: Frontend time: 1.871s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 191, 128]) layer.0.v_cache: torch.Size([1, 8, 191, 128]) layer.1.k_cache: torch.Size([1, 8, 191, 128]) layer.1.v_cache: torch.Size([1, 8, 191, 128]) layer.2.k_cache: torch.Size([1, 8, 191, 128]) layer.2.v_cache: torch.Size([1, 8, 191, 128]) layer.3.k_cache: torch.Size([1, 8, 191, 128]) layer.3.v_cache: torch.Size([1, 8, 191, 128]) layer.4.k_cache: torch.Size([1, 8, 191, 128]) layer.4.v_cache: torch.Size([1, 8, 191, 128]) layer.4.output: torch.Size([1, 191, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.614s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 191, 128]) layer.0.v_cache: torch.Size([1, 8, 191, 128]) layer.1.k_cache: torch.Size([1, 8, 191, 128]) layer.1.v_cache: torch.Size([1, 8, 191, 128]) layer.2.k_cache: torch.Size([1, 8, 191, 128]) layer.2.v_cache: torch.Size([1, 8, 191, 128]) layer.3.k_cache: torch.Size([1, 8, 191, 128]) layer.3.v_cache: torch.Size([1, 8, 191, 128]) layer.4.k_cache: torch.Size([1, 8, 191, 128]) layer.4.v_cache: torch.Size([1, 8, 191, 128]) layer.4.output: torch.Size([1, 191, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02670301 28.73522885 layer.0.v_cache 0.00000027 0.00061608 layer.1.k_cache 0.00305067 2.74176217 layer.1.v_cache 0.00000084 0.00259535 layer.2.k_cache 0.00117351 1.38553863 layer.2.v_cache 0.00000111 0.00373825 layer.3.k_cache 0.00132215 1.62338784 layer.3.v_cache 0.00000209 0.00613487 layer.4.k_cache 0.00346212 3.09364135 layer.4.v_cache 0.00000303 0.01031983 layer.4.output 0.00018565 0.20928456 ------------------------------------------------------------------------------------- TOTAL 0.00260439 2.74572153 (elements=2,738,176) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2738176 Total Bytes 55408 BPFP 0.1619 bits/point EBPFP 0.3238 equivalent bits/point MSE 2.745722 ---------------------- -------------------------------------------------------- Time: 3.496s Load: 0.011s, Pack+Encode: 1.871s, Decode+Unpack: 1.614s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 191, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7457 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample18-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample18-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample19-layer4-item1.zst (23/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample19-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 194, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 194, 128) Output shape: (1, 194, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 194, 128]) -> torch.Size([1, 1, 194, 1024]) layer.0.v_cache: torch.Size([1, 8, 194, 128]) -> torch.Size([1, 1, 194, 1024]) layer.1.k_cache: torch.Size([1, 8, 194, 128]) -> torch.Size([1, 1, 194, 1024]) layer.1.v_cache: torch.Size([1, 8, 194, 128]) -> torch.Size([1, 1, 194, 1024]) layer.2.k_cache: torch.Size([1, 8, 194, 128]) -> torch.Size([1, 1, 194, 1024]) layer.2.v_cache: torch.Size([1, 8, 194, 128]) -> torch.Size([1, 1, 194, 1024]) layer.3.k_cache: torch.Size([1, 8, 194, 128]) -> torch.Size([1, 1, 194, 1024]) layer.3.v_cache: torch.Size([1, 8, 194, 128]) -> torch.Size([1, 1, 194, 1024]) layer.4.k_cache: torch.Size([1, 8, 194, 128]) -> torch.Size([1, 1, 194, 1024]) layer.4.v_cache: torch.Size([1, 8, 194, 128]) -> torch.Size([1, 1, 194, 1024]) layer.4.output: torch.Size([1, 194, 4096]) -> torch.Size([1, 1, 194, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,156B, BPFP=0.0466 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,536B, BPFP=0.2632 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,064B, BPFP=0.2442 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 7,180B, BPFP=0.2891 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,184B, BPFP=0.2893 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 8,704B, BPFP=0.3505 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,120B, BPFP=0.3270 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,792B, BPFP=0.3541 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,860B, BPFP=0.2360 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 9,344B, BPFP=0.3763 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,284B, BPFP=0.0431 ⌛️ [2/4] FRONTEND: Frontend time: 2.261s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 194, 128]) layer.0.v_cache: torch.Size([1, 8, 194, 128]) layer.1.k_cache: torch.Size([1, 8, 194, 128]) layer.1.v_cache: torch.Size([1, 8, 194, 128]) layer.2.k_cache: torch.Size([1, 8, 194, 128]) layer.2.v_cache: torch.Size([1, 8, 194, 128]) layer.3.k_cache: torch.Size([1, 8, 194, 128]) layer.3.v_cache: torch.Size([1, 8, 194, 128]) layer.4.k_cache: torch.Size([1, 8, 194, 128]) layer.4.v_cache: torch.Size([1, 8, 194, 128]) layer.4.output: torch.Size([1, 194, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.851s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 194, 128]) layer.0.v_cache: torch.Size([1, 8, 194, 128]) layer.1.k_cache: torch.Size([1, 8, 194, 128]) layer.1.v_cache: torch.Size([1, 8, 194, 128]) layer.2.k_cache: torch.Size([1, 8, 194, 128]) layer.2.v_cache: torch.Size([1, 8, 194, 128]) layer.3.k_cache: torch.Size([1, 8, 194, 128]) layer.3.v_cache: torch.Size([1, 8, 194, 128]) layer.4.k_cache: torch.Size([1, 8, 194, 128]) layer.4.v_cache: torch.Size([1, 8, 194, 128]) layer.4.output: torch.Size([1, 194, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02707747 27.27690832 layer.0.v_cache 0.00000027 0.00061094 layer.1.k_cache 0.00295501 3.19923857 layer.1.v_cache 0.00000076 0.00251101 layer.2.k_cache 0.00116763 1.48842173 layer.2.v_cache 0.00000111 0.00375837 layer.3.k_cache 0.00133119 1.71387993 layer.3.v_cache 0.00000207 0.00607027 layer.4.k_cache 0.00346189 3.29251004 layer.4.v_cache 0.00000304 0.01025723 layer.4.output 0.00015844 0.18536558 ------------------------------------------------------------------------------------- TOTAL 0.00261673 2.69540205 (elements=2,781,184) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2781184 Total Bytes 73224 BPFP 0.2106 bits/point EBPFP 0.4213 equivalent bits/point MSE 2.695402 ---------------------- -------------------------------------------------------- Time: 4.122s Load: 0.010s, Pack+Encode: 2.261s, Decode+Unpack: 1.851s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 194, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 194, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6954 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample19-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample19-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample2-layer4-item1.zst (24/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample2-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 241, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 241, 128) Output shape: (1, 241, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 241, 128]) -> torch.Size([1, 1, 241, 1024]) layer.0.v_cache: torch.Size([1, 8, 241, 128]) -> torch.Size([1, 1, 241, 1024]) layer.1.k_cache: torch.Size([1, 8, 241, 128]) -> torch.Size([1, 1, 241, 1024]) layer.1.v_cache: torch.Size([1, 8, 241, 128]) -> torch.Size([1, 1, 241, 1024]) layer.2.k_cache: torch.Size([1, 8, 241, 128]) -> torch.Size([1, 1, 241, 1024]) layer.2.v_cache: torch.Size([1, 8, 241, 128]) -> torch.Size([1, 1, 241, 1024]) layer.3.k_cache: torch.Size([1, 8, 241, 128]) -> torch.Size([1, 1, 241, 1024]) layer.3.v_cache: torch.Size([1, 8, 241, 128]) -> torch.Size([1, 1, 241, 1024]) layer.4.k_cache: torch.Size([1, 8, 241, 128]) -> torch.Size([1, 1, 241, 1024]) layer.4.v_cache: torch.Size([1, 8, 241, 128]) -> torch.Size([1, 1, 241, 1024]) layer.4.output: torch.Size([1, 241, 4096]) -> torch.Size([1, 1, 241, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,316B, BPFP=0.0427 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,256B, BPFP=0.2352 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,856B, BPFP=0.2223 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 8,716B, BPFP=0.2825 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,108B, BPFP=0.2304 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 10,440B, BPFP=0.3384 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 9,260B, BPFP=0.3002 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 9,884B, BPFP=0.3204 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,488B, BPFP=0.2103 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 10,100B, BPFP=0.3274 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,836B, BPFP=0.0392 ⌛️ [2/4] FRONTEND: Frontend time: 2.397s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 241, 128]) layer.0.v_cache: torch.Size([1, 8, 241, 128]) layer.1.k_cache: torch.Size([1, 8, 241, 128]) layer.1.v_cache: torch.Size([1, 8, 241, 128]) layer.2.k_cache: torch.Size([1, 8, 241, 128]) layer.2.v_cache: torch.Size([1, 8, 241, 128]) layer.3.k_cache: torch.Size([1, 8, 241, 128]) layer.3.v_cache: torch.Size([1, 8, 241, 128]) layer.4.k_cache: torch.Size([1, 8, 241, 128]) layer.4.v_cache: torch.Size([1, 8, 241, 128]) layer.4.output: torch.Size([1, 241, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.809s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 241, 128]) layer.0.v_cache: torch.Size([1, 8, 241, 128]) layer.1.k_cache: torch.Size([1, 8, 241, 128]) layer.1.v_cache: torch.Size([1, 8, 241, 128]) layer.2.k_cache: torch.Size([1, 8, 241, 128]) layer.2.v_cache: torch.Size([1, 8, 241, 128]) layer.3.k_cache: torch.Size([1, 8, 241, 128]) layer.3.v_cache: torch.Size([1, 8, 241, 128]) layer.4.k_cache: torch.Size([1, 8, 241, 128]) layer.4.v_cache: torch.Size([1, 8, 241, 128]) layer.4.output: torch.Size([1, 241, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02558961 25.21669760 layer.0.v_cache 0.00000026 0.00061425 layer.1.k_cache 0.00297512 3.31307540 layer.1.v_cache 0.00000081 0.00263998 layer.2.k_cache 0.00122557 1.46943911 layer.2.v_cache 0.00000115 0.00397969 layer.3.k_cache 0.00132258 1.78377080 layer.3.v_cache 0.00000273 0.00650339 layer.4.k_cache 0.00351858 3.46995703 layer.4.v_cache 0.00000307 0.01061363 layer.4.output 0.00016366 0.19255780 ------------------------------------------------------------------------------------- TOTAL 0.00252101 2.57482300 (elements=3,454,976) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 3454976 Total Bytes 82260 BPFP 0.1905 bits/point EBPFP 0.3809 equivalent bits/point MSE 2.574823 ---------------------- -------------------------------------------------------- Time: 4.219s Load: 0.012s, Pack+Encode: 2.397s, Decode+Unpack: 1.809s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 241, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 241, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5748 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample2-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample2-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample20-layer4-item1.zst (25/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample20-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 197, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 197, 128) Output shape: (1, 197, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.0.v_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.1.k_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.1.v_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.2.k_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.2.v_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.3.k_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.3.v_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.4.k_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.4.v_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.4.output: torch.Size([1, 197, 4096]) -> torch.Size([1, 1, 197, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,212B, BPFP=0.0481 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,604B, BPFP=0.2619 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,804B, BPFP=0.2302 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 7,124B, BPFP=0.2825 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,372B, BPFP=0.2924 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 9,532B, BPFP=0.3780 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,632B, BPFP=0.3423 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,860B, BPFP=0.3514 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,396B, BPFP=0.2536 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 9,796B, BPFP=0.3885 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,032B, BPFP=0.0400 ⌛️ [2/4] FRONTEND: Frontend time: 2.386s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 197, 128]) layer.0.v_cache: torch.Size([1, 8, 197, 128]) layer.1.k_cache: torch.Size([1, 8, 197, 128]) layer.1.v_cache: torch.Size([1, 8, 197, 128]) layer.2.k_cache: torch.Size([1, 8, 197, 128]) layer.2.v_cache: torch.Size([1, 8, 197, 128]) layer.3.k_cache: torch.Size([1, 8, 197, 128]) layer.3.v_cache: torch.Size([1, 8, 197, 128]) layer.4.k_cache: torch.Size([1, 8, 197, 128]) layer.4.v_cache: torch.Size([1, 8, 197, 128]) layer.4.output: torch.Size([1, 197, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.860s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 197, 128]) layer.0.v_cache: torch.Size([1, 8, 197, 128]) layer.1.k_cache: torch.Size([1, 8, 197, 128]) layer.1.v_cache: torch.Size([1, 8, 197, 128]) layer.2.k_cache: torch.Size([1, 8, 197, 128]) layer.2.v_cache: torch.Size([1, 8, 197, 128]) layer.3.k_cache: torch.Size([1, 8, 197, 128]) layer.3.v_cache: torch.Size([1, 8, 197, 128]) layer.4.k_cache: torch.Size([1, 8, 197, 128]) layer.4.v_cache: torch.Size([1, 8, 197, 128]) layer.4.output: torch.Size([1, 197, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02755299 33.68636977 layer.0.v_cache 0.00000026 0.00061411 layer.1.k_cache 0.00304278 3.20610643 layer.1.v_cache 0.00000080 0.00249860 layer.2.k_cache 0.00114678 1.48079964 layer.2.v_cache 0.00000109 0.00356314 layer.3.k_cache 0.00134691 1.69354589 layer.3.v_cache 0.00000250 0.00592007 layer.4.k_cache 0.00338668 3.44370514 layer.4.v_cache 0.00000296 0.00988505 layer.4.output 0.00017873 0.18799763 ------------------------------------------------------------------------------------- TOTAL 0.00265705 3.16321417 (elements=2,824,192) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2824192 Total Bytes 75364 BPFP 0.2135 bits/point EBPFP 0.4270 equivalent bits/point MSE 3.163214 ---------------------- -------------------------------------------------------- Time: 4.256s Load: 0.011s, Pack+Encode: 2.386s, Decode+Unpack: 1.860s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 197, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 3.1632 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample20-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample20-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample21-layer4-item1.zst (26/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample21-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 182, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 182, 128) Output shape: (1, 182, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.0.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.1.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.1.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.2.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.2.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.3.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.3.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.4.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.4.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.4.output: torch.Size([1, 182, 4096]) -> torch.Size([1, 1, 182, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,156B, BPFP=0.0496 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,992B, BPFP=0.2572 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,176B, BPFP=0.2222 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 7,140B, BPFP=0.3065 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,640B, BPFP=0.2421 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,064B, BPFP=0.3032 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,384B, BPFP=0.2740 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,560B, BPFP=0.3245 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,700B, BPFP=0.2018 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,748B, BPFP=0.3326 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,668B, BPFP=0.0501 ⌛️ [2/4] FRONTEND: Frontend time: 1.881s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 182, 128]) layer.0.v_cache: torch.Size([1, 8, 182, 128]) layer.1.k_cache: torch.Size([1, 8, 182, 128]) layer.1.v_cache: torch.Size([1, 8, 182, 128]) layer.2.k_cache: torch.Size([1, 8, 182, 128]) layer.2.v_cache: torch.Size([1, 8, 182, 128]) layer.3.k_cache: torch.Size([1, 8, 182, 128]) layer.3.v_cache: torch.Size([1, 8, 182, 128]) layer.4.k_cache: torch.Size([1, 8, 182, 128]) layer.4.v_cache: torch.Size([1, 8, 182, 128]) layer.4.output: torch.Size([1, 182, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.663s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 182, 128]) layer.0.v_cache: torch.Size([1, 8, 182, 128]) layer.1.k_cache: torch.Size([1, 8, 182, 128]) layer.1.v_cache: torch.Size([1, 8, 182, 128]) layer.2.k_cache: torch.Size([1, 8, 182, 128]) layer.2.v_cache: torch.Size([1, 8, 182, 128]) layer.3.k_cache: torch.Size([1, 8, 182, 128]) layer.3.v_cache: torch.Size([1, 8, 182, 128]) layer.4.k_cache: torch.Size([1, 8, 182, 128]) layer.4.v_cache: torch.Size([1, 8, 182, 128]) layer.4.output: torch.Size([1, 182, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02714607 27.73127093 layer.0.v_cache 0.00000026 0.00061815 layer.1.k_cache 0.00303102 3.99590327 layer.1.v_cache 0.00000080 0.00266935 layer.2.k_cache 0.00117008 1.56595620 layer.2.v_cache 0.00000108 0.00370689 layer.3.k_cache 0.00136151 1.89468887 layer.3.v_cache 0.00000204 0.00627380 layer.4.k_cache 0.00337920 3.72979602 layer.4.v_cache 0.00000313 0.01053334 layer.4.output 0.00018729 0.20616219 ------------------------------------------------------------------------------------- TOTAL 0.00263174 2.84043326 (elements=2,609,152) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2609152 Total Bytes 63228 BPFP 0.1939 bits/point EBPFP 0.3877 equivalent bits/point MSE 2.840433 ---------------------- -------------------------------------------------------- Time: 3.555s Load: 0.011s, Pack+Encode: 1.881s, Decode+Unpack: 1.663s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 182, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8404 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample21-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample21-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample22-layer4-item1.zst (27/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample22-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 191, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 191, 128) Output shape: (1, 191, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.0.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.1.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.1.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.2.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.2.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.3.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.3.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.4.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.4.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.4.output: torch.Size([1, 191, 4096]) -> torch.Size([1, 1, 191, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,088B, BPFP=0.0445 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,452B, BPFP=0.2230 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 3,812B, BPFP=0.1559 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,176B, BPFP=0.2526 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 4,852B, BPFP=0.1985 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,212B, BPFP=0.2950 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 5,248B, BPFP=0.2147 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,284B, BPFP=0.2979 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 3,780B, BPFP=0.1546 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,408B, BPFP=0.3030 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,488B, BPFP=0.0459 ⌛️ [2/4] FRONTEND: Frontend time: 1.871s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 191, 128]) layer.0.v_cache: torch.Size([1, 8, 191, 128]) layer.1.k_cache: torch.Size([1, 8, 191, 128]) layer.1.v_cache: torch.Size([1, 8, 191, 128]) layer.2.k_cache: torch.Size([1, 8, 191, 128]) layer.2.v_cache: torch.Size([1, 8, 191, 128]) layer.3.k_cache: torch.Size([1, 8, 191, 128]) layer.3.v_cache: torch.Size([1, 8, 191, 128]) layer.4.k_cache: torch.Size([1, 8, 191, 128]) layer.4.v_cache: torch.Size([1, 8, 191, 128]) layer.4.output: torch.Size([1, 191, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.489s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 191, 128]) layer.0.v_cache: torch.Size([1, 8, 191, 128]) layer.1.k_cache: torch.Size([1, 8, 191, 128]) layer.1.v_cache: torch.Size([1, 8, 191, 128]) layer.2.k_cache: torch.Size([1, 8, 191, 128]) layer.2.v_cache: torch.Size([1, 8, 191, 128]) layer.3.k_cache: torch.Size([1, 8, 191, 128]) layer.3.v_cache: torch.Size([1, 8, 191, 128]) layer.4.k_cache: torch.Size([1, 8, 191, 128]) layer.4.v_cache: torch.Size([1, 8, 191, 128]) layer.4.output: torch.Size([1, 191, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02679409 27.78588484 layer.0.v_cache 0.00000027 0.00062148 layer.1.k_cache 0.00297761 2.80556276 layer.1.v_cache 0.00000080 0.00254707 layer.2.k_cache 0.00115400 1.44502490 layer.2.v_cache 0.00000116 0.00364417 layer.3.k_cache 0.00135265 1.65746000 layer.3.v_cache 0.00000211 0.00599668 layer.4.k_cache 0.00347104 3.52225323 layer.4.v_cache 0.00000296 0.00996172 layer.4.output 0.00020113 0.20978704 ------------------------------------------------------------------------------------- TOTAL 0.00261152 2.71986464 (elements=2,738,176) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2738176 Total Bytes 56800 BPFP 0.1659 bits/point EBPFP 0.3319 equivalent bits/point MSE 2.719865 ---------------------- -------------------------------------------------------- Time: 3.371s Load: 0.011s, Pack+Encode: 1.871s, Decode+Unpack: 1.489s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 191, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7199 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample22-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample22-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample23-layer4-item1.zst (28/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample23-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 185, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 185, 128) Output shape: (1, 185, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 185, 128]) -> torch.Size([1, 1, 185, 1024]) layer.0.v_cache: torch.Size([1, 8, 185, 128]) -> torch.Size([1, 1, 185, 1024]) layer.1.k_cache: torch.Size([1, 8, 185, 128]) -> torch.Size([1, 1, 185, 1024]) layer.1.v_cache: torch.Size([1, 8, 185, 128]) -> torch.Size([1, 1, 185, 1024]) layer.2.k_cache: torch.Size([1, 8, 185, 128]) -> torch.Size([1, 1, 185, 1024]) layer.2.v_cache: torch.Size([1, 8, 185, 128]) -> torch.Size([1, 1, 185, 1024]) layer.3.k_cache: torch.Size([1, 8, 185, 128]) -> torch.Size([1, 1, 185, 1024]) layer.3.v_cache: torch.Size([1, 8, 185, 128]) -> torch.Size([1, 1, 185, 1024]) layer.4.k_cache: torch.Size([1, 8, 185, 128]) -> torch.Size([1, 1, 185, 1024]) layer.4.v_cache: torch.Size([1, 8, 185, 128]) -> torch.Size([1, 1, 185, 1024]) layer.4.output: torch.Size([1, 185, 4096]) -> torch.Size([1, 1, 185, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,208B, BPFP=0.0510 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,564B, BPFP=0.2350 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,080B, BPFP=0.2145 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,684B, BPFP=0.2823 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,344B, BPFP=0.2257 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,372B, BPFP=0.3113 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,348B, BPFP=0.2681 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,244B, BPFP=0.3059 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,708B, BPFP=0.1988 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,236B, BPFP=0.3056 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,736B, BPFP=0.0500 ⌛️ [2/4] FRONTEND: Frontend time: 1.965s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 185, 128]) layer.0.v_cache: torch.Size([1, 8, 185, 128]) layer.1.k_cache: torch.Size([1, 8, 185, 128]) layer.1.v_cache: torch.Size([1, 8, 185, 128]) layer.2.k_cache: torch.Size([1, 8, 185, 128]) layer.2.v_cache: torch.Size([1, 8, 185, 128]) layer.3.k_cache: torch.Size([1, 8, 185, 128]) layer.3.v_cache: torch.Size([1, 8, 185, 128]) layer.4.k_cache: torch.Size([1, 8, 185, 128]) layer.4.v_cache: torch.Size([1, 8, 185, 128]) layer.4.output: torch.Size([1, 185, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.414s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 185, 128]) layer.0.v_cache: torch.Size([1, 8, 185, 128]) layer.1.k_cache: torch.Size([1, 8, 185, 128]) layer.1.v_cache: torch.Size([1, 8, 185, 128]) layer.2.k_cache: torch.Size([1, 8, 185, 128]) layer.2.v_cache: torch.Size([1, 8, 185, 128]) layer.3.k_cache: torch.Size([1, 8, 185, 128]) layer.3.v_cache: torch.Size([1, 8, 185, 128]) layer.4.k_cache: torch.Size([1, 8, 185, 128]) layer.4.v_cache: torch.Size([1, 8, 185, 128]) layer.4.output: torch.Size([1, 185, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02696336 26.42374894 layer.0.v_cache 0.00000026 0.00061335 layer.1.k_cache 0.00303561 3.79151611 layer.1.v_cache 0.00000078 0.00269248 layer.2.k_cache 0.00115494 1.57561134 layer.2.v_cache 0.00000115 0.00394323 layer.3.k_cache 0.00132996 1.94830669 layer.3.v_cache 0.00000212 0.00643379 layer.4.k_cache 0.00355001 3.89952722 layer.4.v_cache 0.00000299 0.01053149 layer.4.output 0.00021000 0.19535079 ------------------------------------------------------------------------------------- TOTAL 0.00263437 2.74602342 (elements=2,652,160) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2652160 Total Bytes 61524 BPFP 0.1856 bits/point EBPFP 0.3712 equivalent bits/point MSE 2.746023 ---------------------- -------------------------------------------------------- Time: 3.390s Load: 0.011s, Pack+Encode: 1.965s, Decode+Unpack: 1.414s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 185, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 185, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7460 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample23-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample23-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample24-layer4-item1.zst (29/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample24-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 184, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 184, 128) Output shape: (1, 184, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.0.v_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.1.k_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.1.v_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.2.k_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.2.v_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.3.k_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.3.v_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.4.k_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.4.v_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.4.output: torch.Size([1, 184, 4096]) -> torch.Size([1, 1, 184, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,224B, BPFP=0.0520 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,856B, BPFP=0.2486 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,992B, BPFP=0.2120 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,640B, BPFP=0.2819 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,436B, BPFP=0.2308 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,316B, BPFP=0.3106 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,536B, BPFP=0.2775 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,692B, BPFP=0.3266 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,048B, BPFP=0.2143 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,016B, BPFP=0.2979 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,764B, BPFP=0.0506 ⌛️ [2/4] FRONTEND: Frontend time: 1.901s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 184, 128]) layer.0.v_cache: torch.Size([1, 8, 184, 128]) layer.1.k_cache: torch.Size([1, 8, 184, 128]) layer.1.v_cache: torch.Size([1, 8, 184, 128]) layer.2.k_cache: torch.Size([1, 8, 184, 128]) layer.2.v_cache: torch.Size([1, 8, 184, 128]) layer.3.k_cache: torch.Size([1, 8, 184, 128]) layer.3.v_cache: torch.Size([1, 8, 184, 128]) layer.4.k_cache: torch.Size([1, 8, 184, 128]) layer.4.v_cache: torch.Size([1, 8, 184, 128]) layer.4.output: torch.Size([1, 184, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.587s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 184, 128]) layer.0.v_cache: torch.Size([1, 8, 184, 128]) layer.1.k_cache: torch.Size([1, 8, 184, 128]) layer.1.v_cache: torch.Size([1, 8, 184, 128]) layer.2.k_cache: torch.Size([1, 8, 184, 128]) layer.2.v_cache: torch.Size([1, 8, 184, 128]) layer.3.k_cache: torch.Size([1, 8, 184, 128]) layer.3.v_cache: torch.Size([1, 8, 184, 128]) layer.4.k_cache: torch.Size([1, 8, 184, 128]) layer.4.v_cache: torch.Size([1, 8, 184, 128]) layer.4.output: torch.Size([1, 184, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02665583 26.06273883 layer.0.v_cache 0.00000026 0.00061239 layer.1.k_cache 0.00299416 3.92137876 layer.1.v_cache 0.00000077 0.00254864 layer.2.k_cache 0.00115256 1.58554359 layer.2.v_cache 0.00000113 0.00366196 layer.3.k_cache 0.00134142 1.94404635 layer.3.v_cache 0.00000202 0.00608853 layer.4.k_cache 0.00346119 3.87254566 layer.4.v_cache 0.00000292 0.01010711 layer.4.output 0.00021714 0.20392660 ------------------------------------------------------------------------------------- TOTAL 0.00260577 2.73035559 (elements=2,637,824) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2637824 Total Bytes 62520 BPFP 0.1896 bits/point EBPFP 0.3792 equivalent bits/point MSE 2.730356 ---------------------- -------------------------------------------------------- Time: 3.499s Load: 0.011s, Pack+Encode: 1.901s, Decode+Unpack: 1.587s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 184, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7304 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample24-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample24-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample25-layer4-item1.zst (30/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample25-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 195, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 195, 128) Output shape: (1, 195, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 195, 128]) -> torch.Size([1, 1, 195, 1024]) layer.0.v_cache: torch.Size([1, 8, 195, 128]) -> torch.Size([1, 1, 195, 1024]) layer.1.k_cache: torch.Size([1, 8, 195, 128]) -> torch.Size([1, 1, 195, 1024]) layer.1.v_cache: torch.Size([1, 8, 195, 128]) -> torch.Size([1, 1, 195, 1024]) layer.2.k_cache: torch.Size([1, 8, 195, 128]) -> torch.Size([1, 1, 195, 1024]) layer.2.v_cache: torch.Size([1, 8, 195, 128]) -> torch.Size([1, 1, 195, 1024]) layer.3.k_cache: torch.Size([1, 8, 195, 128]) -> torch.Size([1, 1, 195, 1024]) layer.3.v_cache: torch.Size([1, 8, 195, 128]) -> torch.Size([1, 1, 195, 1024]) layer.4.k_cache: torch.Size([1, 8, 195, 128]) -> torch.Size([1, 1, 195, 1024]) layer.4.v_cache: torch.Size([1, 8, 195, 128]) -> torch.Size([1, 1, 195, 1024]) layer.4.output: torch.Size([1, 195, 4096]) -> torch.Size([1, 1, 195, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,204B, BPFP=0.0482 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,972B, BPFP=0.2793 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,348B, BPFP=0.2543 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 8,348B, BPFP=0.3345 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,692B, BPFP=0.3082 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 9,244B, BPFP=0.3704 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,280B, BPFP=0.3317 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 9,192B, BPFP=0.3683 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,248B, BPFP=0.2503 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 9,880B, BPFP=0.3958 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,428B, BPFP=0.0544 ⌛️ [2/4] FRONTEND: Frontend time: 2.277s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 195, 128]) layer.0.v_cache: torch.Size([1, 8, 195, 128]) layer.1.k_cache: torch.Size([1, 8, 195, 128]) layer.1.v_cache: torch.Size([1, 8, 195, 128]) layer.2.k_cache: torch.Size([1, 8, 195, 128]) layer.2.v_cache: torch.Size([1, 8, 195, 128]) layer.3.k_cache: torch.Size([1, 8, 195, 128]) layer.3.v_cache: torch.Size([1, 8, 195, 128]) layer.4.k_cache: torch.Size([1, 8, 195, 128]) layer.4.v_cache: torch.Size([1, 8, 195, 128]) layer.4.output: torch.Size([1, 195, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.040s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 195, 128]) layer.0.v_cache: torch.Size([1, 8, 195, 128]) layer.1.k_cache: torch.Size([1, 8, 195, 128]) layer.1.v_cache: torch.Size([1, 8, 195, 128]) layer.2.k_cache: torch.Size([1, 8, 195, 128]) layer.2.v_cache: torch.Size([1, 8, 195, 128]) layer.3.k_cache: torch.Size([1, 8, 195, 128]) layer.3.v_cache: torch.Size([1, 8, 195, 128]) layer.4.k_cache: torch.Size([1, 8, 195, 128]) layer.4.v_cache: torch.Size([1, 8, 195, 128]) layer.4.output: torch.Size([1, 195, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02622716 30.81013121 layer.0.v_cache 0.00000027 0.00062284 layer.1.k_cache 0.00300573 3.09476882 layer.1.v_cache 0.00000076 0.00248501 layer.2.k_cache 0.00116358 1.48983780 layer.2.v_cache 0.00000112 0.00368341 layer.3.k_cache 0.00131201 1.72766285 layer.3.v_cache 0.00000217 0.00624231 layer.4.k_cache 0.00348515 3.17754814 layer.4.v_cache 0.00000317 0.01049349 layer.4.output 0.00014777 0.20143282 ------------------------------------------------------------------------------------- TOTAL 0.00255658 2.93780051 (elements=2,795,520) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2795520 Total Bytes 78836 BPFP 0.2256 bits/point EBPFP 0.4512 equivalent bits/point MSE 2.937801 ---------------------- -------------------------------------------------------- Time: 4.327s Load: 0.010s, Pack+Encode: 2.277s, Decode+Unpack: 2.040s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 195, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 195, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.9378 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample25-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample25-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample26-layer4-item1.zst (31/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample26-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 190, 128) Output shape: (1, 190, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.0.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.1.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.1.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.2.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.2.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.3.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.3.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.4.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.4.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.4.output: torch.Size([1, 190, 4096]) -> torch.Size([1, 1, 190, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,212B, BPFP=0.0498 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,560B, BPFP=0.2286 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,140B, BPFP=0.1702 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,028B, BPFP=0.2479 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,060B, BPFP=0.2081 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,860B, BPFP=0.2821 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 5,624B, BPFP=0.2313 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,164B, BPFP=0.2946 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,036B, BPFP=0.1660 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,976B, BPFP=0.2868 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,692B, BPFP=0.0585 ⌛️ [2/4] FRONTEND: Frontend time: 1.899s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 190, 128]) layer.0.v_cache: torch.Size([1, 8, 190, 128]) layer.1.k_cache: torch.Size([1, 8, 190, 128]) layer.1.v_cache: torch.Size([1, 8, 190, 128]) layer.2.k_cache: torch.Size([1, 8, 190, 128]) layer.2.v_cache: torch.Size([1, 8, 190, 128]) layer.3.k_cache: torch.Size([1, 8, 190, 128]) layer.3.v_cache: torch.Size([1, 8, 190, 128]) layer.4.k_cache: torch.Size([1, 8, 190, 128]) layer.4.v_cache: torch.Size([1, 8, 190, 128]) layer.4.output: torch.Size([1, 190, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.439s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 190, 128]) layer.0.v_cache: torch.Size([1, 8, 190, 128]) layer.1.k_cache: torch.Size([1, 8, 190, 128]) layer.1.v_cache: torch.Size([1, 8, 190, 128]) layer.2.k_cache: torch.Size([1, 8, 190, 128]) layer.2.v_cache: torch.Size([1, 8, 190, 128]) layer.3.k_cache: torch.Size([1, 8, 190, 128]) layer.3.v_cache: torch.Size([1, 8, 190, 128]) layer.4.k_cache: torch.Size([1, 8, 190, 128]) layer.4.v_cache: torch.Size([1, 8, 190, 128]) layer.4.output: torch.Size([1, 190, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02720581 27.56038497 layer.0.v_cache 0.00000028 0.00062744 layer.1.k_cache 0.00301432 3.06528770 layer.1.v_cache 0.00000084 0.00259049 layer.2.k_cache 0.00115620 1.54433803 layer.2.v_cache 0.00000110 0.00383011 layer.3.k_cache 0.00139960 1.79050213 layer.3.v_cache 0.00000211 0.00638866 layer.4.k_cache 0.00346420 3.64164429 layer.4.v_cache 0.00000301 0.01043989 layer.4.output 0.00019765 0.21753811 ------------------------------------------------------------------------------------- TOTAL 0.00264558 2.74972758 (elements=2,723,840) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2723840 Total Bytes 58352 BPFP 0.1714 bits/point EBPFP 0.3428 equivalent bits/point MSE 2.749728 ---------------------- -------------------------------------------------------- Time: 3.349s Load: 0.010s, Pack+Encode: 1.899s, Decode+Unpack: 1.439s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7497 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample26-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample26-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample27-layer4-item1.zst (32/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample27-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 177, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 177, 128) Output shape: (1, 177, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.0.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.1.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.1.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.2.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.2.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.3.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.3.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.4.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.4.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.4.output: torch.Size([1, 177, 4096]) -> torch.Size([1, 1, 177, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,064B, BPFP=0.0470 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,192B, BPFP=0.2733 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,724B, BPFP=0.2526 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,860B, BPFP=0.3028 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,628B, BPFP=0.2484 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,476B, BPFP=0.3300 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,628B, BPFP=0.2925 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,312B, BPFP=0.3227 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,240B, BPFP=0.2313 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,540B, BPFP=0.3328 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,192B, BPFP=0.0463 ⌛️ [2/4] FRONTEND: Frontend time: 1.989s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 177, 128]) layer.0.v_cache: torch.Size([1, 8, 177, 128]) layer.1.k_cache: torch.Size([1, 8, 177, 128]) layer.1.v_cache: torch.Size([1, 8, 177, 128]) layer.2.k_cache: torch.Size([1, 8, 177, 128]) layer.2.v_cache: torch.Size([1, 8, 177, 128]) layer.3.k_cache: torch.Size([1, 8, 177, 128]) layer.3.v_cache: torch.Size([1, 8, 177, 128]) layer.4.k_cache: torch.Size([1, 8, 177, 128]) layer.4.v_cache: torch.Size([1, 8, 177, 128]) layer.4.output: torch.Size([1, 177, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.436s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 177, 128]) layer.0.v_cache: torch.Size([1, 8, 177, 128]) layer.1.k_cache: torch.Size([1, 8, 177, 128]) layer.1.v_cache: torch.Size([1, 8, 177, 128]) layer.2.k_cache: torch.Size([1, 8, 177, 128]) layer.2.v_cache: torch.Size([1, 8, 177, 128]) layer.3.k_cache: torch.Size([1, 8, 177, 128]) layer.3.v_cache: torch.Size([1, 8, 177, 128]) layer.4.k_cache: torch.Size([1, 8, 177, 128]) layer.4.v_cache: torch.Size([1, 8, 177, 128]) layer.4.output: torch.Size([1, 177, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02683750 24.82589325 layer.0.v_cache 0.00000026 0.00061362 layer.1.k_cache 0.00307604 3.71616894 layer.1.v_cache 0.00000078 0.00257801 layer.2.k_cache 0.00123050 1.52140334 layer.2.v_cache 0.00000110 0.00372138 layer.3.k_cache 0.00133191 1.77701961 layer.3.v_cache 0.00000206 0.00607548 layer.4.k_cache 0.00350474 3.64680990 layer.4.v_cache 0.00000300 0.01005244 layer.4.output 0.00016181 0.19686025 ------------------------------------------------------------------------------------- TOTAL 0.00261679 2.59269836 (elements=2,537,472) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2537472 Total Bytes 63856 BPFP 0.2013 bits/point EBPFP 0.4026 equivalent bits/point MSE 2.592698 ---------------------- -------------------------------------------------------- Time: 3.435s Load: 0.011s, Pack+Encode: 1.989s, Decode+Unpack: 1.436s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 177, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5927 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample27-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample27-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample28-layer4-item1.zst (33/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample28-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 182, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 182, 128) Output shape: (1, 182, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.0.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.1.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.1.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.2.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.2.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.3.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.3.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.4.k_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.4.v_cache: torch.Size([1, 8, 182, 128]) -> torch.Size([1, 1, 182, 1024]) layer.4.output: torch.Size([1, 182, 4096]) -> torch.Size([1, 1, 182, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,140B, BPFP=0.0489 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,336B, BPFP=0.2291 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,176B, BPFP=0.2222 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,828B, BPFP=0.2931 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,560B, BPFP=0.2387 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,500B, BPFP=0.3219 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,616B, BPFP=0.2840 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,604B, BPFP=0.3264 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,712B, BPFP=0.2023 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,452B, BPFP=0.3199 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,820B, BPFP=0.0517 ⌛️ [2/4] FRONTEND: Frontend time: 1.905s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 182, 128]) layer.0.v_cache: torch.Size([1, 8, 182, 128]) layer.1.k_cache: torch.Size([1, 8, 182, 128]) layer.1.v_cache: torch.Size([1, 8, 182, 128]) layer.2.k_cache: torch.Size([1, 8, 182, 128]) layer.2.v_cache: torch.Size([1, 8, 182, 128]) layer.3.k_cache: torch.Size([1, 8, 182, 128]) layer.3.v_cache: torch.Size([1, 8, 182, 128]) layer.4.k_cache: torch.Size([1, 8, 182, 128]) layer.4.v_cache: torch.Size([1, 8, 182, 128]) layer.4.output: torch.Size([1, 182, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.608s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 182, 128]) layer.0.v_cache: torch.Size([1, 8, 182, 128]) layer.1.k_cache: torch.Size([1, 8, 182, 128]) layer.1.v_cache: torch.Size([1, 8, 182, 128]) layer.2.k_cache: torch.Size([1, 8, 182, 128]) layer.2.v_cache: torch.Size([1, 8, 182, 128]) layer.3.k_cache: torch.Size([1, 8, 182, 128]) layer.3.v_cache: torch.Size([1, 8, 182, 128]) layer.4.k_cache: torch.Size([1, 8, 182, 128]) layer.4.v_cache: torch.Size([1, 8, 182, 128]) layer.4.output: torch.Size([1, 182, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02800116 27.30576601 layer.0.v_cache 0.00000026 0.00061915 layer.1.k_cache 0.00305201 3.70562040 layer.1.v_cache 0.00000078 0.00254717 layer.2.k_cache 0.00115625 1.53119165 layer.2.v_cache 0.00000114 0.00389355 layer.3.k_cache 0.00134464 1.87564774 layer.3.v_cache 0.00000201 0.00620850 layer.4.k_cache 0.00345014 3.85909498 layer.4.v_cache 0.00000301 0.01058916 layer.4.output 0.00022774 0.20815409 ------------------------------------------------------------------------------------- TOTAL 0.00270874 2.79527105 (elements=2,609,152) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2609152 Total Bytes 62744 BPFP 0.1924 bits/point EBPFP 0.3848 equivalent bits/point MSE 2.795271 ---------------------- -------------------------------------------------------- Time: 3.523s Load: 0.009s, Pack+Encode: 1.905s, Decode+Unpack: 1.608s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 182, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 182, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7953 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample28-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample28-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample29-layer4-item1.zst (34/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample29-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 186, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 186, 128) Output shape: (1, 186, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 186, 128]) -> torch.Size([1, 1, 186, 1024]) layer.0.v_cache: torch.Size([1, 8, 186, 128]) -> torch.Size([1, 1, 186, 1024]) layer.1.k_cache: torch.Size([1, 8, 186, 128]) -> torch.Size([1, 1, 186, 1024]) layer.1.v_cache: torch.Size([1, 8, 186, 128]) -> torch.Size([1, 1, 186, 1024]) layer.2.k_cache: torch.Size([1, 8, 186, 128]) -> torch.Size([1, 1, 186, 1024]) layer.2.v_cache: torch.Size([1, 8, 186, 128]) -> torch.Size([1, 1, 186, 1024]) layer.3.k_cache: torch.Size([1, 8, 186, 128]) -> torch.Size([1, 1, 186, 1024]) layer.3.v_cache: torch.Size([1, 8, 186, 128]) -> torch.Size([1, 1, 186, 1024]) layer.4.k_cache: torch.Size([1, 8, 186, 128]) -> torch.Size([1, 1, 186, 1024]) layer.4.v_cache: torch.Size([1, 8, 186, 128]) -> torch.Size([1, 1, 186, 1024]) layer.4.output: torch.Size([1, 186, 4096]) -> torch.Size([1, 1, 186, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,148B, BPFP=0.0482 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,716B, BPFP=0.2401 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,632B, BPFP=0.1946 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,620B, BPFP=0.2781 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,708B, BPFP=0.2398 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,028B, BPFP=0.2952 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,180B, BPFP=0.2596 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,780B, BPFP=0.2848 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,752B, BPFP=0.1996 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,416B, BPFP=0.3115 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,772B, BPFP=0.0501 ⌛️ [2/4] FRONTEND: Frontend time: 1.870s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 186, 128]) layer.0.v_cache: torch.Size([1, 8, 186, 128]) layer.1.k_cache: torch.Size([1, 8, 186, 128]) layer.1.v_cache: torch.Size([1, 8, 186, 128]) layer.2.k_cache: torch.Size([1, 8, 186, 128]) layer.2.v_cache: torch.Size([1, 8, 186, 128]) layer.3.k_cache: torch.Size([1, 8, 186, 128]) layer.3.v_cache: torch.Size([1, 8, 186, 128]) layer.4.k_cache: torch.Size([1, 8, 186, 128]) layer.4.v_cache: torch.Size([1, 8, 186, 128]) layer.4.output: torch.Size([1, 186, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.493s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 186, 128]) layer.0.v_cache: torch.Size([1, 8, 186, 128]) layer.1.k_cache: torch.Size([1, 8, 186, 128]) layer.1.v_cache: torch.Size([1, 8, 186, 128]) layer.2.k_cache: torch.Size([1, 8, 186, 128]) layer.2.v_cache: torch.Size([1, 8, 186, 128]) layer.3.k_cache: torch.Size([1, 8, 186, 128]) layer.3.v_cache: torch.Size([1, 8, 186, 128]) layer.4.k_cache: torch.Size([1, 8, 186, 128]) layer.4.v_cache: torch.Size([1, 8, 186, 128]) layer.4.output: torch.Size([1, 186, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02748687 26.69007004 layer.0.v_cache 0.00000027 0.00061881 layer.1.k_cache 0.00295559 3.82718782 layer.1.v_cache 0.00000084 0.00267811 layer.2.k_cache 0.00114893 1.60117496 layer.2.v_cache 0.00000110 0.00393968 layer.3.k_cache 0.00136328 1.92943269 layer.3.v_cache 0.00000212 0.00650740 layer.4.k_cache 0.00348854 3.79296087 layer.4.v_cache 0.00000314 0.01088055 layer.4.output 0.00017293 0.20712627 ------------------------------------------------------------------------------------- TOTAL 0.00265303 2.76385400 (elements=2,666,496) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2666496 Total Bytes 60752 BPFP 0.1823 bits/point EBPFP 0.3645 equivalent bits/point MSE 2.763854 ---------------------- -------------------------------------------------------- Time: 3.373s Load: 0.010s, Pack+Encode: 1.870s, Decode+Unpack: 1.493s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 186, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 186, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7639 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample29-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample29-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample3-layer4-item1.zst (35/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample3-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 225, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 225, 128) Output shape: (1, 225, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 225, 128]) -> torch.Size([1, 1, 225, 1024]) layer.0.v_cache: torch.Size([1, 8, 225, 128]) -> torch.Size([1, 1, 225, 1024]) layer.1.k_cache: torch.Size([1, 8, 225, 128]) -> torch.Size([1, 1, 225, 1024]) layer.1.v_cache: torch.Size([1, 8, 225, 128]) -> torch.Size([1, 1, 225, 1024]) layer.2.k_cache: torch.Size([1, 8, 225, 128]) -> torch.Size([1, 1, 225, 1024]) layer.2.v_cache: torch.Size([1, 8, 225, 128]) -> torch.Size([1, 1, 225, 1024]) layer.3.k_cache: torch.Size([1, 8, 225, 128]) -> torch.Size([1, 1, 225, 1024]) layer.3.v_cache: torch.Size([1, 8, 225, 128]) -> torch.Size([1, 1, 225, 1024]) layer.4.k_cache: torch.Size([1, 8, 225, 128]) -> torch.Size([1, 1, 225, 1024]) layer.4.v_cache: torch.Size([1, 8, 225, 128]) -> torch.Size([1, 1, 225, 1024]) layer.4.output: torch.Size([1, 225, 4096]) -> torch.Size([1, 1, 225, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,236B, BPFP=0.0429 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,156B, BPFP=0.2485 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,032B, BPFP=0.2094 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 7,524B, BPFP=0.2612 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,524B, BPFP=0.2612 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 8,336B, BPFP=0.2894 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,320B, BPFP=0.2889 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 9,008B, BPFP=0.3128 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,736B, BPFP=0.1992 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 9,308B, BPFP=0.3232 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,700B, BPFP=0.0408 ⌛️ [2/4] FRONTEND: Frontend time: 2.402s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 225, 128]) layer.0.v_cache: torch.Size([1, 8, 225, 128]) layer.1.k_cache: torch.Size([1, 8, 225, 128]) layer.1.v_cache: torch.Size([1, 8, 225, 128]) layer.2.k_cache: torch.Size([1, 8, 225, 128]) layer.2.v_cache: torch.Size([1, 8, 225, 128]) layer.3.k_cache: torch.Size([1, 8, 225, 128]) layer.3.v_cache: torch.Size([1, 8, 225, 128]) layer.4.k_cache: torch.Size([1, 8, 225, 128]) layer.4.v_cache: torch.Size([1, 8, 225, 128]) layer.4.output: torch.Size([1, 225, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.822s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 225, 128]) layer.0.v_cache: torch.Size([1, 8, 225, 128]) layer.1.k_cache: torch.Size([1, 8, 225, 128]) layer.1.v_cache: torch.Size([1, 8, 225, 128]) layer.2.k_cache: torch.Size([1, 8, 225, 128]) layer.2.v_cache: torch.Size([1, 8, 225, 128]) layer.3.k_cache: torch.Size([1, 8, 225, 128]) layer.3.v_cache: torch.Size([1, 8, 225, 128]) layer.4.k_cache: torch.Size([1, 8, 225, 128]) layer.4.v_cache: torch.Size([1, 8, 225, 128]) layer.4.output: torch.Size([1, 225, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02545540 28.83332899 layer.0.v_cache 0.00000025 0.00060997 layer.1.k_cache 0.00297359 3.10775499 layer.1.v_cache 0.00000080 0.00253042 layer.2.k_cache 0.00113723 1.42600667 layer.2.v_cache 0.00000119 0.00388022 layer.3.k_cache 0.00129491 1.70475369 layer.3.v_cache 0.00000221 0.00649621 layer.4.k_cache 0.00357589 3.30681261 layer.4.v_cache 0.00000317 0.01091760 layer.4.output 0.00016372 0.17433567 ------------------------------------------------------------------------------------- TOTAL 0.00250711 2.79288815 (elements=3,225,600) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 3225600 Total Bytes 74880 BPFP 0.1857 bits/point EBPFP 0.3714 equivalent bits/point MSE 2.792888 ---------------------- -------------------------------------------------------- Time: 4.236s Load: 0.012s, Pack+Encode: 2.402s, Decode+Unpack: 1.822s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 225, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 225, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7929 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample3-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample3-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample30-layer4-item1.zst (36/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample30-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 179, 128) Output shape: (1, 179, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.output: torch.Size([1, 179, 4096]) -> torch.Size([1, 1, 179, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,096B, BPFP=0.0478 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,320B, BPFP=0.2758 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,612B, BPFP=0.2449 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,460B, BPFP=0.2819 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,784B, BPFP=0.2524 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,728B, BPFP=0.3373 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,936B, BPFP=0.3027 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,416B, BPFP=0.3673 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,360B, BPFP=0.2339 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 8,392B, BPFP=0.3663 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,264B, BPFP=0.0465 ⌛️ [2/4] FRONTEND: Frontend time: 1.935s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) layer.4.output: torch.Size([1, 179, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.530s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) layer.4.output: torch.Size([1, 179, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02765450 25.42915285 layer.0.v_cache 0.00000027 0.00062592 layer.1.k_cache 0.00304088 3.62952649 layer.1.v_cache 0.00000084 0.00271854 layer.2.k_cache 0.00119692 1.50304340 layer.2.v_cache 0.00000108 0.00384479 layer.3.k_cache 0.00131095 1.82238838 layer.3.v_cache 0.00000209 0.00659882 layer.4.k_cache 0.00344948 3.52987995 layer.4.v_cache 0.00000315 0.01091118 layer.4.output 0.00017252 0.19013668 ------------------------------------------------------------------------------------- TOTAL 0.00266787 2.62137407 (elements=2,566,144) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2566144 Total Bytes 66368 BPFP 0.2069 bits/point EBPFP 0.4138 equivalent bits/point MSE 2.621374 ---------------------- -------------------------------------------------------- Time: 3.474s Load: 0.009s, Pack+Encode: 1.935s, Decode+Unpack: 1.530s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6214 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample30-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample30-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample31-layer4-item1.zst (37/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample31-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 187, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 187, 128) Output shape: (1, 187, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 187, 128]) -> torch.Size([1, 1, 187, 1024]) layer.0.v_cache: torch.Size([1, 8, 187, 128]) -> torch.Size([1, 1, 187, 1024]) layer.1.k_cache: torch.Size([1, 8, 187, 128]) -> torch.Size([1, 1, 187, 1024]) layer.1.v_cache: torch.Size([1, 8, 187, 128]) -> torch.Size([1, 1, 187, 1024]) layer.2.k_cache: torch.Size([1, 8, 187, 128]) -> torch.Size([1, 1, 187, 1024]) layer.2.v_cache: torch.Size([1, 8, 187, 128]) -> torch.Size([1, 1, 187, 1024]) layer.3.k_cache: torch.Size([1, 8, 187, 128]) -> torch.Size([1, 1, 187, 1024]) layer.3.v_cache: torch.Size([1, 8, 187, 128]) -> torch.Size([1, 1, 187, 1024]) layer.4.k_cache: torch.Size([1, 8, 187, 128]) -> torch.Size([1, 1, 187, 1024]) layer.4.v_cache: torch.Size([1, 8, 187, 128]) -> torch.Size([1, 1, 187, 1024]) layer.4.output: torch.Size([1, 187, 4096]) -> torch.Size([1, 1, 187, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,132B, BPFP=0.0473 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,560B, BPFP=0.2323 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,568B, BPFP=0.1908 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,460B, BPFP=0.2699 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,252B, BPFP=0.2194 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,096B, BPFP=0.2965 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,096B, BPFP=0.2547 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,236B, BPFP=0.3023 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,668B, BPFP=0.1950 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,556B, BPFP=0.3157 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,132B, BPFP=0.0432 ⌛️ [2/4] FRONTEND: Frontend time: 1.884s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 187, 128]) layer.0.v_cache: torch.Size([1, 8, 187, 128]) layer.1.k_cache: torch.Size([1, 8, 187, 128]) layer.1.v_cache: torch.Size([1, 8, 187, 128]) layer.2.k_cache: torch.Size([1, 8, 187, 128]) layer.2.v_cache: torch.Size([1, 8, 187, 128]) layer.3.k_cache: torch.Size([1, 8, 187, 128]) layer.3.v_cache: torch.Size([1, 8, 187, 128]) layer.4.k_cache: torch.Size([1, 8, 187, 128]) layer.4.v_cache: torch.Size([1, 8, 187, 128]) layer.4.output: torch.Size([1, 187, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.610s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 187, 128]) layer.0.v_cache: torch.Size([1, 8, 187, 128]) layer.1.k_cache: torch.Size([1, 8, 187, 128]) layer.1.v_cache: torch.Size([1, 8, 187, 128]) layer.2.k_cache: torch.Size([1, 8, 187, 128]) layer.2.v_cache: torch.Size([1, 8, 187, 128]) layer.3.k_cache: torch.Size([1, 8, 187, 128]) layer.3.v_cache: torch.Size([1, 8, 187, 128]) layer.4.k_cache: torch.Size([1, 8, 187, 128]) layer.4.v_cache: torch.Size([1, 8, 187, 128]) layer.4.output: torch.Size([1, 187, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02664549 24.67314662 layer.0.v_cache 0.00000027 0.00062346 layer.1.k_cache 0.00298854 3.55975440 layer.1.v_cache 0.00000078 0.00270032 layer.2.k_cache 0.00115871 1.55890513 layer.2.v_cache 0.00000111 0.00388716 layer.3.k_cache 0.00133556 1.89289440 layer.3.v_cache 0.00000211 0.00623649 layer.4.k_cache 0.00339894 3.79620035 layer.4.v_cache 0.00000310 0.01067294 layer.4.output 0.00017976 0.19036549 ------------------------------------------------------------------------------------- TOTAL 0.00258955 2.59046309 (elements=2,680,832) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2680832 Total Bytes 59756 BPFP 0.1783 bits/point EBPFP 0.3566 equivalent bits/point MSE 2.590463 ---------------------- -------------------------------------------------------- Time: 3.506s Load: 0.011s, Pack+Encode: 1.884s, Decode+Unpack: 1.610s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 187, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 187, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5905 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample31-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample31-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample33-layer4-item1.zst (38/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample33-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 190, 128) Output shape: (1, 190, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.0.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.1.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.1.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.2.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.2.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.3.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.3.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.4.k_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.4.v_cache: torch.Size([1, 8, 190, 128]) -> torch.Size([1, 1, 190, 1024]) layer.4.output: torch.Size([1, 190, 4096]) -> torch.Size([1, 1, 190, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,212B, BPFP=0.0498 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,692B, BPFP=0.2340 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,184B, BPFP=0.1720 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,348B, BPFP=0.2610 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,020B, BPFP=0.2064 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,848B, BPFP=0.2816 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 5,588B, BPFP=0.2298 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,628B, BPFP=0.2725 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,092B, BPFP=0.1683 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,696B, BPFP=0.2753 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,380B, BPFP=0.0450 ⌛️ [2/4] FRONTEND: Frontend time: 1.908s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 190, 128]) layer.0.v_cache: torch.Size([1, 8, 190, 128]) layer.1.k_cache: torch.Size([1, 8, 190, 128]) layer.1.v_cache: torch.Size([1, 8, 190, 128]) layer.2.k_cache: torch.Size([1, 8, 190, 128]) layer.2.v_cache: torch.Size([1, 8, 190, 128]) layer.3.k_cache: torch.Size([1, 8, 190, 128]) layer.3.v_cache: torch.Size([1, 8, 190, 128]) layer.4.k_cache: torch.Size([1, 8, 190, 128]) layer.4.v_cache: torch.Size([1, 8, 190, 128]) layer.4.output: torch.Size([1, 190, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.449s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 190, 128]) layer.0.v_cache: torch.Size([1, 8, 190, 128]) layer.1.k_cache: torch.Size([1, 8, 190, 128]) layer.1.v_cache: torch.Size([1, 8, 190, 128]) layer.2.k_cache: torch.Size([1, 8, 190, 128]) layer.2.v_cache: torch.Size([1, 8, 190, 128]) layer.3.k_cache: torch.Size([1, 8, 190, 128]) layer.3.v_cache: torch.Size([1, 8, 190, 128]) layer.4.k_cache: torch.Size([1, 8, 190, 128]) layer.4.v_cache: torch.Size([1, 8, 190, 128]) layer.4.output: torch.Size([1, 190, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02674542 28.56797903 layer.0.v_cache 0.00000026 0.00063121 layer.1.k_cache 0.00305433 2.94579147 layer.1.v_cache 0.00000079 0.00256716 layer.2.k_cache 0.00120723 1.53200330 layer.2.v_cache 0.00000116 0.00393388 layer.3.k_cache 0.00130860 1.72376323 layer.3.v_cache 0.00000216 0.00628254 layer.4.k_cache 0.00354577 3.25430073 layer.4.v_cache 0.00000304 0.01058553 layer.4.output 0.00015926 0.18517342 ------------------------------------------------------------------------------------- TOTAL 0.00260756 2.77060941 (elements=2,723,840) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2723840 Total Bytes 56688 BPFP 0.1665 bits/point EBPFP 0.3330 equivalent bits/point MSE 2.770609 ---------------------- -------------------------------------------------------- Time: 3.368s Load: 0.011s, Pack+Encode: 1.908s, Decode+Unpack: 1.449s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 190, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 190, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7706 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample33-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample33-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample34-layer4-item1.zst (39/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample34-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 171, 128) Output shape: (1, 171, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.output: torch.Size([1, 171, 4096]) -> torch.Size([1, 1, 171, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,028B, BPFP=0.0470 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,100B, BPFP=0.2787 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,624B, BPFP=0.2569 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,692B, BPFP=0.3057 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,672B, BPFP=0.2591 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,536B, BPFP=0.3443 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,868B, BPFP=0.3138 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,580B, BPFP=0.3463 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,432B, BPFP=0.2482 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,264B, BPFP=0.3319 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,356B, BPFP=0.0498 ⌛️ [2/4] FRONTEND: Frontend time: 1.950s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.458s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02696512 26.14182429 layer.0.v_cache 0.00000027 0.00062458 layer.1.k_cache 0.00309770 3.52403857 layer.1.v_cache 0.00000080 0.00266743 layer.2.k_cache 0.00123733 1.46297798 layer.2.v_cache 0.00000111 0.00380504 layer.3.k_cache 0.00133744 1.74613194 layer.3.v_cache 0.00000211 0.00618570 layer.4.k_cache 0.00342193 3.33874262 layer.4.v_cache 0.00000311 0.01074589 layer.4.output 0.00017162 0.19610854 ------------------------------------------------------------------------------------- TOTAL 0.00262524 2.64444130 (elements=2,451,456) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2451456 Total Bytes 64152 BPFP 0.2094 bits/point EBPFP 0.4187 equivalent bits/point MSE 2.644441 ---------------------- -------------------------------------------------------- Time: 3.418s Load: 0.010s, Pack+Encode: 1.950s, Decode+Unpack: 1.458s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6444 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample34-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample34-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample35-layer4-item1.zst (40/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample35-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 191, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 191, 128) Output shape: (1, 191, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.0.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.1.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.1.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.2.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.2.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.3.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.3.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.4.k_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.4.v_cache: torch.Size([1, 8, 191, 128]) -> torch.Size([1, 1, 191, 1024]) layer.4.output: torch.Size([1, 191, 4096]) -> torch.Size([1, 1, 191, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,064B, BPFP=0.0435 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,032B, BPFP=0.2058 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 3,844B, BPFP=0.1572 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,172B, BPFP=0.2525 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 4,344B, BPFP=0.1777 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,016B, BPFP=0.2870 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 5,232B, BPFP=0.2140 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,852B, BPFP=0.2803 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 3,744B, BPFP=0.1531 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,816B, BPFP=0.2788 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,204B, BPFP=0.0328 ⌛️ [2/4] FRONTEND: Frontend time: 1.895s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 191, 128]) layer.0.v_cache: torch.Size([1, 8, 191, 128]) layer.1.k_cache: torch.Size([1, 8, 191, 128]) layer.1.v_cache: torch.Size([1, 8, 191, 128]) layer.2.k_cache: torch.Size([1, 8, 191, 128]) layer.2.v_cache: torch.Size([1, 8, 191, 128]) layer.3.k_cache: torch.Size([1, 8, 191, 128]) layer.3.v_cache: torch.Size([1, 8, 191, 128]) layer.4.k_cache: torch.Size([1, 8, 191, 128]) layer.4.v_cache: torch.Size([1, 8, 191, 128]) layer.4.output: torch.Size([1, 191, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.639s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 191, 128]) layer.0.v_cache: torch.Size([1, 8, 191, 128]) layer.1.k_cache: torch.Size([1, 8, 191, 128]) layer.1.v_cache: torch.Size([1, 8, 191, 128]) layer.2.k_cache: torch.Size([1, 8, 191, 128]) layer.2.v_cache: torch.Size([1, 8, 191, 128]) layer.3.k_cache: torch.Size([1, 8, 191, 128]) layer.3.v_cache: torch.Size([1, 8, 191, 128]) layer.4.k_cache: torch.Size([1, 8, 191, 128]) layer.4.v_cache: torch.Size([1, 8, 191, 128]) layer.4.output: torch.Size([1, 191, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02691182 26.67027722 layer.0.v_cache 0.00000026 0.00057561 layer.1.k_cache 0.00302612 3.05276729 layer.1.v_cache 0.00000075 0.00232721 layer.2.k_cache 0.00114460 1.42322928 layer.2.v_cache 0.00000105 0.00335267 layer.3.k_cache 0.00140477 1.60614956 layer.3.v_cache 0.00000198 0.00551033 layer.4.k_cache 0.00340838 3.39042048 layer.4.v_cache 0.00000288 0.00924092 layer.4.output 0.00019978 0.18306195 ------------------------------------------------------------------------------------- TOTAL 0.00262155 2.63543560 (elements=2,738,176) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2738176 Total Bytes 53320 BPFP 0.1558 bits/point EBPFP 0.3116 equivalent bits/point MSE 2.635436 ---------------------- -------------------------------------------------------- Time: 3.545s Load: 0.010s, Pack+Encode: 1.895s, Decode+Unpack: 1.639s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 191, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 191, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6354 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample35-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample35-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample36-layer4-item1.zst (41/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample36-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 178, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 178, 128) Output shape: (1, 178, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.0.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.1.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.1.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.2.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.2.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.3.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.3.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.4.k_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.4.v_cache: torch.Size([1, 8, 178, 128]) -> torch.Size([1, 1, 178, 1024]) layer.4.output: torch.Size([1, 178, 4096]) -> torch.Size([1, 1, 178, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,068B, BPFP=0.0469 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,584B, BPFP=0.2451 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,888B, BPFP=0.2584 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,648B, BPFP=0.2918 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,852B, BPFP=0.2568 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,604B, BPFP=0.3337 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,760B, BPFP=0.2967 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,800B, BPFP=0.3423 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,276B, BPFP=0.2316 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,884B, BPFP=0.3460 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,964B, BPFP=0.0435 ⌛️ [2/4] FRONTEND: Frontend time: 1.887s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 178, 128]) layer.0.v_cache: torch.Size([1, 8, 178, 128]) layer.1.k_cache: torch.Size([1, 8, 178, 128]) layer.1.v_cache: torch.Size([1, 8, 178, 128]) layer.2.k_cache: torch.Size([1, 8, 178, 128]) layer.2.v_cache: torch.Size([1, 8, 178, 128]) layer.3.k_cache: torch.Size([1, 8, 178, 128]) layer.3.v_cache: torch.Size([1, 8, 178, 128]) layer.4.k_cache: torch.Size([1, 8, 178, 128]) layer.4.v_cache: torch.Size([1, 8, 178, 128]) layer.4.output: torch.Size([1, 178, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.476s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 178, 128]) layer.0.v_cache: torch.Size([1, 8, 178, 128]) layer.1.k_cache: torch.Size([1, 8, 178, 128]) layer.1.v_cache: torch.Size([1, 8, 178, 128]) layer.2.k_cache: torch.Size([1, 8, 178, 128]) layer.2.v_cache: torch.Size([1, 8, 178, 128]) layer.3.k_cache: torch.Size([1, 8, 178, 128]) layer.3.v_cache: torch.Size([1, 8, 178, 128]) layer.4.k_cache: torch.Size([1, 8, 178, 128]) layer.4.v_cache: torch.Size([1, 8, 178, 128]) layer.4.output: torch.Size([1, 178, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02810422 24.33462902 layer.0.v_cache 0.00000028 0.00061744 layer.1.k_cache 0.00299977 3.51256947 layer.1.v_cache 0.00000079 0.00264753 layer.2.k_cache 0.00115350 1.53320433 layer.2.v_cache 0.00000110 0.00380574 layer.3.k_cache 0.00134504 1.76146535 layer.3.v_cache 0.00000210 0.00637002 layer.4.k_cache 0.00349525 3.54001300 layer.4.v_cache 0.00000313 0.01102700 layer.4.output 0.00017533 0.16908747 ------------------------------------------------------------------------------------- TOTAL 0.00270046 2.52733563 (elements=2,551,808) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2551808 Total Bytes 64328 BPFP 0.2017 bits/point EBPFP 0.4033 equivalent bits/point MSE 2.527336 ---------------------- -------------------------------------------------------- Time: 3.373s Load: 0.010s, Pack+Encode: 1.887s, Decode+Unpack: 1.476s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 178, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 178, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5273 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample36-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample36-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample37-layer4-item1.zst (42/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample37-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 170, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 170, 128) Output shape: (1, 170, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.0.v_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.1.k_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.1.v_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.2.k_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.2.v_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.3.k_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.3.v_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.4.k_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.4.v_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.4.output: torch.Size([1, 170, 4096]) -> torch.Size([1, 1, 170, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,020B, BPFP=0.0469 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,080B, BPFP=0.2794 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,412B, BPFP=0.2487 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,480B, BPFP=0.2978 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,536B, BPFP=0.2544 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,632B, BPFP=0.3507 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,760B, BPFP=0.3107 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,676B, BPFP=0.3528 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,108B, BPFP=0.2347 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,468B, BPFP=0.3432 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,648B, BPFP=0.0534 ⌛️ [2/4] FRONTEND: Frontend time: 1.971s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 170, 128]) layer.0.v_cache: torch.Size([1, 8, 170, 128]) layer.1.k_cache: torch.Size([1, 8, 170, 128]) layer.1.v_cache: torch.Size([1, 8, 170, 128]) layer.2.k_cache: torch.Size([1, 8, 170, 128]) layer.2.v_cache: torch.Size([1, 8, 170, 128]) layer.3.k_cache: torch.Size([1, 8, 170, 128]) layer.3.v_cache: torch.Size([1, 8, 170, 128]) layer.4.k_cache: torch.Size([1, 8, 170, 128]) layer.4.v_cache: torch.Size([1, 8, 170, 128]) layer.4.output: torch.Size([1, 170, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.421s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 170, 128]) layer.0.v_cache: torch.Size([1, 8, 170, 128]) layer.1.k_cache: torch.Size([1, 8, 170, 128]) layer.1.v_cache: torch.Size([1, 8, 170, 128]) layer.2.k_cache: torch.Size([1, 8, 170, 128]) layer.2.v_cache: torch.Size([1, 8, 170, 128]) layer.3.k_cache: torch.Size([1, 8, 170, 128]) layer.3.v_cache: torch.Size([1, 8, 170, 128]) layer.4.k_cache: torch.Size([1, 8, 170, 128]) layer.4.v_cache: torch.Size([1, 8, 170, 128]) layer.4.output: torch.Size([1, 170, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02684929 27.71917222 layer.0.v_cache 0.00000027 0.00061316 layer.1.k_cache 0.00302298 3.54170460 layer.1.v_cache 0.00000079 0.00256208 layer.2.k_cache 0.00119467 1.44203617 layer.2.v_cache 0.00000116 0.00366826 layer.3.k_cache 0.00130934 1.74671846 layer.3.v_cache 0.00000210 0.00611625 layer.4.k_cache 0.00343955 3.35279936 layer.4.v_cache 0.00000302 0.01018573 layer.4.output 0.00019176 0.19975469 ------------------------------------------------------------------------------------- TOTAL 0.00261359 2.75889965 (elements=2,437,120) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2437120 Total Bytes 63820 BPFP 0.2095 bits/point EBPFP 0.4190 equivalent bits/point MSE 2.758900 ---------------------- -------------------------------------------------------- Time: 3.403s Load: 0.010s, Pack+Encode: 1.971s, Decode+Unpack: 1.421s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 170, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7589 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample37-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample37-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample38-layer4-item1.zst (43/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample38-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 175, 128) Output shape: (1, 175, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.output: torch.Size([1, 175, 4096]) -> torch.Size([1, 1, 175, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,016B, BPFP=0.0454 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,900B, BPFP=0.2634 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,420B, BPFP=0.2420 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,952B, BPFP=0.2657 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,324B, BPFP=0.2377 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,672B, BPFP=0.2979 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,592B, BPFP=0.2943 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,332B, BPFP=0.3273 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,820B, BPFP=0.2152 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,388B, BPFP=0.3298 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,052B, BPFP=0.0452 ⌛️ [2/4] FRONTEND: Frontend time: 1.899s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) layer.4.output: torch.Size([1, 175, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.621s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) layer.4.output: torch.Size([1, 175, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02712274 25.11708147 layer.0.v_cache 0.00000026 0.00061292 layer.1.k_cache 0.00313634 3.46995361 layer.1.v_cache 0.00000079 0.00261545 layer.2.k_cache 0.00115072 1.46277536 layer.2.v_cache 0.00000109 0.00371148 layer.3.k_cache 0.00137237 1.77975185 layer.3.v_cache 0.00000208 0.00618014 layer.4.k_cache 0.00360495 3.59956334 layer.4.v_cache 0.00000302 0.01036651 layer.4.output 0.00022301 0.20485053 ------------------------------------------------------------------------------------- TOTAL 0.00266332 2.59085816 (elements=2,508,800) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2508800 Total Bytes 60468 BPFP 0.1928 bits/point EBPFP 0.3856 equivalent bits/point MSE 2.590858 ---------------------- -------------------------------------------------------- Time: 3.529s Load: 0.009s, Pack+Encode: 1.899s, Decode+Unpack: 1.621s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5909 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample38-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample38-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample39-layer4-item1.zst (44/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample39-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 177, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 177, 128) Output shape: (1, 177, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.0.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.1.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.1.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.2.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.2.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.3.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.3.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.4.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.4.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.4.output: torch.Size([1, 177, 4096]) -> torch.Size([1, 1, 177, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,072B, BPFP=0.0473 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,896B, BPFP=0.2602 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,620B, BPFP=0.2481 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,500B, BPFP=0.2869 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,792B, BPFP=0.2556 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,532B, BPFP=0.3325 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,572B, BPFP=0.2901 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,472B, BPFP=0.3298 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,220B, BPFP=0.2304 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,368B, BPFP=0.3252 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,884B, BPFP=0.0539 ⌛️ [2/4] FRONTEND: Frontend time: 1.897s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 177, 128]) layer.0.v_cache: torch.Size([1, 8, 177, 128]) layer.1.k_cache: torch.Size([1, 8, 177, 128]) layer.1.v_cache: torch.Size([1, 8, 177, 128]) layer.2.k_cache: torch.Size([1, 8, 177, 128]) layer.2.v_cache: torch.Size([1, 8, 177, 128]) layer.3.k_cache: torch.Size([1, 8, 177, 128]) layer.3.v_cache: torch.Size([1, 8, 177, 128]) layer.4.k_cache: torch.Size([1, 8, 177, 128]) layer.4.v_cache: torch.Size([1, 8, 177, 128]) layer.4.output: torch.Size([1, 177, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.519s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 177, 128]) layer.0.v_cache: torch.Size([1, 8, 177, 128]) layer.1.k_cache: torch.Size([1, 8, 177, 128]) layer.1.v_cache: torch.Size([1, 8, 177, 128]) layer.2.k_cache: torch.Size([1, 8, 177, 128]) layer.2.v_cache: torch.Size([1, 8, 177, 128]) layer.3.k_cache: torch.Size([1, 8, 177, 128]) layer.3.v_cache: torch.Size([1, 8, 177, 128]) layer.4.k_cache: torch.Size([1, 8, 177, 128]) layer.4.v_cache: torch.Size([1, 8, 177, 128]) layer.4.output: torch.Size([1, 177, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02784215 24.32190148 layer.0.v_cache 0.00000027 0.00063595 layer.1.k_cache 0.00307979 3.59273964 layer.1.v_cache 0.00000080 0.00267001 layer.2.k_cache 0.00119829 1.50501368 layer.2.v_cache 0.00000109 0.00395893 layer.3.k_cache 0.00135088 1.78713834 layer.3.v_cache 0.00000204 0.00626457 layer.4.k_cache 0.00346965 3.61765613 layer.4.v_cache 0.00000308 0.01054598 layer.4.output 0.00016766 0.20022674 ------------------------------------------------------------------------------------- TOTAL 0.00268705 2.54638797 (elements=2,537,472) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2537472 Total Bytes 63928 BPFP 0.2015 bits/point EBPFP 0.4031 equivalent bits/point MSE 2.546388 ---------------------- -------------------------------------------------------- Time: 3.427s Load: 0.011s, Pack+Encode: 1.897s, Decode+Unpack: 1.519s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 177, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5464 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample39-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample39-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample4-layer4-item1.zst (45/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample4-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 258, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.013s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 258, 128) Output shape: (1, 258, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 258, 128]) -> torch.Size([1, 1, 258, 1024]) layer.0.v_cache: torch.Size([1, 8, 258, 128]) -> torch.Size([1, 1, 258, 1024]) layer.1.k_cache: torch.Size([1, 8, 258, 128]) -> torch.Size([1, 1, 258, 1024]) layer.1.v_cache: torch.Size([1, 8, 258, 128]) -> torch.Size([1, 1, 258, 1024]) layer.2.k_cache: torch.Size([1, 8, 258, 128]) -> torch.Size([1, 1, 258, 1024]) layer.2.v_cache: torch.Size([1, 8, 258, 128]) -> torch.Size([1, 1, 258, 1024]) layer.3.k_cache: torch.Size([1, 8, 258, 128]) -> torch.Size([1, 1, 258, 1024]) layer.3.v_cache: torch.Size([1, 8, 258, 128]) -> torch.Size([1, 1, 258, 1024]) layer.4.k_cache: torch.Size([1, 8, 258, 128]) -> torch.Size([1, 1, 258, 1024]) layer.4.v_cache: torch.Size([1, 8, 258, 128]) -> torch.Size([1, 1, 258, 1024]) layer.4.output: torch.Size([1, 258, 4096]) -> torch.Size([1, 1, 258, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,468B, BPFP=0.0445 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 8,556B, BPFP=0.2591 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 7,324B, BPFP=0.2218 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 8,916B, BPFP=0.2700 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 8,720B, BPFP=0.2641 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 10,632B, BPFP=0.3219 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 10,552B, BPFP=0.3195 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 10,264B, BPFP=0.3108 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 7,348B, BPFP=0.2225 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 11,512B, BPFP=0.3486 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,376B, BPFP=0.0407 ⌛️ [2/4] FRONTEND: Frontend time: 2.348s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 258, 128]) layer.0.v_cache: torch.Size([1, 8, 258, 128]) layer.1.k_cache: torch.Size([1, 8, 258, 128]) layer.1.v_cache: torch.Size([1, 8, 258, 128]) layer.2.k_cache: torch.Size([1, 8, 258, 128]) layer.2.v_cache: torch.Size([1, 8, 258, 128]) layer.3.k_cache: torch.Size([1, 8, 258, 128]) layer.3.v_cache: torch.Size([1, 8, 258, 128]) layer.4.k_cache: torch.Size([1, 8, 258, 128]) layer.4.v_cache: torch.Size([1, 8, 258, 128]) layer.4.output: torch.Size([1, 258, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.720s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 258, 128]) layer.0.v_cache: torch.Size([1, 8, 258, 128]) layer.1.k_cache: torch.Size([1, 8, 258, 128]) layer.1.v_cache: torch.Size([1, 8, 258, 128]) layer.2.k_cache: torch.Size([1, 8, 258, 128]) layer.2.v_cache: torch.Size([1, 8, 258, 128]) layer.3.k_cache: torch.Size([1, 8, 258, 128]) layer.3.v_cache: torch.Size([1, 8, 258, 128]) layer.4.k_cache: torch.Size([1, 8, 258, 128]) layer.4.v_cache: torch.Size([1, 8, 258, 128]) layer.4.output: torch.Size([1, 258, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02582260 31.44754572 layer.0.v_cache 0.00000026 0.00062887 layer.1.k_cache 0.00299646 2.93569568 layer.1.v_cache 0.00000083 0.00264421 layer.2.k_cache 0.00117653 1.45966132 layer.2.v_cache 0.00000118 0.00391934 layer.3.k_cache 0.00132295 1.74748715 layer.3.v_cache 0.00000232 0.00653601 layer.4.k_cache 0.00348270 3.22219305 layer.4.v_cache 0.00000323 0.01117505 layer.4.output 0.00015595 0.19634364 ------------------------------------------------------------------------------------- TOTAL 0.00253092 2.97306150 (elements=3,698,688) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 3698688 Total Bytes 90668 BPFP 0.1961 bits/point EBPFP 0.3922 equivalent bits/point MSE 2.973061 ---------------------- -------------------------------------------------------- Time: 4.080s Load: 0.013s, Pack+Encode: 2.348s, Decode+Unpack: 1.720s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 258, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 258, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.9731 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample4-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample4-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample40-layer4-item1.zst (46/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample40-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 173, 128) Output shape: (1, 173, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.output: torch.Size([1, 173, 4096]) -> torch.Size([1, 1, 173, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,032B, BPFP=0.0466 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,480B, BPFP=0.2475 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,652B, BPFP=0.2552 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,004B, BPFP=0.2711 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,500B, BPFP=0.2484 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,068B, BPFP=0.3192 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,708B, BPFP=0.3029 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,860B, BPFP=0.3098 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,112B, BPFP=0.2309 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,708B, BPFP=0.3029 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,872B, BPFP=0.0437 ⌛️ [2/4] FRONTEND: Frontend time: 1.839s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) layer.4.output: torch.Size([1, 173, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.439s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) layer.4.output: torch.Size([1, 173, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02754376 27.39158745 layer.0.v_cache 0.00000027 0.00060865 layer.1.k_cache 0.00303993 3.50640763 layer.1.v_cache 0.00000078 0.00261399 layer.2.k_cache 0.00116504 1.45481317 layer.2.v_cache 0.00000115 0.00389066 layer.3.k_cache 0.00135444 1.73610745 layer.3.v_cache 0.00000213 0.00625563 layer.4.k_cache 0.00351023 3.50865253 layer.4.v_cache 0.00000311 0.01049179 layer.4.output 0.00018471 0.20591679 ------------------------------------------------------------------------------------- TOTAL 0.00266855 2.74607829 (elements=2,480,128) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2480128 Total Bytes 59996 BPFP 0.1935 bits/point EBPFP 0.3871 equivalent bits/point MSE 2.746078 ---------------------- -------------------------------------------------------- Time: 3.288s Load: 0.010s, Pack+Encode: 1.839s, Decode+Unpack: 1.439s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7461 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample40-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample40-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample41-layer4-item1.zst (47/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample41-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 179, 128) Output shape: (1, 179, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.output: torch.Size([1, 179, 4096]) -> torch.Size([1, 1, 179, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,064B, BPFP=0.0464 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,672B, BPFP=0.2476 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,684B, BPFP=0.2481 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,336B, BPFP=0.2765 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,752B, BPFP=0.2510 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,508B, BPFP=0.3277 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,948B, BPFP=0.3032 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,044B, BPFP=0.3511 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,200B, BPFP=0.2270 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,512B, BPFP=0.3279 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,064B, BPFP=0.0443 ⌛️ [2/4] FRONTEND: Frontend time: 1.892s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) layer.4.output: torch.Size([1, 179, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.642s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) layer.4.output: torch.Size([1, 179, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02616998 23.18029853 layer.0.v_cache 0.00000026 0.00062716 layer.1.k_cache 0.00300457 3.71965462 layer.1.v_cache 0.00000081 0.00257609 layer.2.k_cache 0.00116479 1.49505223 layer.2.v_cache 0.00000111 0.00378350 layer.3.k_cache 0.00135393 1.80996073 layer.3.v_cache 0.00000203 0.00635014 layer.4.k_cache 0.00348554 3.79298034 layer.4.v_cache 0.00000307 0.01051777 layer.4.output 0.00016512 0.18544064 ------------------------------------------------------------------------------------- TOTAL 0.00256047 2.48311169 (elements=2,566,144) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2566144 Total Bytes 63784 BPFP 0.1988 bits/point EBPFP 0.3977 equivalent bits/point MSE 2.483112 ---------------------- -------------------------------------------------------- Time: 3.544s Load: 0.010s, Pack+Encode: 1.892s, Decode+Unpack: 1.642s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.4831 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample41-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample41-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample42-layer4-item1.zst (48/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample42-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 157, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 157, 128) Output shape: (1, 157, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.0.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.1.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.1.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.2.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.2.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.3.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.3.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.4.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.4.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.4.output: torch.Size([1, 157, 4096]) -> torch.Size([1, 1, 157, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,012B, BPFP=0.0504 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,424B, BPFP=0.2699 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,612B, BPFP=0.2295 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,140B, BPFP=0.3055 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,556B, BPFP=0.2765 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,876B, BPFP=0.3422 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,168B, BPFP=0.3069 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,552B, BPFP=0.3758 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,780B, BPFP=0.2379 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,880B, BPFP=0.3424 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,364B, BPFP=0.0543 ⌛️ [2/4] FRONTEND: Frontend time: 1.903s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 157, 128]) layer.0.v_cache: torch.Size([1, 8, 157, 128]) layer.1.k_cache: torch.Size([1, 8, 157, 128]) layer.1.v_cache: torch.Size([1, 8, 157, 128]) layer.2.k_cache: torch.Size([1, 8, 157, 128]) layer.2.v_cache: torch.Size([1, 8, 157, 128]) layer.3.k_cache: torch.Size([1, 8, 157, 128]) layer.3.v_cache: torch.Size([1, 8, 157, 128]) layer.4.k_cache: torch.Size([1, 8, 157, 128]) layer.4.v_cache: torch.Size([1, 8, 157, 128]) layer.4.output: torch.Size([1, 157, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.426s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 157, 128]) layer.0.v_cache: torch.Size([1, 8, 157, 128]) layer.1.k_cache: torch.Size([1, 8, 157, 128]) layer.1.v_cache: torch.Size([1, 8, 157, 128]) layer.2.k_cache: torch.Size([1, 8, 157, 128]) layer.2.v_cache: torch.Size([1, 8, 157, 128]) layer.3.k_cache: torch.Size([1, 8, 157, 128]) layer.3.v_cache: torch.Size([1, 8, 157, 128]) layer.4.k_cache: torch.Size([1, 8, 157, 128]) layer.4.v_cache: torch.Size([1, 8, 157, 128]) layer.4.output: torch.Size([1, 157, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02756658 34.31348900 layer.0.v_cache 0.00000027 0.00061211 layer.1.k_cache 0.00308485 3.42440893 layer.1.v_cache 0.00000086 0.00252649 layer.2.k_cache 0.00117752 1.47885015 layer.2.v_cache 0.00000111 0.00369551 layer.3.k_cache 0.00136410 1.75873036 layer.3.v_cache 0.00000207 0.00610768 layer.4.k_cache 0.00338255 3.31387971 layer.4.v_cache 0.00000300 0.01044409 layer.4.output 0.00016164 0.19558339 ------------------------------------------------------------------------------------- TOTAL 0.00265925 3.22107697 (elements=2,250,752) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2250752 Total Bytes 59364 BPFP 0.2110 bits/point EBPFP 0.4220 equivalent bits/point MSE 3.221077 ---------------------- -------------------------------------------------------- Time: 3.337s Load: 0.009s, Pack+Encode: 1.903s, Decode+Unpack: 1.426s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 157, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 3.2211 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample42-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample42-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample43-layer4-item1.zst (49/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample43-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 174, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 174, 128) Output shape: (1, 174, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.0.v_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.1.k_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.1.v_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.2.k_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.2.v_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.3.k_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.3.v_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.4.k_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.4.v_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.4.output: torch.Size([1, 174, 4096]) -> torch.Size([1, 1, 174, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,012B, BPFP=0.0454 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,424B, BPFP=0.2435 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,488B, BPFP=0.2464 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,824B, BPFP=0.2615 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,464B, BPFP=0.2453 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,184B, BPFP=0.3226 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,440B, BPFP=0.2892 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,424B, BPFP=0.3333 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,004B, BPFP=0.2247 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,944B, BPFP=0.3118 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,756B, BPFP=0.0534 ⌛️ [2/4] FRONTEND: Frontend time: 1.944s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 174, 128]) layer.0.v_cache: torch.Size([1, 8, 174, 128]) layer.1.k_cache: torch.Size([1, 8, 174, 128]) layer.1.v_cache: torch.Size([1, 8, 174, 128]) layer.2.k_cache: torch.Size([1, 8, 174, 128]) layer.2.v_cache: torch.Size([1, 8, 174, 128]) layer.3.k_cache: torch.Size([1, 8, 174, 128]) layer.3.v_cache: torch.Size([1, 8, 174, 128]) layer.4.k_cache: torch.Size([1, 8, 174, 128]) layer.4.v_cache: torch.Size([1, 8, 174, 128]) layer.4.output: torch.Size([1, 174, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.572s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 174, 128]) layer.0.v_cache: torch.Size([1, 8, 174, 128]) layer.1.k_cache: torch.Size([1, 8, 174, 128]) layer.1.v_cache: torch.Size([1, 8, 174, 128]) layer.2.k_cache: torch.Size([1, 8, 174, 128]) layer.2.v_cache: torch.Size([1, 8, 174, 128]) layer.3.k_cache: torch.Size([1, 8, 174, 128]) layer.3.v_cache: torch.Size([1, 8, 174, 128]) layer.4.k_cache: torch.Size([1, 8, 174, 128]) layer.4.v_cache: torch.Size([1, 8, 174, 128]) layer.4.output: torch.Size([1, 174, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02674412 26.36441777 layer.0.v_cache 0.00000027 0.00063859 layer.1.k_cache 0.00309791 3.57000347 layer.1.v_cache 0.00000087 0.00267540 layer.2.k_cache 0.00117281 1.48681676 layer.2.v_cache 0.00000114 0.00393271 layer.3.k_cache 0.00132088 1.76760426 layer.3.v_cache 0.00000225 0.00655862 layer.4.k_cache 0.00347387 3.54179680 layer.4.v_cache 0.00000314 0.01062944 layer.4.output 0.00018404 0.20688381 ------------------------------------------------------------------------------------- TOTAL 0.00261096 2.68447208 (elements=2,494,464) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2494464 Total Bytes 60964 BPFP 0.1955 bits/point EBPFP 0.3910 equivalent bits/point MSE 2.684472 ---------------------- -------------------------------------------------------- Time: 3.525s Load: 0.009s, Pack+Encode: 1.944s, Decode+Unpack: 1.572s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 174, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6845 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample43-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample43-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample44-layer4-item1.zst (50/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample44-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 177, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 177, 128) Output shape: (1, 177, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.0.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.1.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.1.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.2.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.2.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.3.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.3.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.4.k_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.4.v_cache: torch.Size([1, 8, 177, 128]) -> torch.Size([1, 1, 177, 1024]) layer.4.output: torch.Size([1, 177, 4096]) -> torch.Size([1, 1, 177, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,040B, BPFP=0.0459 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,052B, BPFP=0.2671 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,824B, BPFP=0.2571 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,508B, BPFP=0.2873 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,700B, BPFP=0.2516 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 8,020B, BPFP=0.3540 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,756B, BPFP=0.2982 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,456B, BPFP=0.3291 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,168B, BPFP=0.2281 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,796B, BPFP=0.3441 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,848B, BPFP=0.0425 ⌛️ [2/4] FRONTEND: Frontend time: 1.864s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 177, 128]) layer.0.v_cache: torch.Size([1, 8, 177, 128]) layer.1.k_cache: torch.Size([1, 8, 177, 128]) layer.1.v_cache: torch.Size([1, 8, 177, 128]) layer.2.k_cache: torch.Size([1, 8, 177, 128]) layer.2.v_cache: torch.Size([1, 8, 177, 128]) layer.3.k_cache: torch.Size([1, 8, 177, 128]) layer.3.v_cache: torch.Size([1, 8, 177, 128]) layer.4.k_cache: torch.Size([1, 8, 177, 128]) layer.4.v_cache: torch.Size([1, 8, 177, 128]) layer.4.output: torch.Size([1, 177, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.517s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 177, 128]) layer.0.v_cache: torch.Size([1, 8, 177, 128]) layer.1.k_cache: torch.Size([1, 8, 177, 128]) layer.1.v_cache: torch.Size([1, 8, 177, 128]) layer.2.k_cache: torch.Size([1, 8, 177, 128]) layer.2.v_cache: torch.Size([1, 8, 177, 128]) layer.3.k_cache: torch.Size([1, 8, 177, 128]) layer.3.v_cache: torch.Size([1, 8, 177, 128]) layer.4.k_cache: torch.Size([1, 8, 177, 128]) layer.4.v_cache: torch.Size([1, 8, 177, 128]) layer.4.output: torch.Size([1, 177, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02710410 24.96956380 layer.0.v_cache 0.00000026 0.00061445 layer.1.k_cache 0.00310284 3.70789333 layer.1.v_cache 0.00000080 0.00261122 layer.2.k_cache 0.00117362 1.54880261 layer.2.v_cache 0.00000106 0.00380658 layer.3.k_cache 0.00136356 1.80671287 layer.3.v_cache 0.00000205 0.00602411 layer.4.k_cache 0.00349505 3.61150192 layer.4.v_cache 0.00000305 0.01049055 layer.4.output 0.00021483 0.20452844 ------------------------------------------------------------------------------------- TOTAL 0.00265041 2.60615252 (elements=2,537,472) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2537472 Total Bytes 64168 BPFP 0.2023 bits/point EBPFP 0.4046 equivalent bits/point MSE 2.606153 ---------------------- -------------------------------------------------------- Time: 3.390s Load: 0.009s, Pack+Encode: 1.864s, Decode+Unpack: 1.517s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 177, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 177, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6062 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample44-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample44-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample45-layer4-item1.zst (51/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample45-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 165, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 165, 128) Output shape: (1, 165, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.0.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.1.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.1.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.2.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.2.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.3.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.3.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.4.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.4.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.4.output: torch.Size([1, 165, 4096]) -> torch.Size([1, 1, 165, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,020B, BPFP=0.0483 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,016B, BPFP=0.2848 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,460B, BPFP=0.2585 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,340B, BPFP=0.3002 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,848B, BPFP=0.2769 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,752B, BPFP=0.3197 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,128B, BPFP=0.3375 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,020B, BPFP=0.3324 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,164B, BPFP=0.2445 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,112B, BPFP=0.3367 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,592B, BPFP=0.0425 ⌛️ [2/4] FRONTEND: Frontend time: 1.958s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 165, 128]) layer.0.v_cache: torch.Size([1, 8, 165, 128]) layer.1.k_cache: torch.Size([1, 8, 165, 128]) layer.1.v_cache: torch.Size([1, 8, 165, 128]) layer.2.k_cache: torch.Size([1, 8, 165, 128]) layer.2.v_cache: torch.Size([1, 8, 165, 128]) layer.3.k_cache: torch.Size([1, 8, 165, 128]) layer.3.v_cache: torch.Size([1, 8, 165, 128]) layer.4.k_cache: torch.Size([1, 8, 165, 128]) layer.4.v_cache: torch.Size([1, 8, 165, 128]) layer.4.output: torch.Size([1, 165, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.440s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 165, 128]) layer.0.v_cache: torch.Size([1, 8, 165, 128]) layer.1.k_cache: torch.Size([1, 8, 165, 128]) layer.1.v_cache: torch.Size([1, 8, 165, 128]) layer.2.k_cache: torch.Size([1, 8, 165, 128]) layer.2.v_cache: torch.Size([1, 8, 165, 128]) layer.3.k_cache: torch.Size([1, 8, 165, 128]) layer.3.v_cache: torch.Size([1, 8, 165, 128]) layer.4.k_cache: torch.Size([1, 8, 165, 128]) layer.4.v_cache: torch.Size([1, 8, 165, 128]) layer.4.output: torch.Size([1, 165, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02820997 25.96450343 layer.0.v_cache 0.00000025 0.00057290 layer.1.k_cache 0.00309343 3.19037568 layer.1.v_cache 0.00000073 0.00224480 layer.2.k_cache 0.00118989 1.47254435 layer.2.v_cache 0.00000098 0.00329714 layer.3.k_cache 0.00136500 1.66631581 layer.3.v_cache 0.00000190 0.00541074 layer.4.k_cache 0.00352692 3.56427557 layer.4.v_cache 0.00000287 0.00939320 layer.4.output 0.00017443 0.18889223 ------------------------------------------------------------------------------------- TOTAL 0.00272069 2.61675018 (elements=2,365,440) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2365440 Total Bytes 61452 BPFP 0.2078 bits/point EBPFP 0.4157 equivalent bits/point MSE 2.616750 ---------------------- -------------------------------------------------------- Time: 3.407s Load: 0.009s, Pack+Encode: 1.958s, Decode+Unpack: 1.440s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 165, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6168 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample45-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample45-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample47-layer4-item1.zst (52/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample47-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 175, 128) Output shape: (1, 175, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.output: torch.Size([1, 175, 4096]) -> torch.Size([1, 1, 175, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,036B, BPFP=0.0462 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,592B, BPFP=0.2496 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,488B, BPFP=0.2450 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,088B, BPFP=0.2718 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,392B, BPFP=0.2407 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,968B, BPFP=0.3111 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,252B, BPFP=0.2791 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,372B, BPFP=0.3291 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,736B, BPFP=0.2114 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,260B, BPFP=0.3241 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,636B, BPFP=0.0406 ⌛️ [2/4] FRONTEND: Frontend time: 1.878s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) layer.4.output: torch.Size([1, 175, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.577s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) layer.4.output: torch.Size([1, 175, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02684375 26.21250000 layer.0.v_cache 0.00000026 0.00058681 layer.1.k_cache 0.00307220 3.45914934 layer.1.v_cache 0.00000076 0.00239211 layer.2.k_cache 0.00114531 1.45625924 layer.2.v_cache 0.00000107 0.00353923 layer.3.k_cache 0.00135081 1.72216448 layer.3.v_cache 0.00000217 0.00605372 layer.4.k_cache 0.00336849 3.54325021 layer.4.v_cache 0.00000291 0.00997588 layer.4.output 0.00018329 0.18777222 ------------------------------------------------------------------------------------- TOTAL 0.00260863 2.65478285 (elements=2,508,800) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2508800 Total Bytes 59820 BPFP 0.1908 bits/point EBPFP 0.3815 equivalent bits/point MSE 2.654783 ---------------------- -------------------------------------------------------- Time: 3.465s Load: 0.009s, Pack+Encode: 1.878s, Decode+Unpack: 1.577s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6548 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample47-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample47-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample48-layer4-item1.zst (53/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample48-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 179, 128) Output shape: (1, 179, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.output: torch.Size([1, 179, 4096]) -> torch.Size([1, 1, 179, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,088B, BPFP=0.0475 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,680B, BPFP=0.2479 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,604B, BPFP=0.2446 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,828B, BPFP=0.2980 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,696B, BPFP=0.2486 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,908B, BPFP=0.3451 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,840B, BPFP=0.2985 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,532B, BPFP=0.3287 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,256B, BPFP=0.2294 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,612B, BPFP=0.3322 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,388B, BPFP=0.0479 ⌛️ [2/4] FRONTEND: Frontend time: 1.898s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) layer.4.output: torch.Size([1, 179, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.436s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) layer.4.output: torch.Size([1, 179, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02697467 24.25603941 layer.0.v_cache 0.00000027 0.00060999 layer.1.k_cache 0.00293179 3.56501847 layer.1.v_cache 0.00000082 0.00259010 layer.2.k_cache 0.00115626 1.52967587 layer.2.v_cache 0.00000116 0.00375565 layer.3.k_cache 0.00134185 1.84830172 layer.3.v_cache 0.00000205 0.00614879 layer.4.k_cache 0.00342791 3.76930655 layer.4.v_cache 0.00000307 0.01049132 layer.4.output 0.00018990 0.19016692 ------------------------------------------------------------------------------------- TOTAL 0.00261425 2.55375754 (elements=2,566,144) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2566144 Total Bytes 64432 BPFP 0.2009 bits/point EBPFP 0.4017 equivalent bits/point MSE 2.553758 ---------------------- -------------------------------------------------------- Time: 3.343s Load: 0.010s, Pack+Encode: 1.898s, Decode+Unpack: 1.436s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5538 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample48-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample48-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample49-layer4-item1.zst (54/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample49-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 169, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 169, 128) Output shape: (1, 169, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.output: torch.Size([1, 169, 4096]) -> torch.Size([1, 1, 169, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 996B, BPFP=0.0460 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,148B, BPFP=0.2842 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,876B, BPFP=0.2716 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,080B, BPFP=0.2811 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,528B, BPFP=0.2555 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,164B, BPFP=0.3312 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,656B, BPFP=0.3077 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,148B, BPFP=0.3304 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,508B, BPFP=0.2546 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,284B, BPFP=0.3367 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,768B, BPFP=0.0435 ⌛️ [2/4] FRONTEND: Frontend time: 1.942s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) layer.4.output: torch.Size([1, 169, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.561s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) layer.4.output: torch.Size([1, 169, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02698109 24.74685651 layer.0.v_cache 0.00000027 0.00062553 layer.1.k_cache 0.00303223 3.75616275 layer.1.v_cache 0.00000079 0.00268677 layer.2.k_cache 0.00117179 1.46049490 layer.2.v_cache 0.00000111 0.00386817 layer.3.k_cache 0.00133305 1.71603014 layer.3.v_cache 0.00000207 0.00624991 layer.4.k_cache 0.00348428 3.18559960 layer.4.v_cache 0.00000320 0.01092366 layer.4.output 0.00019031 0.18713733 ------------------------------------------------------------------------------------- TOTAL 0.00262651 2.54557481 (elements=2,422,784) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2422784 Total Bytes 62156 BPFP 0.2052 bits/point EBPFP 0.4105 equivalent bits/point MSE 2.545575 ---------------------- -------------------------------------------------------- Time: 3.512s Load: 0.009s, Pack+Encode: 1.942s, Decode+Unpack: 1.561s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 169, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5456 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample49-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample49-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample5-layer4-item1.zst (55/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample5-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 221, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 221, 128) Output shape: (1, 221, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 221, 128]) -> torch.Size([1, 1, 221, 1024]) layer.0.v_cache: torch.Size([1, 8, 221, 128]) -> torch.Size([1, 1, 221, 1024]) layer.1.k_cache: torch.Size([1, 8, 221, 128]) -> torch.Size([1, 1, 221, 1024]) layer.1.v_cache: torch.Size([1, 8, 221, 128]) -> torch.Size([1, 1, 221, 1024]) layer.2.k_cache: torch.Size([1, 8, 221, 128]) -> torch.Size([1, 1, 221, 1024]) layer.2.v_cache: torch.Size([1, 8, 221, 128]) -> torch.Size([1, 1, 221, 1024]) layer.3.k_cache: torch.Size([1, 8, 221, 128]) -> torch.Size([1, 1, 221, 1024]) layer.3.v_cache: torch.Size([1, 8, 221, 128]) -> torch.Size([1, 1, 221, 1024]) layer.4.k_cache: torch.Size([1, 8, 221, 128]) -> torch.Size([1, 1, 221, 1024]) layer.4.v_cache: torch.Size([1, 8, 221, 128]) -> torch.Size([1, 1, 221, 1024]) layer.4.output: torch.Size([1, 221, 4096]) -> torch.Size([1, 1, 221, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,244B, BPFP=0.0440 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,896B, BPFP=0.2438 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,944B, BPFP=0.2101 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 7,500B, BPFP=0.2651 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 6,804B, BPFP=0.2405 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 8,488B, BPFP=0.3001 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,176B, BPFP=0.2890 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,684B, BPFP=0.3070 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,696B, BPFP=0.2014 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 9,084B, BPFP=0.3211 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,340B, BPFP=0.0384 ⌛️ [2/4] FRONTEND: Frontend time: 2.246s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 221, 128]) layer.0.v_cache: torch.Size([1, 8, 221, 128]) layer.1.k_cache: torch.Size([1, 8, 221, 128]) layer.1.v_cache: torch.Size([1, 8, 221, 128]) layer.2.k_cache: torch.Size([1, 8, 221, 128]) layer.2.v_cache: torch.Size([1, 8, 221, 128]) layer.3.k_cache: torch.Size([1, 8, 221, 128]) layer.3.v_cache: torch.Size([1, 8, 221, 128]) layer.4.k_cache: torch.Size([1, 8, 221, 128]) layer.4.v_cache: torch.Size([1, 8, 221, 128]) layer.4.output: torch.Size([1, 221, 4096]) ⌛️ [3/4] BACKEND: Backend time: 2.079s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 221, 128]) layer.0.v_cache: torch.Size([1, 8, 221, 128]) layer.1.k_cache: torch.Size([1, 8, 221, 128]) layer.1.v_cache: torch.Size([1, 8, 221, 128]) layer.2.k_cache: torch.Size([1, 8, 221, 128]) layer.2.v_cache: torch.Size([1, 8, 221, 128]) layer.3.k_cache: torch.Size([1, 8, 221, 128]) layer.3.v_cache: torch.Size([1, 8, 221, 128]) layer.4.k_cache: torch.Size([1, 8, 221, 128]) layer.4.v_cache: torch.Size([1, 8, 221, 128]) layer.4.output: torch.Size([1, 221, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02713613 26.55960787 layer.0.v_cache 0.00000026 0.00059978 layer.1.k_cache 0.00294223 3.09482240 layer.1.v_cache 0.00000078 0.00242067 layer.2.k_cache 0.00116104 1.41032969 layer.2.v_cache 0.00000109 0.00360804 layer.3.k_cache 0.00132333 1.68327739 layer.3.v_cache 0.00000207 0.00601863 layer.4.k_cache 0.00348143 3.58237481 layer.4.v_cache 0.00000308 0.01038963 layer.4.output 0.00015143 0.20200684 ------------------------------------------------------------------------------------- TOTAL 0.00261837 2.65439116 (elements=3,168,256) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 3168256 Total Bytes 72856 BPFP 0.1840 bits/point EBPFP 0.3679 equivalent bits/point MSE 2.654391 ---------------------- -------------------------------------------------------- Time: 4.336s Load: 0.011s, Pack+Encode: 2.246s, Decode+Unpack: 2.079s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 221, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 221, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6544 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample5-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample5-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample50-layer4-item1.zst (56/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample50-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 175, 128) Output shape: (1, 175, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.output: torch.Size([1, 175, 4096]) -> torch.Size([1, 1, 175, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,024B, BPFP=0.0457 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,248B, BPFP=0.2343 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,568B, BPFP=0.2486 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,116B, BPFP=0.2730 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,460B, BPFP=0.2437 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,536B, BPFP=0.2918 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,456B, BPFP=0.2882 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,752B, BPFP=0.3014 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,060B, BPFP=0.2259 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,040B, BPFP=0.3143 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,984B, BPFP=0.0445 ⌛️ [2/4] FRONTEND: Frontend time: 1.895s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) layer.4.output: torch.Size([1, 175, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.457s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) layer.4.output: torch.Size([1, 175, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02716402 24.51893136 layer.0.v_cache 0.00000026 0.00058753 layer.1.k_cache 0.00305459 3.52178292 layer.1.v_cache 0.00000078 0.00253882 layer.2.k_cache 0.00115271 1.47992467 layer.2.v_cache 0.00000108 0.00369503 layer.3.k_cache 0.00136835 1.71955043 layer.3.v_cache 0.00000209 0.00611423 layer.4.k_cache 0.00338224 3.42123186 layer.4.v_cache 0.00000309 0.01039034 layer.4.output 0.00017352 0.17875359 ------------------------------------------------------------------------------------- TOTAL 0.00263024 2.52855440 (elements=2,508,800) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2508800 Total Bytes 59244 BPFP 0.1889 bits/point EBPFP 0.3778 equivalent bits/point MSE 2.528554 ---------------------- -------------------------------------------------------- Time: 3.361s Load: 0.010s, Pack+Encode: 1.895s, Decode+Unpack: 1.457s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5286 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample50-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample50-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample51-layer4-item1.zst (57/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample51-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 184, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 184, 128) Output shape: (1, 184, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.0.v_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.1.k_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.1.v_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.2.k_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.2.v_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.3.k_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.3.v_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.4.k_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.4.v_cache: torch.Size([1, 8, 184, 128]) -> torch.Size([1, 1, 184, 1024]) layer.4.output: torch.Size([1, 184, 4096]) -> torch.Size([1, 1, 184, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,220B, BPFP=0.0518 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,872B, BPFP=0.2493 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,988B, BPFP=0.2118 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,304B, BPFP=0.2677 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,476B, BPFP=0.2325 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,200B, BPFP=0.3057 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,496B, BPFP=0.2758 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,324B, BPFP=0.3110 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,952B, BPFP=0.2103 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,176B, BPFP=0.3047 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,224B, BPFP=0.0555 ⌛️ [2/4] FRONTEND: Frontend time: 1.943s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 184, 128]) layer.0.v_cache: torch.Size([1, 8, 184, 128]) layer.1.k_cache: torch.Size([1, 8, 184, 128]) layer.1.v_cache: torch.Size([1, 8, 184, 128]) layer.2.k_cache: torch.Size([1, 8, 184, 128]) layer.2.v_cache: torch.Size([1, 8, 184, 128]) layer.3.k_cache: torch.Size([1, 8, 184, 128]) layer.3.v_cache: torch.Size([1, 8, 184, 128]) layer.4.k_cache: torch.Size([1, 8, 184, 128]) layer.4.v_cache: torch.Size([1, 8, 184, 128]) layer.4.output: torch.Size([1, 184, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.579s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 184, 128]) layer.0.v_cache: torch.Size([1, 8, 184, 128]) layer.1.k_cache: torch.Size([1, 8, 184, 128]) layer.1.v_cache: torch.Size([1, 8, 184, 128]) layer.2.k_cache: torch.Size([1, 8, 184, 128]) layer.2.v_cache: torch.Size([1, 8, 184, 128]) layer.3.k_cache: torch.Size([1, 8, 184, 128]) layer.3.v_cache: torch.Size([1, 8, 184, 128]) layer.4.k_cache: torch.Size([1, 8, 184, 128]) layer.4.v_cache: torch.Size([1, 8, 184, 128]) layer.4.output: torch.Size([1, 184, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02750326 26.90563434 layer.0.v_cache 0.00000026 0.00063486 layer.1.k_cache 0.00297607 3.63073929 layer.1.v_cache 0.00000082 0.00280465 layer.2.k_cache 0.00115877 1.60006614 layer.2.v_cache 0.00000113 0.00387127 layer.3.k_cache 0.00134157 1.98294532 layer.3.v_cache 0.00000216 0.00689008 layer.4.k_cache 0.00343623 3.85154956 layer.4.v_cache 0.00000338 0.01121109 layer.4.output 0.00018269 0.19392240 ------------------------------------------------------------------------------------- TOTAL 0.00265389 2.76943116 (elements=2,637,824) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2637824 Total Bytes 62232 BPFP 0.1887 bits/point EBPFP 0.3775 equivalent bits/point MSE 2.769431 ---------------------- -------------------------------------------------------- Time: 3.531s Load: 0.010s, Pack+Encode: 1.943s, Decode+Unpack: 1.579s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 184, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 184, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7694 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample51-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample51-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample52-layer4-item1.zst (58/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample52-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 169, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 169, 128) Output shape: (1, 169, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.output: torch.Size([1, 169, 4096]) -> torch.Size([1, 1, 169, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,012B, BPFP=0.0468 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,616B, BPFP=0.2596 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,500B, BPFP=0.2543 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,320B, BPFP=0.2922 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,696B, BPFP=0.2633 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,932B, BPFP=0.3205 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,608B, BPFP=0.3055 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,120B, BPFP=0.3291 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,376B, BPFP=0.2485 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,188B, BPFP=0.3323 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,948B, BPFP=0.0456 ⌛️ [2/4] FRONTEND: Frontend time: 1.875s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) layer.4.output: torch.Size([1, 169, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.590s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) layer.4.output: torch.Size([1, 169, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02710367 27.80128109 layer.0.v_cache 0.00000027 0.00059899 layer.1.k_cache 0.00315269 3.48552672 layer.1.v_cache 0.00000078 0.00254672 layer.2.k_cache 0.00113874 1.47989469 layer.2.v_cache 0.00000109 0.00369517 layer.3.k_cache 0.00134854 1.74434287 layer.3.v_cache 0.00000205 0.00606967 layer.4.k_cache 0.00348762 3.41418421 layer.4.v_cache 0.00000318 0.01050813 layer.4.output 0.00019737 0.20751012 ------------------------------------------------------------------------------------- TOTAL 0.00264486 2.76990634 (elements=2,422,784) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2422784 Total Bytes 61316 BPFP 0.2025 bits/point EBPFP 0.4049 equivalent bits/point MSE 2.769906 ---------------------- -------------------------------------------------------- Time: 3.474s Load: 0.009s, Pack+Encode: 1.875s, Decode+Unpack: 1.590s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 169, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7699 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample52-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample52-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample53-layer4-item1.zst (59/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample53-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 173, 128) Output shape: (1, 173, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.output: torch.Size([1, 173, 4096]) -> torch.Size([1, 1, 173, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,032B, BPFP=0.0466 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,044B, BPFP=0.2729 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,612B, BPFP=0.2534 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,768B, BPFP=0.3056 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,800B, BPFP=0.2619 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,032B, BPFP=0.3176 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,624B, BPFP=0.2991 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,468B, BPFP=0.3372 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,084B, BPFP=0.2296 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,920B, BPFP=0.3125 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,352B, BPFP=0.0604 ⌛️ [2/4] FRONTEND: Frontend time: 1.975s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) layer.4.output: torch.Size([1, 173, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.432s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) layer.4.output: torch.Size([1, 173, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02649080 26.32781453 layer.0.v_cache 0.00000026 0.00063524 layer.1.k_cache 0.00311801 3.38041943 layer.1.v_cache 0.00000082 0.00265410 layer.2.k_cache 0.00118182 1.47616895 layer.2.v_cache 0.00000117 0.00378311 layer.3.k_cache 0.00133769 1.76603619 layer.3.v_cache 0.00000224 0.00640270 layer.4.k_cache 0.00352722 3.67174446 layer.4.v_cache 0.00000310 0.01078144 layer.4.output 0.00018175 0.19460899 ------------------------------------------------------------------------------------- TOTAL 0.00259930 2.67320544 (elements=2,480,128) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2480128 Total Bytes 63736 BPFP 0.2056 bits/point EBPFP 0.4112 equivalent bits/point MSE 2.673205 ---------------------- -------------------------------------------------------- Time: 3.416s Load: 0.009s, Pack+Encode: 1.975s, Decode+Unpack: 1.432s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6732 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample53-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample53-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample54-layer4-item1.zst (60/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample54-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 176, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 176, 128) Output shape: (1, 176, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 176, 128]) -> torch.Size([1, 1, 176, 1024]) layer.0.v_cache: torch.Size([1, 8, 176, 128]) -> torch.Size([1, 1, 176, 1024]) layer.1.k_cache: torch.Size([1, 8, 176, 128]) -> torch.Size([1, 1, 176, 1024]) layer.1.v_cache: torch.Size([1, 8, 176, 128]) -> torch.Size([1, 1, 176, 1024]) layer.2.k_cache: torch.Size([1, 8, 176, 128]) -> torch.Size([1, 1, 176, 1024]) layer.2.v_cache: torch.Size([1, 8, 176, 128]) -> torch.Size([1, 1, 176, 1024]) layer.3.k_cache: torch.Size([1, 8, 176, 128]) -> torch.Size([1, 1, 176, 1024]) layer.3.v_cache: torch.Size([1, 8, 176, 128]) -> torch.Size([1, 1, 176, 1024]) layer.4.k_cache: torch.Size([1, 8, 176, 128]) -> torch.Size([1, 1, 176, 1024]) layer.4.v_cache: torch.Size([1, 8, 176, 128]) -> torch.Size([1, 1, 176, 1024]) layer.4.output: torch.Size([1, 176, 4096]) -> torch.Size([1, 1, 176, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,080B, BPFP=0.0479 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,108B, BPFP=0.2267 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,880B, BPFP=0.2610 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,072B, BPFP=0.2695 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,764B, BPFP=0.2559 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,328B, BPFP=0.3253 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,688B, BPFP=0.2969 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,536B, BPFP=0.3345 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,192B, BPFP=0.2305 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,244B, BPFP=0.3216 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,324B, BPFP=0.0480 ⌛️ [2/4] FRONTEND: Frontend time: 1.903s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 176, 128]) layer.0.v_cache: torch.Size([1, 8, 176, 128]) layer.1.k_cache: torch.Size([1, 8, 176, 128]) layer.1.v_cache: torch.Size([1, 8, 176, 128]) layer.2.k_cache: torch.Size([1, 8, 176, 128]) layer.2.v_cache: torch.Size([1, 8, 176, 128]) layer.3.k_cache: torch.Size([1, 8, 176, 128]) layer.3.v_cache: torch.Size([1, 8, 176, 128]) layer.4.k_cache: torch.Size([1, 8, 176, 128]) layer.4.v_cache: torch.Size([1, 8, 176, 128]) layer.4.output: torch.Size([1, 176, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.692s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 176, 128]) layer.0.v_cache: torch.Size([1, 8, 176, 128]) layer.1.k_cache: torch.Size([1, 8, 176, 128]) layer.1.v_cache: torch.Size([1, 8, 176, 128]) layer.2.k_cache: torch.Size([1, 8, 176, 128]) layer.2.v_cache: torch.Size([1, 8, 176, 128]) layer.3.k_cache: torch.Size([1, 8, 176, 128]) layer.3.v_cache: torch.Size([1, 8, 176, 128]) layer.4.k_cache: torch.Size([1, 8, 176, 128]) layer.4.v_cache: torch.Size([1, 8, 176, 128]) layer.4.output: torch.Size([1, 176, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02681748 25.82058438 layer.0.v_cache 0.00000027 0.00062884 layer.1.k_cache 0.00299919 3.79845255 layer.1.v_cache 0.00000086 0.00265599 layer.2.k_cache 0.00118180 1.45852245 layer.2.v_cache 0.00000111 0.00387658 layer.3.k_cache 0.00131753 1.74838916 layer.3.v_cache 0.00000211 0.00638411 layer.4.k_cache 0.00351615 3.51119752 layer.4.v_cache 0.00000302 0.01049953 layer.4.output 0.00017220 0.19263363 ------------------------------------------------------------------------------------- TOTAL 0.00260917 2.65226612 (elements=2,523,136) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2523136 Total Bytes 62216 BPFP 0.1973 bits/point EBPFP 0.3945 equivalent bits/point MSE 2.652266 ---------------------- -------------------------------------------------------- Time: 3.605s Load: 0.010s, Pack+Encode: 1.903s, Decode+Unpack: 1.692s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 176, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 176, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6523 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample54-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample54-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample55-layer4-item1.zst (61/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample55-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 164, 128) Output shape: (1, 164, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.output: torch.Size([1, 164, 4096]) -> torch.Size([1, 1, 164, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,024B, BPFP=0.0488 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,800B, BPFP=0.2763 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,228B, BPFP=0.2490 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,592B, BPFP=0.3140 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,900B, BPFP=0.2811 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,100B, BPFP=0.3382 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,020B, BPFP=0.3344 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,376B, BPFP=0.3514 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,468B, BPFP=0.2605 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,576B, BPFP=0.3609 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,648B, BPFP=0.0434 ⌛️ [2/4] FRONTEND: Frontend time: 1.896s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) layer.4.output: torch.Size([1, 164, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.440s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) layer.4.output: torch.Size([1, 164, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02653997 25.55380323 layer.0.v_cache 0.00000026 0.00059690 layer.1.k_cache 0.00301226 3.15024846 layer.1.v_cache 0.00000076 0.00254660 layer.2.k_cache 0.00116255 1.46792072 layer.2.v_cache 0.00000105 0.00356151 layer.3.k_cache 0.00132644 1.75859758 layer.3.v_cache 0.00000197 0.00586522 layer.4.k_cache 0.00346445 3.48483946 layer.4.v_cache 0.00000298 0.01001705 layer.4.output 0.00016888 0.18552813 ------------------------------------------------------------------------------------- TOTAL 0.00258487 2.58429352 (elements=2,351,104) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2351104 Total Bytes 62732 BPFP 0.2135 bits/point EBPFP 0.4269 equivalent bits/point MSE 2.584294 ---------------------- -------------------------------------------------------- Time: 3.346s Load: 0.010s, Pack+Encode: 1.896s, Decode+Unpack: 1.440s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5843 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample55-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample55-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample56-layer4-item1.zst (62/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample56-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 179, 128) Output shape: (1, 179, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) -> torch.Size([1, 1, 179, 1024]) layer.4.output: torch.Size([1, 179, 4096]) -> torch.Size([1, 1, 179, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,060B, BPFP=0.0463 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,612B, BPFP=0.2449 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,424B, BPFP=0.2367 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,968B, BPFP=0.3041 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,812B, BPFP=0.2537 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 8,116B, BPFP=0.3542 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,696B, BPFP=0.2922 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,088B, BPFP=0.3530 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,108B, BPFP=0.2229 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,824B, BPFP=0.3415 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,972B, BPFP=0.0433 ⌛️ [2/4] FRONTEND: Frontend time: 1.969s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) layer.4.output: torch.Size([1, 179, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.507s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 179, 128]) layer.0.v_cache: torch.Size([1, 8, 179, 128]) layer.1.k_cache: torch.Size([1, 8, 179, 128]) layer.1.v_cache: torch.Size([1, 8, 179, 128]) layer.2.k_cache: torch.Size([1, 8, 179, 128]) layer.2.v_cache: torch.Size([1, 8, 179, 128]) layer.3.k_cache: torch.Size([1, 8, 179, 128]) layer.3.v_cache: torch.Size([1, 8, 179, 128]) layer.4.k_cache: torch.Size([1, 8, 179, 128]) layer.4.v_cache: torch.Size([1, 8, 179, 128]) layer.4.output: torch.Size([1, 179, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02669685 25.03320858 layer.0.v_cache 0.00000027 0.00059235 layer.1.k_cache 0.00301279 3.54068794 layer.1.v_cache 0.00000079 0.00251156 layer.2.k_cache 0.00116913 1.52314869 layer.2.v_cache 0.00000107 0.00358827 layer.3.k_cache 0.00133970 1.79264891 layer.3.v_cache 0.00000203 0.00596302 layer.4.k_cache 0.00352847 3.80800795 layer.4.v_cache 0.00000295 0.01016149 layer.4.output 0.00017485 0.17126031 ------------------------------------------------------------------------------------- TOTAL 0.00260382 2.60039714 (elements=2,566,144) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2566144 Total Bytes 64680 BPFP 0.2016 bits/point EBPFP 0.4033 equivalent bits/point MSE 2.600397 ---------------------- -------------------------------------------------------- Time: 3.486s Load: 0.009s, Pack+Encode: 1.969s, Decode+Unpack: 1.507s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 179, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 179, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6004 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample56-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample56-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample57-layer4-item1.zst (63/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample57-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 188, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 188, 128) Output shape: (1, 188, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 188, 128]) -> torch.Size([1, 1, 188, 1024]) layer.0.v_cache: torch.Size([1, 8, 188, 128]) -> torch.Size([1, 1, 188, 1024]) layer.1.k_cache: torch.Size([1, 8, 188, 128]) -> torch.Size([1, 1, 188, 1024]) layer.1.v_cache: torch.Size([1, 8, 188, 128]) -> torch.Size([1, 1, 188, 1024]) layer.2.k_cache: torch.Size([1, 8, 188, 128]) -> torch.Size([1, 1, 188, 1024]) layer.2.v_cache: torch.Size([1, 8, 188, 128]) -> torch.Size([1, 1, 188, 1024]) layer.3.k_cache: torch.Size([1, 8, 188, 128]) -> torch.Size([1, 1, 188, 1024]) layer.3.v_cache: torch.Size([1, 8, 188, 128]) -> torch.Size([1, 1, 188, 1024]) layer.4.k_cache: torch.Size([1, 8, 188, 128]) -> torch.Size([1, 1, 188, 1024]) layer.4.v_cache: torch.Size([1, 8, 188, 128]) -> torch.Size([1, 1, 188, 1024]) layer.4.output: torch.Size([1, 188, 4096]) -> torch.Size([1, 1, 188, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,204B, BPFP=0.0500 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,180B, BPFP=0.2153 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,672B, BPFP=0.1941 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,696B, BPFP=0.2783 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,240B, BPFP=0.2178 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,652B, BPFP=0.2764 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 5,852B, BPFP=0.2432 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,912B, BPFP=0.2872 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,548B, BPFP=0.1890 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,396B, BPFP=0.3073 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,080B, BPFP=0.0424 ⌛️ [2/4] FRONTEND: Frontend time: 1.863s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 188, 128]) layer.0.v_cache: torch.Size([1, 8, 188, 128]) layer.1.k_cache: torch.Size([1, 8, 188, 128]) layer.1.v_cache: torch.Size([1, 8, 188, 128]) layer.2.k_cache: torch.Size([1, 8, 188, 128]) layer.2.v_cache: torch.Size([1, 8, 188, 128]) layer.3.k_cache: torch.Size([1, 8, 188, 128]) layer.3.v_cache: torch.Size([1, 8, 188, 128]) layer.4.k_cache: torch.Size([1, 8, 188, 128]) layer.4.v_cache: torch.Size([1, 8, 188, 128]) layer.4.output: torch.Size([1, 188, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.588s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 188, 128]) layer.0.v_cache: torch.Size([1, 8, 188, 128]) layer.1.k_cache: torch.Size([1, 8, 188, 128]) layer.1.v_cache: torch.Size([1, 8, 188, 128]) layer.2.k_cache: torch.Size([1, 8, 188, 128]) layer.2.v_cache: torch.Size([1, 8, 188, 128]) layer.3.k_cache: torch.Size([1, 8, 188, 128]) layer.3.v_cache: torch.Size([1, 8, 188, 128]) layer.4.k_cache: torch.Size([1, 8, 188, 128]) layer.4.v_cache: torch.Size([1, 8, 188, 128]) layer.4.output: torch.Size([1, 188, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02614099 24.71021287 layer.0.v_cache 0.00000027 0.00062453 layer.1.k_cache 0.00304419 3.60051127 layer.1.v_cache 0.00000081 0.00270277 layer.2.k_cache 0.00119225 1.54628526 layer.2.v_cache 0.00000108 0.00373312 layer.3.k_cache 0.00133471 1.85869128 layer.3.v_cache 0.00000207 0.00612330 layer.4.k_cache 0.00351311 3.71757994 layer.4.v_cache 0.00000296 0.01030271 layer.4.output 0.00017320 0.18226764 ------------------------------------------------------------------------------------- TOTAL 0.00256609 2.58470268 (elements=2,695,168) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2695168 Total Bytes 58432 BPFP 0.1734 bits/point EBPFP 0.3469 equivalent bits/point MSE 2.584703 ---------------------- -------------------------------------------------------- Time: 3.461s Load: 0.009s, Pack+Encode: 1.863s, Decode+Unpack: 1.588s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 188, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 188, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5847 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample57-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample57-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample59-layer4-item1.zst (64/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample59-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 197, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 197, 128) Output shape: (1, 197, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.0.v_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.1.k_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.1.v_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.2.k_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.2.v_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.3.k_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.3.v_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.4.k_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.4.v_cache: torch.Size([1, 8, 197, 128]) -> torch.Size([1, 1, 197, 1024]) layer.4.output: torch.Size([1, 197, 4096]) -> torch.Size([1, 1, 197, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,208B, BPFP=0.0479 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,600B, BPFP=0.2617 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,124B, BPFP=0.2429 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 7,752B, BPFP=0.3074 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,448B, BPFP=0.2954 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 8,844B, BPFP=0.3507 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,296B, BPFP=0.3290 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,976B, BPFP=0.3560 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,296B, BPFP=0.2497 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 9,392B, BPFP=0.3725 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,336B, BPFP=0.0430 ⌛️ [2/4] FRONTEND: Frontend time: 2.335s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 197, 128]) layer.0.v_cache: torch.Size([1, 8, 197, 128]) layer.1.k_cache: torch.Size([1, 8, 197, 128]) layer.1.v_cache: torch.Size([1, 8, 197, 128]) layer.2.k_cache: torch.Size([1, 8, 197, 128]) layer.2.v_cache: torch.Size([1, 8, 197, 128]) layer.3.k_cache: torch.Size([1, 8, 197, 128]) layer.3.v_cache: torch.Size([1, 8, 197, 128]) layer.4.k_cache: torch.Size([1, 8, 197, 128]) layer.4.v_cache: torch.Size([1, 8, 197, 128]) layer.4.output: torch.Size([1, 197, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.782s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 197, 128]) layer.0.v_cache: torch.Size([1, 8, 197, 128]) layer.1.k_cache: torch.Size([1, 8, 197, 128]) layer.1.v_cache: torch.Size([1, 8, 197, 128]) layer.2.k_cache: torch.Size([1, 8, 197, 128]) layer.2.v_cache: torch.Size([1, 8, 197, 128]) layer.3.k_cache: torch.Size([1, 8, 197, 128]) layer.3.v_cache: torch.Size([1, 8, 197, 128]) layer.4.k_cache: torch.Size([1, 8, 197, 128]) layer.4.v_cache: torch.Size([1, 8, 197, 128]) layer.4.output: torch.Size([1, 197, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02717010 29.08881514 layer.0.v_cache 0.00000028 0.00060349 layer.1.k_cache 0.00303619 3.21362584 layer.1.v_cache 0.00000078 0.00236217 layer.2.k_cache 0.00117018 1.48280110 layer.2.v_cache 0.00000106 0.00339867 layer.3.k_cache 0.00130579 1.68831453 layer.3.v_cache 0.00000202 0.00569547 layer.4.k_cache 0.00343555 3.30413509 layer.4.v_cache 0.00000289 0.00944918 layer.4.output 0.00018088 0.16281864 ------------------------------------------------------------------------------------- TOTAL 0.00263203 2.81789109 (elements=2,824,192) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2824192 Total Bytes 75272 BPFP 0.2132 bits/point EBPFP 0.4264 equivalent bits/point MSE 2.817891 ---------------------- -------------------------------------------------------- Time: 4.128s Load: 0.010s, Pack+Encode: 2.335s, Decode+Unpack: 1.782s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 197, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 197, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8179 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample59-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample59-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample6-layer4-item1.zst (65/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample6-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 204, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.012s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 204, 128) Output shape: (1, 204, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.0.v_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.1.k_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.1.v_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.2.k_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.2.v_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.3.k_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.3.v_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.4.k_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.4.v_cache: torch.Size([1, 8, 204, 128]) -> torch.Size([1, 1, 204, 1024]) layer.4.output: torch.Size([1, 204, 4096]) -> torch.Size([1, 1, 204, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,228B, BPFP=0.0470 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,064B, BPFP=0.2705 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,408B, BPFP=0.2454 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 9,000B, BPFP=0.3447 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,912B, BPFP=0.3030 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 10,368B, BPFP=0.3971 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,456B, BPFP=0.3238 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 10,276B, BPFP=0.3935 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,076B, BPFP=0.2327 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 10,820B, BPFP=0.4144 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,672B, BPFP=0.0447 ⌛️ [2/4] FRONTEND: Frontend time: 2.395s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 204, 128]) layer.0.v_cache: torch.Size([1, 8, 204, 128]) layer.1.k_cache: torch.Size([1, 8, 204, 128]) layer.1.v_cache: torch.Size([1, 8, 204, 128]) layer.2.k_cache: torch.Size([1, 8, 204, 128]) layer.2.v_cache: torch.Size([1, 8, 204, 128]) layer.3.k_cache: torch.Size([1, 8, 204, 128]) layer.3.v_cache: torch.Size([1, 8, 204, 128]) layer.4.k_cache: torch.Size([1, 8, 204, 128]) layer.4.v_cache: torch.Size([1, 8, 204, 128]) layer.4.output: torch.Size([1, 204, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.879s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 204, 128]) layer.0.v_cache: torch.Size([1, 8, 204, 128]) layer.1.k_cache: torch.Size([1, 8, 204, 128]) layer.1.v_cache: torch.Size([1, 8, 204, 128]) layer.2.k_cache: torch.Size([1, 8, 204, 128]) layer.2.v_cache: torch.Size([1, 8, 204, 128]) layer.3.k_cache: torch.Size([1, 8, 204, 128]) layer.3.v_cache: torch.Size([1, 8, 204, 128]) layer.4.k_cache: torch.Size([1, 8, 204, 128]) layer.4.v_cache: torch.Size([1, 8, 204, 128]) layer.4.output: torch.Size([1, 204, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02651415 30.95187797 layer.0.v_cache 0.00000026 0.00061940 layer.1.k_cache 0.00301626 3.10901956 layer.1.v_cache 0.00000082 0.00255537 layer.2.k_cache 0.00116162 1.45202622 layer.2.v_cache 0.00000113 0.00369477 layer.3.k_cache 0.00136563 1.68355650 layer.3.v_cache 0.00000207 0.00615065 layer.4.k_cache 0.00348750 3.48473762 layer.4.v_cache 0.00000309 0.01040355 layer.4.output 0.00017661 0.18801119 ------------------------------------------------------------------------------------- TOTAL 0.00258993 2.96119188 (elements=2,924,544) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2924544 Total Bytes 82280 BPFP 0.2251 bits/point EBPFP 0.4501 equivalent bits/point MSE 2.961192 ---------------------- -------------------------------------------------------- Time: 4.285s Load: 0.012s, Pack+Encode: 2.395s, Decode+Unpack: 1.879s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 204, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 204, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.9612 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample6-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample6-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample60-layer4-item1.zst (66/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample60-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 171, 128) Output shape: (1, 171, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.output: torch.Size([1, 171, 4096]) -> torch.Size([1, 1, 171, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,032B, BPFP=0.0471 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,908B, BPFP=0.2699 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,684B, BPFP=0.2597 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,668B, BPFP=0.3046 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,604B, BPFP=0.2560 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,544B, BPFP=0.3447 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,840B, BPFP=0.3125 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,420B, BPFP=0.3390 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,016B, BPFP=0.2292 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,108B, BPFP=0.3247 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,056B, BPFP=0.0463 ⌛️ [2/4] FRONTEND: Frontend time: 1.879s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.658s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02690465 26.96904126 layer.0.v_cache 0.00000028 0.00062160 layer.1.k_cache 0.00309750 3.33135879 layer.1.v_cache 0.00000081 0.00248945 layer.2.k_cache 0.00115835 1.44866801 layer.2.v_cache 0.00000109 0.00367407 layer.3.k_cache 0.00135207 1.74915514 layer.3.v_cache 0.00000214 0.00624336 layer.4.k_cache 0.00342229 3.45603256 layer.4.v_cache 0.00000302 0.01022203 layer.4.output 0.00017203 0.19598299 ------------------------------------------------------------------------------------- TOTAL 0.00261645 2.69724559 (elements=2,451,456) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2451456 Total Bytes 62880 BPFP 0.2052 bits/point EBPFP 0.4104 equivalent bits/point MSE 2.697246 ---------------------- -------------------------------------------------------- Time: 3.545s Load: 0.009s, Pack+Encode: 1.879s, Decode+Unpack: 1.658s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6972 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample60-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample60-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample62-layer4-item1.zst (67/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample62-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 180, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 180, 128) Output shape: (1, 180, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 180, 128]) -> torch.Size([1, 1, 180, 1024]) layer.0.v_cache: torch.Size([1, 8, 180, 128]) -> torch.Size([1, 1, 180, 1024]) layer.1.k_cache: torch.Size([1, 8, 180, 128]) -> torch.Size([1, 1, 180, 1024]) layer.1.v_cache: torch.Size([1, 8, 180, 128]) -> torch.Size([1, 1, 180, 1024]) layer.2.k_cache: torch.Size([1, 8, 180, 128]) -> torch.Size([1, 1, 180, 1024]) layer.2.v_cache: torch.Size([1, 8, 180, 128]) -> torch.Size([1, 1, 180, 1024]) layer.3.k_cache: torch.Size([1, 8, 180, 128]) -> torch.Size([1, 1, 180, 1024]) layer.3.v_cache: torch.Size([1, 8, 180, 128]) -> torch.Size([1, 1, 180, 1024]) layer.4.k_cache: torch.Size([1, 8, 180, 128]) -> torch.Size([1, 1, 180, 1024]) layer.4.v_cache: torch.Size([1, 8, 180, 128]) -> torch.Size([1, 1, 180, 1024]) layer.4.output: torch.Size([1, 180, 4096]) -> torch.Size([1, 1, 180, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,088B, BPFP=0.0472 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,016B, BPFP=0.2611 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,412B, BPFP=0.2349 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,404B, BPFP=0.2780 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,920B, BPFP=0.2569 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,792B, BPFP=0.3382 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,960B, BPFP=0.3021 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,056B, BPFP=0.3497 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,232B, BPFP=0.2271 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 8,036B, BPFP=0.3488 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,292B, BPFP=0.0466 ⌛️ [2/4] FRONTEND: Frontend time: 1.889s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 180, 128]) layer.0.v_cache: torch.Size([1, 8, 180, 128]) layer.1.k_cache: torch.Size([1, 8, 180, 128]) layer.1.v_cache: torch.Size([1, 8, 180, 128]) layer.2.k_cache: torch.Size([1, 8, 180, 128]) layer.2.v_cache: torch.Size([1, 8, 180, 128]) layer.3.k_cache: torch.Size([1, 8, 180, 128]) layer.3.v_cache: torch.Size([1, 8, 180, 128]) layer.4.k_cache: torch.Size([1, 8, 180, 128]) layer.4.v_cache: torch.Size([1, 8, 180, 128]) layer.4.output: torch.Size([1, 180, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.436s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 180, 128]) layer.0.v_cache: torch.Size([1, 8, 180, 128]) layer.1.k_cache: torch.Size([1, 8, 180, 128]) layer.1.v_cache: torch.Size([1, 8, 180, 128]) layer.2.k_cache: torch.Size([1, 8, 180, 128]) layer.2.v_cache: torch.Size([1, 8, 180, 128]) layer.3.k_cache: torch.Size([1, 8, 180, 128]) layer.3.v_cache: torch.Size([1, 8, 180, 128]) layer.4.k_cache: torch.Size([1, 8, 180, 128]) layer.4.v_cache: torch.Size([1, 8, 180, 128]) layer.4.output: torch.Size([1, 180, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02696468 24.80825738 layer.0.v_cache 0.00000027 0.00061771 layer.1.k_cache 0.00304294 3.48813443 layer.1.v_cache 0.00000080 0.00262908 layer.2.k_cache 0.00116858 1.56383752 layer.2.v_cache 0.00000111 0.00373654 layer.3.k_cache 0.00132793 1.82545217 layer.3.v_cache 0.00000207 0.00635151 layer.4.k_cache 0.00340096 3.76721530 layer.4.v_cache 0.00000322 0.01078406 layer.4.output 0.00018870 0.19380110 ------------------------------------------------------------------------------------- TOTAL 0.00261910 2.58944429 (elements=2,580,480) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2580480 Total Bytes 65208 BPFP 0.2022 bits/point EBPFP 0.4043 equivalent bits/point MSE 2.589444 ---------------------- -------------------------------------------------------- Time: 3.334s Load: 0.009s, Pack+Encode: 1.889s, Decode+Unpack: 1.436s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 180, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 180, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5894 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample62-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample62-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample63-layer4-item1.zst (68/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample63-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 165, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 165, 128) Output shape: (1, 165, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.0.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.1.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.1.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.2.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.2.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.3.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.3.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.4.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.4.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.4.output: torch.Size([1, 165, 4096]) -> torch.Size([1, 1, 165, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,024B, BPFP=0.0485 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,604B, BPFP=0.2653 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,300B, BPFP=0.2509 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,772B, BPFP=0.2733 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,704B, BPFP=0.2701 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,768B, BPFP=0.3205 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,760B, BPFP=0.3201 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,208B, BPFP=0.3413 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,056B, BPFP=0.2394 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,996B, BPFP=0.3312 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,252B, BPFP=0.0503 ⌛️ [2/4] FRONTEND: Frontend time: 1.974s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 165, 128]) layer.0.v_cache: torch.Size([1, 8, 165, 128]) layer.1.k_cache: torch.Size([1, 8, 165, 128]) layer.1.v_cache: torch.Size([1, 8, 165, 128]) layer.2.k_cache: torch.Size([1, 8, 165, 128]) layer.2.v_cache: torch.Size([1, 8, 165, 128]) layer.3.k_cache: torch.Size([1, 8, 165, 128]) layer.3.v_cache: torch.Size([1, 8, 165, 128]) layer.4.k_cache: torch.Size([1, 8, 165, 128]) layer.4.v_cache: torch.Size([1, 8, 165, 128]) layer.4.output: torch.Size([1, 165, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.519s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 165, 128]) layer.0.v_cache: torch.Size([1, 8, 165, 128]) layer.1.k_cache: torch.Size([1, 8, 165, 128]) layer.1.v_cache: torch.Size([1, 8, 165, 128]) layer.2.k_cache: torch.Size([1, 8, 165, 128]) layer.2.v_cache: torch.Size([1, 8, 165, 128]) layer.3.k_cache: torch.Size([1, 8, 165, 128]) layer.3.v_cache: torch.Size([1, 8, 165, 128]) layer.4.k_cache: torch.Size([1, 8, 165, 128]) layer.4.v_cache: torch.Size([1, 8, 165, 128]) layer.4.output: torch.Size([1, 165, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02707518 24.85131096 layer.0.v_cache 0.00000026 0.00061931 layer.1.k_cache 0.00315113 3.14255667 layer.1.v_cache 0.00000089 0.00268484 layer.2.k_cache 0.00116140 1.48100124 layer.2.v_cache 0.00000111 0.00380966 layer.3.k_cache 0.00133385 1.70946230 layer.3.v_cache 0.00000208 0.00620468 layer.4.k_cache 0.00353406 3.69344113 layer.4.v_cache 0.00000307 0.01048436 layer.4.output 0.00018338 0.18627231 ------------------------------------------------------------------------------------- TOTAL 0.00264261 2.54619031 (elements=2,365,440) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2365440 Total Bytes 60444 BPFP 0.2044 bits/point EBPFP 0.4088 equivalent bits/point MSE 2.546190 ---------------------- -------------------------------------------------------- Time: 3.502s Load: 0.009s, Pack+Encode: 1.974s, Decode+Unpack: 1.519s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 165, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5462 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample63-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample63-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample64-layer4-item1.zst (69/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample64-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 171, 128) Output shape: (1, 171, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.output: torch.Size([1, 171, 4096]) -> torch.Size([1, 1, 171, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,044B, BPFP=0.0477 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,624B, BPFP=0.2569 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,948B, BPFP=0.2717 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,756B, BPFP=0.3087 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,596B, BPFP=0.2557 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,316B, BPFP=0.3342 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,788B, BPFP=0.3101 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,628B, BPFP=0.3485 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,228B, BPFP=0.2389 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,204B, BPFP=0.3291 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,864B, BPFP=0.0441 ⌛️ [2/4] FRONTEND: Frontend time: 1.882s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.586s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02638843 28.24664771 layer.0.v_cache 0.00000026 0.00060827 layer.1.k_cache 0.00303726 3.50572588 layer.1.v_cache 0.00000078 0.00256006 layer.2.k_cache 0.00118268 1.43096103 layer.2.v_cache 0.00000107 0.00366163 layer.3.k_cache 0.00135949 1.76942560 layer.3.v_cache 0.00000203 0.00601150 layer.4.k_cache 0.00348397 3.48931920 layer.4.v_cache 0.00000306 0.01013259 layer.4.output 0.00019589 0.19432951 ------------------------------------------------------------------------------------- TOTAL 0.00258876 2.80302654 (elements=2,451,456) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2451456 Total Bytes 62996 BPFP 0.2056 bits/point EBPFP 0.4112 equivalent bits/point MSE 2.803027 ---------------------- -------------------------------------------------------- Time: 3.479s Load: 0.010s, Pack+Encode: 1.882s, Decode+Unpack: 1.586s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8030 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample64-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample64-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample65-layer4-item1.zst (70/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample65-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 166, 128) Output shape: (1, 166, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.0.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.1.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.1.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.2.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.2.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.3.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.3.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.4.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.4.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.4.output: torch.Size([1, 166, 4096]) -> torch.Size([1, 1, 166, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 992B, BPFP=0.0467 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,916B, BPFP=0.2784 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,128B, BPFP=0.2413 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,584B, BPFP=0.3099 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,668B, BPFP=0.2668 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,276B, BPFP=0.3424 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,384B, BPFP=0.3475 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,296B, BPFP=0.3434 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,348B, BPFP=0.2517 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,476B, BPFP=0.3518 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,572B, BPFP=0.0420 ⌛️ [2/4] FRONTEND: Frontend time: 1.928s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 166, 128]) layer.0.v_cache: torch.Size([1, 8, 166, 128]) layer.1.k_cache: torch.Size([1, 8, 166, 128]) layer.1.v_cache: torch.Size([1, 8, 166, 128]) layer.2.k_cache: torch.Size([1, 8, 166, 128]) layer.2.v_cache: torch.Size([1, 8, 166, 128]) layer.3.k_cache: torch.Size([1, 8, 166, 128]) layer.3.v_cache: torch.Size([1, 8, 166, 128]) layer.4.k_cache: torch.Size([1, 8, 166, 128]) layer.4.v_cache: torch.Size([1, 8, 166, 128]) layer.4.output: torch.Size([1, 166, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.418s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 166, 128]) layer.0.v_cache: torch.Size([1, 8, 166, 128]) layer.1.k_cache: torch.Size([1, 8, 166, 128]) layer.1.v_cache: torch.Size([1, 8, 166, 128]) layer.2.k_cache: torch.Size([1, 8, 166, 128]) layer.2.v_cache: torch.Size([1, 8, 166, 128]) layer.3.k_cache: torch.Size([1, 8, 166, 128]) layer.3.v_cache: torch.Size([1, 8, 166, 128]) layer.4.k_cache: torch.Size([1, 8, 166, 128]) layer.4.v_cache: torch.Size([1, 8, 166, 128]) layer.4.output: torch.Size([1, 166, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02749621 24.60948972 layer.0.v_cache 0.00000027 0.00060606 layer.1.k_cache 0.00299443 3.10361637 layer.1.v_cache 0.00000078 0.00254251 layer.2.k_cache 0.00116568 1.43979829 layer.2.v_cache 0.00000116 0.00376537 layer.3.k_cache 0.00133954 1.76899039 layer.3.v_cache 0.00000205 0.00608859 layer.4.k_cache 0.00345084 3.51365092 layer.4.v_cache 0.00000307 0.01032832 layer.4.output 0.00017739 0.17693502 ------------------------------------------------------------------------------------- TOTAL 0.00265454 2.51190119 (elements=2,379,776) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2379776 Total Bytes 62640 BPFP 0.2106 bits/point EBPFP 0.4211 equivalent bits/point MSE 2.511901 ---------------------- -------------------------------------------------------- Time: 3.355s Load: 0.009s, Pack+Encode: 1.928s, Decode+Unpack: 1.418s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5119 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample65-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample65-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample66-layer4-item1.zst (71/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample66-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 164, 128) Output shape: (1, 164, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.output: torch.Size([1, 164, 4096]) -> torch.Size([1, 1, 164, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,004B, BPFP=0.0478 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,460B, BPFP=0.3077 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,696B, BPFP=0.2713 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,204B, BPFP=0.2955 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,796B, BPFP=0.2761 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,004B, BPFP=0.3337 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,068B, BPFP=0.3367 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,836B, BPFP=0.3256 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,172B, BPFP=0.2464 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,196B, BPFP=0.3428 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,744B, BPFP=0.0446 ⌛️ [2/4] FRONTEND: Frontend time: 1.919s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) layer.4.output: torch.Size([1, 164, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.558s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) layer.4.output: torch.Size([1, 164, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02831957 26.53317931 layer.0.v_cache 0.00000026 0.00061109 layer.1.k_cache 0.00319086 3.59117685 layer.1.v_cache 0.00000080 0.00257727 layer.2.k_cache 0.00118664 1.44639420 layer.2.v_cache 0.00000113 0.00370395 layer.3.k_cache 0.00134609 1.76054736 layer.3.v_cache 0.00000208 0.00624202 layer.4.k_cache 0.00339094 3.48194253 layer.4.v_cache 0.00000313 0.01074567 layer.4.output 0.00016991 0.19054241 ------------------------------------------------------------------------------------- TOTAL 0.00272294 2.68566356 (elements=2,351,104) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2351104 Total Bytes 62180 BPFP 0.2116 bits/point EBPFP 0.4232 equivalent bits/point MSE 2.685664 ---------------------- -------------------------------------------------------- Time: 3.487s Load: 0.010s, Pack+Encode: 1.919s, Decode+Unpack: 1.558s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6857 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample66-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample66-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample67-layer4-item1.zst (72/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample67-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 169, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 169, 128) Output shape: (1, 169, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.output: torch.Size([1, 169, 4096]) -> torch.Size([1, 1, 169, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,036B, BPFP=0.0479 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,960B, BPFP=0.2755 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,568B, BPFP=0.2574 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,420B, BPFP=0.2968 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,648B, BPFP=0.2611 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,360B, BPFP=0.3402 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,844B, BPFP=0.3164 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,508B, BPFP=0.3471 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,284B, BPFP=0.2443 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,356B, BPFP=0.3401 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,476B, BPFP=0.0402 ⌛️ [2/4] FRONTEND: Frontend time: 1.885s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) layer.4.output: torch.Size([1, 169, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.504s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) layer.4.output: torch.Size([1, 169, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02632406 24.66179156 layer.0.v_cache 0.00000026 0.00059323 layer.1.k_cache 0.00304039 3.29629174 layer.1.v_cache 0.00000079 0.00257647 layer.2.k_cache 0.00117699 1.44200820 layer.2.v_cache 0.00000106 0.00355328 layer.3.k_cache 0.00132669 1.71558303 layer.3.v_cache 0.00000197 0.00576549 layer.4.k_cache 0.00349736 3.40975573 layer.4.v_cache 0.00000306 0.01009920 layer.4.output 0.00016247 0.18020485 ------------------------------------------------------------------------------------- TOTAL 0.00257304 2.51920267 (elements=2,422,784) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2422784 Total Bytes 62460 BPFP 0.2062 bits/point EBPFP 0.4125 equivalent bits/point MSE 2.519203 ---------------------- -------------------------------------------------------- Time: 3.398s Load: 0.009s, Pack+Encode: 1.885s, Decode+Unpack: 1.504s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 169, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5192 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample67-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample67-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample68-layer4-item1.zst (73/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample68-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 167, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 167, 128) Output shape: (1, 167, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.0.v_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.1.k_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.1.v_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.2.k_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.2.v_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.3.k_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.3.v_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.4.k_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.4.v_cache: torch.Size([1, 8, 167, 128]) -> torch.Size([1, 1, 167, 1024]) layer.4.output: torch.Size([1, 167, 4096]) -> torch.Size([1, 1, 167, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,028B, BPFP=0.0481 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,736B, BPFP=0.2683 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,328B, BPFP=0.2493 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,872B, BPFP=0.2747 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,532B, BPFP=0.2588 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,624B, BPFP=0.3099 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,832B, BPFP=0.3196 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,312B, BPFP=0.3421 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,332B, BPFP=0.2494 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,508B, BPFP=0.3512 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,172B, BPFP=0.0488 ⌛️ [2/4] FRONTEND: Frontend time: 1.959s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 167, 128]) layer.0.v_cache: torch.Size([1, 8, 167, 128]) layer.1.k_cache: torch.Size([1, 8, 167, 128]) layer.1.v_cache: torch.Size([1, 8, 167, 128]) layer.2.k_cache: torch.Size([1, 8, 167, 128]) layer.2.v_cache: torch.Size([1, 8, 167, 128]) layer.3.k_cache: torch.Size([1, 8, 167, 128]) layer.3.v_cache: torch.Size([1, 8, 167, 128]) layer.4.k_cache: torch.Size([1, 8, 167, 128]) layer.4.v_cache: torch.Size([1, 8, 167, 128]) layer.4.output: torch.Size([1, 167, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.525s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 167, 128]) layer.0.v_cache: torch.Size([1, 8, 167, 128]) layer.1.k_cache: torch.Size([1, 8, 167, 128]) layer.1.v_cache: torch.Size([1, 8, 167, 128]) layer.2.k_cache: torch.Size([1, 8, 167, 128]) layer.2.v_cache: torch.Size([1, 8, 167, 128]) layer.3.k_cache: torch.Size([1, 8, 167, 128]) layer.3.v_cache: torch.Size([1, 8, 167, 128]) layer.4.k_cache: torch.Size([1, 8, 167, 128]) layer.4.v_cache: torch.Size([1, 8, 167, 128]) layer.4.output: torch.Size([1, 167, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02744324 25.84483182 layer.0.v_cache 0.00000027 0.00062521 layer.1.k_cache 0.00306544 3.28273750 layer.1.v_cache 0.00000083 0.00265091 layer.2.k_cache 0.00116567 1.43972404 layer.2.v_cache 0.00000114 0.00383157 layer.3.k_cache 0.00131614 1.72362824 layer.3.v_cache 0.00000212 0.00636900 layer.4.k_cache 0.00345897 3.37681717 layer.4.v_cache 0.00000319 0.01067060 layer.4.output 0.00018850 0.19981444 ------------------------------------------------------------------------------------- TOTAL 0.00265793 2.60651027 (elements=2,394,112) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2394112 Total Bytes 61276 BPFP 0.2048 bits/point EBPFP 0.4095 equivalent bits/point MSE 2.606510 ---------------------- -------------------------------------------------------- Time: 3.495s Load: 0.011s, Pack+Encode: 1.959s, Decode+Unpack: 1.525s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 167, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 167, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6065 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample68-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample68-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample69-layer4-item1.zst (74/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample69-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 189, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 189, 128) Output shape: (1, 189, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 189, 128]) -> torch.Size([1, 1, 189, 1024]) layer.0.v_cache: torch.Size([1, 8, 189, 128]) -> torch.Size([1, 1, 189, 1024]) layer.1.k_cache: torch.Size([1, 8, 189, 128]) -> torch.Size([1, 1, 189, 1024]) layer.1.v_cache: torch.Size([1, 8, 189, 128]) -> torch.Size([1, 1, 189, 1024]) layer.2.k_cache: torch.Size([1, 8, 189, 128]) -> torch.Size([1, 1, 189, 1024]) layer.2.v_cache: torch.Size([1, 8, 189, 128]) -> torch.Size([1, 1, 189, 1024]) layer.3.k_cache: torch.Size([1, 8, 189, 128]) -> torch.Size([1, 1, 189, 1024]) layer.3.v_cache: torch.Size([1, 8, 189, 128]) -> torch.Size([1, 1, 189, 1024]) layer.4.k_cache: torch.Size([1, 8, 189, 128]) -> torch.Size([1, 1, 189, 1024]) layer.4.v_cache: torch.Size([1, 8, 189, 128]) -> torch.Size([1, 1, 189, 1024]) layer.4.output: torch.Size([1, 189, 4096]) -> torch.Size([1, 1, 189, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,136B, BPFP=0.0470 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,848B, BPFP=0.2417 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,284B, BPFP=0.1771 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,368B, BPFP=0.2632 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,168B, BPFP=0.2136 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,008B, BPFP=0.2897 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,088B, BPFP=0.2517 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,980B, BPFP=0.2885 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,296B, BPFP=0.1776 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,004B, BPFP=0.2895 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 6,760B, BPFP=0.0699 ⌛️ [2/4] FRONTEND: Frontend time: 1.864s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 189, 128]) layer.0.v_cache: torch.Size([1, 8, 189, 128]) layer.1.k_cache: torch.Size([1, 8, 189, 128]) layer.1.v_cache: torch.Size([1, 8, 189, 128]) layer.2.k_cache: torch.Size([1, 8, 189, 128]) layer.2.v_cache: torch.Size([1, 8, 189, 128]) layer.3.k_cache: torch.Size([1, 8, 189, 128]) layer.3.v_cache: torch.Size([1, 8, 189, 128]) layer.4.k_cache: torch.Size([1, 8, 189, 128]) layer.4.v_cache: torch.Size([1, 8, 189, 128]) layer.4.output: torch.Size([1, 189, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.673s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 189, 128]) layer.0.v_cache: torch.Size([1, 8, 189, 128]) layer.1.k_cache: torch.Size([1, 8, 189, 128]) layer.1.v_cache: torch.Size([1, 8, 189, 128]) layer.2.k_cache: torch.Size([1, 8, 189, 128]) layer.2.v_cache: torch.Size([1, 8, 189, 128]) layer.3.k_cache: torch.Size([1, 8, 189, 128]) layer.3.v_cache: torch.Size([1, 8, 189, 128]) layer.4.k_cache: torch.Size([1, 8, 189, 128]) layer.4.v_cache: torch.Size([1, 8, 189, 128]) layer.4.output: torch.Size([1, 189, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02668087 28.09286128 layer.0.v_cache 0.00000027 0.00065353 layer.1.k_cache 0.00297688 3.28484591 layer.1.v_cache 0.00000086 0.00277711 layer.2.k_cache 0.00116960 1.60575035 layer.2.v_cache 0.00000119 0.00389518 layer.3.k_cache 0.00130480 1.87690210 layer.3.v_cache 0.00000220 0.00670479 layer.4.k_cache 0.00357267 3.45088188 layer.4.v_cache 0.00000319 0.01115761 layer.4.output 0.00018018 0.21633055 ------------------------------------------------------------------------------------- TOTAL 0.00260238 2.80012514 (elements=2,709,504) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2709504 Total Bytes 60940 BPFP 0.1799 bits/point EBPFP 0.3599 equivalent bits/point MSE 2.800125 ---------------------- -------------------------------------------------------- Time: 3.548s Load: 0.010s, Pack+Encode: 1.864s, Decode+Unpack: 1.673s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 189, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 189, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8001 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample69-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample69-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample70-layer4-item1.zst (75/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample70-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 173, 128) Output shape: (1, 173, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.output: torch.Size([1, 173, 4096]) -> torch.Size([1, 1, 173, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,028B, BPFP=0.0464 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,776B, BPFP=0.2608 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,668B, BPFP=0.2560 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,252B, BPFP=0.2823 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,640B, BPFP=0.2547 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,000B, BPFP=0.3161 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,740B, BPFP=0.3044 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,332B, BPFP=0.3311 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,220B, BPFP=0.2357 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,296B, BPFP=0.3295 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,272B, BPFP=0.0595 ⌛️ [2/4] FRONTEND: Frontend time: 1.922s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) layer.4.output: torch.Size([1, 173, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.430s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) layer.4.output: torch.Size([1, 173, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02663673 26.78000813 layer.0.v_cache 0.00000027 0.00061992 layer.1.k_cache 0.00311122 3.40321059 layer.1.v_cache 0.00000080 0.00254329 layer.2.k_cache 0.00117338 1.46580073 layer.2.v_cache 0.00000114 0.00367275 layer.3.k_cache 0.00131734 1.76776688 layer.3.v_cache 0.00000215 0.00617995 layer.4.k_cache 0.00350522 3.53735563 layer.4.v_cache 0.00000298 0.01008051 layer.4.output 0.00018231 0.21395321 ------------------------------------------------------------------------------------- TOTAL 0.00260575 2.70236080 (elements=2,480,128) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2480128 Total Bytes 63224 BPFP 0.2039 bits/point EBPFP 0.4079 equivalent bits/point MSE 2.702361 ---------------------- -------------------------------------------------------- Time: 3.362s Load: 0.010s, Pack+Encode: 1.922s, Decode+Unpack: 1.430s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7024 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample70-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample70-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample71-layer4-item1.zst (76/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample71-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 166, 128) Output shape: (1, 166, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.0.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.1.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.1.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.2.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.2.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.3.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.3.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.4.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.4.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.4.output: torch.Size([1, 166, 4096]) -> torch.Size([1, 1, 166, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 988B, BPFP=0.0465 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,752B, BPFP=0.2707 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,088B, BPFP=0.2395 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,016B, BPFP=0.2831 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,600B, BPFP=0.2636 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,020B, BPFP=0.3304 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,088B, BPFP=0.3336 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,904B, BPFP=0.3249 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,300B, BPFP=0.2494 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,824B, BPFP=0.3212 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,748B, BPFP=0.0441 ⌛️ [2/4] FRONTEND: Frontend time: 1.936s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 166, 128]) layer.0.v_cache: torch.Size([1, 8, 166, 128]) layer.1.k_cache: torch.Size([1, 8, 166, 128]) layer.1.v_cache: torch.Size([1, 8, 166, 128]) layer.2.k_cache: torch.Size([1, 8, 166, 128]) layer.2.v_cache: torch.Size([1, 8, 166, 128]) layer.3.k_cache: torch.Size([1, 8, 166, 128]) layer.3.v_cache: torch.Size([1, 8, 166, 128]) layer.4.k_cache: torch.Size([1, 8, 166, 128]) layer.4.v_cache: torch.Size([1, 8, 166, 128]) layer.4.output: torch.Size([1, 166, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.590s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 166, 128]) layer.0.v_cache: torch.Size([1, 8, 166, 128]) layer.1.k_cache: torch.Size([1, 8, 166, 128]) layer.1.v_cache: torch.Size([1, 8, 166, 128]) layer.2.k_cache: torch.Size([1, 8, 166, 128]) layer.2.v_cache: torch.Size([1, 8, 166, 128]) layer.3.k_cache: torch.Size([1, 8, 166, 128]) layer.3.v_cache: torch.Size([1, 8, 166, 128]) layer.4.k_cache: torch.Size([1, 8, 166, 128]) layer.4.v_cache: torch.Size([1, 8, 166, 128]) layer.4.output: torch.Size([1, 166, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02747041 26.34529132 layer.0.v_cache 0.00000026 0.00062173 layer.1.k_cache 0.00305587 3.44977616 layer.1.v_cache 0.00000080 0.00255244 layer.2.k_cache 0.00117233 1.45966587 layer.2.v_cache 0.00000108 0.00373838 layer.3.k_cache 0.00134723 1.74764279 layer.3.v_cache 0.00000213 0.00633347 layer.4.k_cache 0.00344432 3.49655115 layer.4.v_cache 0.00000302 0.01038059 layer.4.output 0.00018389 0.19086169 ------------------------------------------------------------------------------------- TOTAL 0.00265950 2.66328576 (elements=2,379,776) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2379776 Total Bytes 60328 BPFP 0.2028 bits/point EBPFP 0.4056 equivalent bits/point MSE 2.663286 ---------------------- -------------------------------------------------------- Time: 3.535s Load: 0.010s, Pack+Encode: 1.936s, Decode+Unpack: 1.590s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6633 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample71-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample71-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample72-layer4-item1.zst (77/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample72-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 171, 128) Output shape: (1, 171, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.output: torch.Size([1, 171, 4096]) -> torch.Size([1, 1, 171, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,028B, BPFP=0.0470 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,084B, BPFP=0.2780 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,604B, BPFP=0.2560 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,900B, BPFP=0.2696 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,652B, BPFP=0.2582 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,760B, BPFP=0.3088 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,748B, BPFP=0.3083 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,076B, BPFP=0.3233 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,216B, BPFP=0.2383 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,156B, BPFP=0.3269 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,792B, BPFP=0.0433 ⌛️ [2/4] FRONTEND: Frontend time: 1.882s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.532s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02713734 25.80228321 layer.0.v_cache 0.00000027 0.00063017 layer.1.k_cache 0.00309222 3.41391009 layer.1.v_cache 0.00000080 0.00254983 layer.2.k_cache 0.00115323 1.46303679 layer.2.v_cache 0.00000110 0.00374327 layer.3.k_cache 0.00132483 1.74918227 layer.3.v_cache 0.00000201 0.00595340 layer.4.k_cache 0.00341669 3.39321632 layer.4.v_cache 0.00000301 0.01013021 layer.4.output 0.00019064 0.21288795 ------------------------------------------------------------------------------------- TOTAL 0.00263529 2.62115624 (elements=2,451,456) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2451456 Total Bytes 61016 BPFP 0.1991 bits/point EBPFP 0.3982 equivalent bits/point MSE 2.621156 ---------------------- -------------------------------------------------------- Time: 3.424s Load: 0.010s, Pack+Encode: 1.882s, Decode+Unpack: 1.532s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6212 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample72-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample72-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample73-layer4-item1.zst (78/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample73-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 168, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 168, 128) Output shape: (1, 168, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 168, 128]) -> torch.Size([1, 1, 168, 1024]) layer.0.v_cache: torch.Size([1, 8, 168, 128]) -> torch.Size([1, 1, 168, 1024]) layer.1.k_cache: torch.Size([1, 8, 168, 128]) -> torch.Size([1, 1, 168, 1024]) layer.1.v_cache: torch.Size([1, 8, 168, 128]) -> torch.Size([1, 1, 168, 1024]) layer.2.k_cache: torch.Size([1, 8, 168, 128]) -> torch.Size([1, 1, 168, 1024]) layer.2.v_cache: torch.Size([1, 8, 168, 128]) -> torch.Size([1, 1, 168, 1024]) layer.3.k_cache: torch.Size([1, 8, 168, 128]) -> torch.Size([1, 1, 168, 1024]) layer.3.v_cache: torch.Size([1, 8, 168, 128]) -> torch.Size([1, 1, 168, 1024]) layer.4.k_cache: torch.Size([1, 8, 168, 128]) -> torch.Size([1, 1, 168, 1024]) layer.4.v_cache: torch.Size([1, 8, 168, 128]) -> torch.Size([1, 1, 168, 1024]) layer.4.output: torch.Size([1, 168, 4096]) -> torch.Size([1, 1, 168, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,024B, BPFP=0.0476 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,036B, BPFP=0.2807 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,548B, BPFP=0.2580 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,272B, BPFP=0.2917 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,728B, BPFP=0.2664 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,936B, BPFP=0.3690 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,156B, BPFP=0.3328 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 8,584B, BPFP=0.3992 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,552B, BPFP=0.2582 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 8,028B, BPFP=0.3733 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,776B, BPFP=0.0439 ⌛️ [2/4] FRONTEND: Frontend time: 1.942s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 168, 128]) layer.0.v_cache: torch.Size([1, 8, 168, 128]) layer.1.k_cache: torch.Size([1, 8, 168, 128]) layer.1.v_cache: torch.Size([1, 8, 168, 128]) layer.2.k_cache: torch.Size([1, 8, 168, 128]) layer.2.v_cache: torch.Size([1, 8, 168, 128]) layer.3.k_cache: torch.Size([1, 8, 168, 128]) layer.3.v_cache: torch.Size([1, 8, 168, 128]) layer.4.k_cache: torch.Size([1, 8, 168, 128]) layer.4.v_cache: torch.Size([1, 8, 168, 128]) layer.4.output: torch.Size([1, 168, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.417s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 168, 128]) layer.0.v_cache: torch.Size([1, 8, 168, 128]) layer.1.k_cache: torch.Size([1, 8, 168, 128]) layer.1.v_cache: torch.Size([1, 8, 168, 128]) layer.2.k_cache: torch.Size([1, 8, 168, 128]) layer.2.v_cache: torch.Size([1, 8, 168, 128]) layer.3.k_cache: torch.Size([1, 8, 168, 128]) layer.3.v_cache: torch.Size([1, 8, 168, 128]) layer.4.k_cache: torch.Size([1, 8, 168, 128]) layer.4.v_cache: torch.Size([1, 8, 168, 128]) layer.4.output: torch.Size([1, 168, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02628393 27.45021275 layer.0.v_cache 0.00000027 0.00060714 layer.1.k_cache 0.00308696 3.53128088 layer.1.v_cache 0.00000088 0.00252632 layer.2.k_cache 0.00116165 1.44120425 layer.2.v_cache 0.00000115 0.00374199 layer.3.k_cache 0.00130884 1.73856136 layer.3.v_cache 0.00000264 0.00657544 layer.4.k_cache 0.00339222 3.37623887 layer.4.v_cache 0.00000308 0.01068404 layer.4.output 0.00017467 0.17950057 ------------------------------------------------------------------------------------- TOTAL 0.00256716 2.73425967 (elements=2,408,448) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2408448 Total Bytes 65640 BPFP 0.2180 bits/point EBPFP 0.4361 equivalent bits/point MSE 2.734260 ---------------------- -------------------------------------------------------- Time: 3.368s Load: 0.010s, Pack+Encode: 1.942s, Decode+Unpack: 1.417s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 168, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 168, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7343 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample73-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample73-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample76-layer4-item1.zst (79/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample76-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 164, 128) Output shape: (1, 164, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) -> torch.Size([1, 1, 164, 1024]) layer.4.output: torch.Size([1, 164, 4096]) -> torch.Size([1, 1, 164, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,012B, BPFP=0.0482 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,644B, BPFP=0.2689 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,488B, BPFP=0.2614 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,516B, BPFP=0.3104 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,800B, BPFP=0.2763 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,644B, BPFP=0.3641 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,396B, BPFP=0.3523 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,528B, BPFP=0.3586 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,264B, BPFP=0.2508 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,828B, BPFP=0.3729 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,836B, BPFP=0.0457 ⌛️ [2/4] FRONTEND: Frontend time: 1.878s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) layer.4.output: torch.Size([1, 164, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.678s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 164, 128]) layer.0.v_cache: torch.Size([1, 8, 164, 128]) layer.1.k_cache: torch.Size([1, 8, 164, 128]) layer.1.v_cache: torch.Size([1, 8, 164, 128]) layer.2.k_cache: torch.Size([1, 8, 164, 128]) layer.2.v_cache: torch.Size([1, 8, 164, 128]) layer.3.k_cache: torch.Size([1, 8, 164, 128]) layer.3.v_cache: torch.Size([1, 8, 164, 128]) layer.4.k_cache: torch.Size([1, 8, 164, 128]) layer.4.v_cache: torch.Size([1, 8, 164, 128]) layer.4.output: torch.Size([1, 164, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02737338 25.47896520 layer.0.v_cache 0.00000027 0.00061423 layer.1.k_cache 0.00299196 3.37512505 layer.1.v_cache 0.00000082 0.00255872 layer.2.k_cache 0.00116550 1.47287164 layer.2.v_cache 0.00000112 0.00371869 layer.3.k_cache 0.00132297 1.78337897 layer.3.v_cache 0.00000204 0.00604299 layer.4.k_cache 0.00347967 3.40697163 layer.4.v_cache 0.00000307 0.01045159 layer.4.output 0.00016775 0.18125169 ------------------------------------------------------------------------------------- TOTAL 0.00264370 2.59040753 (elements=2,351,104) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2351104 Total Bytes 63956 BPFP 0.2176 bits/point EBPFP 0.4352 equivalent bits/point MSE 2.590408 ---------------------- -------------------------------------------------------- Time: 3.564s Load: 0.009s, Pack+Encode: 1.878s, Decode+Unpack: 1.678s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 164, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 164, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5904 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample76-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample76-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample77-layer4-item1.zst (80/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample77-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 175, 128) Output shape: (1, 175, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.output: torch.Size([1, 175, 4096]) -> torch.Size([1, 1, 175, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,016B, BPFP=0.0454 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,644B, BPFP=0.2520 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,384B, BPFP=0.2404 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,720B, BPFP=0.2554 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,640B, BPFP=0.2518 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,784B, BPFP=0.3029 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,804B, BPFP=0.3038 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,148B, BPFP=0.3191 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,824B, BPFP=0.2154 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,060B, BPFP=0.3152 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,712B, BPFP=0.0526 ⌛️ [2/4] FRONTEND: Frontend time: 1.901s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) layer.4.output: torch.Size([1, 175, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.440s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) layer.4.output: torch.Size([1, 175, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02704878 28.09571987 layer.0.v_cache 0.00000028 0.00062272 layer.1.k_cache 0.00295012 3.33343680 layer.1.v_cache 0.00000089 0.00259440 layer.2.k_cache 0.00115680 1.48914028 layer.2.v_cache 0.00000112 0.00378982 layer.3.k_cache 0.00133986 1.79226702 layer.3.v_cache 0.00000218 0.00636765 layer.4.k_cache 0.00335044 3.54499337 layer.4.v_cache 0.00000310 0.01059045 layer.4.output 0.00017962 0.22165682 ------------------------------------------------------------------------------------- TOTAL 0.00261229 2.79758212 (elements=2,508,800) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2508800 Total Bytes 60736 BPFP 0.1937 bits/point EBPFP 0.3873 equivalent bits/point MSE 2.797582 ---------------------- -------------------------------------------------------- Time: 3.352s Load: 0.011s, Pack+Encode: 1.901s, Decode+Unpack: 1.440s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7976 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample77-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample77-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample78-layer4-item1.zst (81/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample78-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 166, 128) Output shape: (1, 166, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.0.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.1.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.1.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.2.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.2.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.3.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.3.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.4.k_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.4.v_cache: torch.Size([1, 8, 166, 128]) -> torch.Size([1, 1, 166, 1024]) layer.4.output: torch.Size([1, 166, 4096]) -> torch.Size([1, 1, 166, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 992B, BPFP=0.0467 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,608B, BPFP=0.2639 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,468B, BPFP=0.2573 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,004B, BPFP=0.2826 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,576B, BPFP=0.2624 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,984B, BPFP=0.3287 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,996B, BPFP=0.3293 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,284B, BPFP=0.3428 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,132B, BPFP=0.2415 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,976B, BPFP=0.3283 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,616B, BPFP=0.0425 ⌛️ [2/4] FRONTEND: Frontend time: 1.947s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 166, 128]) layer.0.v_cache: torch.Size([1, 8, 166, 128]) layer.1.k_cache: torch.Size([1, 8, 166, 128]) layer.1.v_cache: torch.Size([1, 8, 166, 128]) layer.2.k_cache: torch.Size([1, 8, 166, 128]) layer.2.v_cache: torch.Size([1, 8, 166, 128]) layer.3.k_cache: torch.Size([1, 8, 166, 128]) layer.3.v_cache: torch.Size([1, 8, 166, 128]) layer.4.k_cache: torch.Size([1, 8, 166, 128]) layer.4.v_cache: torch.Size([1, 8, 166, 128]) layer.4.output: torch.Size([1, 166, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.528s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 166, 128]) layer.0.v_cache: torch.Size([1, 8, 166, 128]) layer.1.k_cache: torch.Size([1, 8, 166, 128]) layer.1.v_cache: torch.Size([1, 8, 166, 128]) layer.2.k_cache: torch.Size([1, 8, 166, 128]) layer.2.v_cache: torch.Size([1, 8, 166, 128]) layer.3.k_cache: torch.Size([1, 8, 166, 128]) layer.3.v_cache: torch.Size([1, 8, 166, 128]) layer.4.k_cache: torch.Size([1, 8, 166, 128]) layer.4.v_cache: torch.Size([1, 8, 166, 128]) layer.4.output: torch.Size([1, 166, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02721562 24.68704702 layer.0.v_cache 0.00000027 0.00061549 layer.1.k_cache 0.00308722 3.33599633 layer.1.v_cache 0.00000079 0.00253355 layer.2.k_cache 0.00116669 1.45272083 layer.2.v_cache 0.00000107 0.00360848 layer.3.k_cache 0.00132320 1.73754019 layer.3.v_cache 0.00000211 0.00612722 layer.4.k_cache 0.00344917 3.38379642 layer.4.v_cache 0.00000308 0.01048037 layer.4.output 0.00019248 0.18578796 ------------------------------------------------------------------------------------- TOTAL 0.00264423 2.52597269 (elements=2,379,776) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2379776 Total Bytes 60636 BPFP 0.2038 bits/point EBPFP 0.4077 equivalent bits/point MSE 2.525973 ---------------------- -------------------------------------------------------- Time: 3.485s Load: 0.009s, Pack+Encode: 1.947s, Decode+Unpack: 1.528s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 166, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 166, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5260 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample78-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample78-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample79-layer4-item1.zst (82/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample79-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 157, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 157, 128) Output shape: (1, 157, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.0.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.1.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.1.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.2.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.2.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.3.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.3.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.4.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.4.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.4.output: torch.Size([1, 157, 4096]) -> torch.Size([1, 1, 157, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,012B, BPFP=0.0504 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,852B, BPFP=0.2912 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,636B, BPFP=0.2307 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,836B, BPFP=0.2904 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,520B, BPFP=0.2747 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,984B, BPFP=0.3475 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,096B, BPFP=0.3033 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,568B, BPFP=0.3268 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,252B, BPFP=0.2613 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,620B, BPFP=0.3294 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,972B, BPFP=0.0494 ⌛️ [2/4] FRONTEND: Frontend time: 1.886s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 157, 128]) layer.0.v_cache: torch.Size([1, 8, 157, 128]) layer.1.k_cache: torch.Size([1, 8, 157, 128]) layer.1.v_cache: torch.Size([1, 8, 157, 128]) layer.2.k_cache: torch.Size([1, 8, 157, 128]) layer.2.v_cache: torch.Size([1, 8, 157, 128]) layer.3.k_cache: torch.Size([1, 8, 157, 128]) layer.3.v_cache: torch.Size([1, 8, 157, 128]) layer.4.k_cache: torch.Size([1, 8, 157, 128]) layer.4.v_cache: torch.Size([1, 8, 157, 128]) layer.4.output: torch.Size([1, 157, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.589s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 157, 128]) layer.0.v_cache: torch.Size([1, 8, 157, 128]) layer.1.k_cache: torch.Size([1, 8, 157, 128]) layer.1.v_cache: torch.Size([1, 8, 157, 128]) layer.2.k_cache: torch.Size([1, 8, 157, 128]) layer.2.v_cache: torch.Size([1, 8, 157, 128]) layer.3.k_cache: torch.Size([1, 8, 157, 128]) layer.3.v_cache: torch.Size([1, 8, 157, 128]) layer.4.k_cache: torch.Size([1, 8, 157, 128]) layer.4.v_cache: torch.Size([1, 8, 157, 128]) layer.4.output: torch.Size([1, 157, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02715074 30.43485022 layer.0.v_cache 0.00000026 0.00062115 layer.1.k_cache 0.00316500 3.32926367 layer.1.v_cache 0.00000094 0.00252653 layer.2.k_cache 0.00119041 1.47760175 layer.2.v_cache 0.00000113 0.00364403 layer.3.k_cache 0.00134161 1.74307776 layer.3.v_cache 0.00000209 0.00598958 layer.4.k_cache 0.00347424 3.73651317 layer.4.v_cache 0.00000301 0.01011381 layer.4.output 0.00014407 0.18632855 ------------------------------------------------------------------------------------- TOTAL 0.00263612 2.96353685 (elements=2,250,752) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2250752 Total Bytes 58348 BPFP 0.2074 bits/point EBPFP 0.4148 equivalent bits/point MSE 2.963537 ---------------------- -------------------------------------------------------- Time: 3.483s Load: 0.009s, Pack+Encode: 1.886s, Decode+Unpack: 1.589s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 157, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.9635 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample79-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample79-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample8-layer4-item1.zst (83/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample8-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 235, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.014s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 235, 128) Output shape: (1, 235, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 235, 128]) -> torch.Size([1, 1, 235, 1024]) layer.0.v_cache: torch.Size([1, 8, 235, 128]) -> torch.Size([1, 1, 235, 1024]) layer.1.k_cache: torch.Size([1, 8, 235, 128]) -> torch.Size([1, 1, 235, 1024]) layer.1.v_cache: torch.Size([1, 8, 235, 128]) -> torch.Size([1, 1, 235, 1024]) layer.2.k_cache: torch.Size([1, 8, 235, 128]) -> torch.Size([1, 1, 235, 1024]) layer.2.v_cache: torch.Size([1, 8, 235, 128]) -> torch.Size([1, 1, 235, 1024]) layer.3.k_cache: torch.Size([1, 8, 235, 128]) -> torch.Size([1, 1, 235, 1024]) layer.3.v_cache: torch.Size([1, 8, 235, 128]) -> torch.Size([1, 1, 235, 1024]) layer.4.k_cache: torch.Size([1, 8, 235, 128]) -> torch.Size([1, 1, 235, 1024]) layer.4.v_cache: torch.Size([1, 8, 235, 128]) -> torch.Size([1, 1, 235, 1024]) layer.4.output: torch.Size([1, 235, 4096]) -> torch.Size([1, 1, 235, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,260B, BPFP=0.0419 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,152B, BPFP=0.2378 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,476B, BPFP=0.2153 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 8,428B, BPFP=0.2802 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,204B, BPFP=0.2395 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 9,068B, BPFP=0.3015 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,648B, BPFP=0.2875 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 9,312B, BPFP=0.3096 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,420B, BPFP=0.2134 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 9,116B, BPFP=0.3031 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,508B, BPFP=0.0375 ⌛️ [2/4] FRONTEND: Frontend time: 2.352s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 235, 128]) layer.0.v_cache: torch.Size([1, 8, 235, 128]) layer.1.k_cache: torch.Size([1, 8, 235, 128]) layer.1.v_cache: torch.Size([1, 8, 235, 128]) layer.2.k_cache: torch.Size([1, 8, 235, 128]) layer.2.v_cache: torch.Size([1, 8, 235, 128]) layer.3.k_cache: torch.Size([1, 8, 235, 128]) layer.3.v_cache: torch.Size([1, 8, 235, 128]) layer.4.k_cache: torch.Size([1, 8, 235, 128]) layer.4.v_cache: torch.Size([1, 8, 235, 128]) layer.4.output: torch.Size([1, 235, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.771s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 235, 128]) layer.0.v_cache: torch.Size([1, 8, 235, 128]) layer.1.k_cache: torch.Size([1, 8, 235, 128]) layer.1.v_cache: torch.Size([1, 8, 235, 128]) layer.2.k_cache: torch.Size([1, 8, 235, 128]) layer.2.v_cache: torch.Size([1, 8, 235, 128]) layer.3.k_cache: torch.Size([1, 8, 235, 128]) layer.3.v_cache: torch.Size([1, 8, 235, 128]) layer.4.k_cache: torch.Size([1, 8, 235, 128]) layer.4.v_cache: torch.Size([1, 8, 235, 128]) layer.4.output: torch.Size([1, 235, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02682624 26.15247881 layer.0.v_cache 0.00000026 0.00060201 layer.1.k_cache 0.00299644 3.22069144 layer.1.v_cache 0.00000075 0.00230256 layer.2.k_cache 0.00117112 1.42302636 layer.2.v_cache 0.00000116 0.00335642 layer.3.k_cache 0.00131946 1.67378137 layer.3.v_cache 0.00000202 0.00565219 layer.4.k_cache 0.00363359 3.48409658 layer.4.v_cache 0.00000293 0.00944147 layer.4.output 0.00013720 0.16742408 ------------------------------------------------------------------------------------- TOTAL 0.00260734 2.61750896 (elements=3,368,960) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 3368960 Total Bytes 77592 BPFP 0.1843 bits/point EBPFP 0.3685 equivalent bits/point MSE 2.617509 ---------------------- -------------------------------------------------------- Time: 4.137s Load: 0.014s, Pack+Encode: 2.352s, Decode+Unpack: 1.771s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 235, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 235, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6175 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample8-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample8-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample81-layer4-item1.zst (84/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample81-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 160, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 160, 128) Output shape: (1, 160, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 160, 128]) -> torch.Size([1, 1, 160, 1024]) layer.0.v_cache: torch.Size([1, 8, 160, 128]) -> torch.Size([1, 1, 160, 1024]) layer.1.k_cache: torch.Size([1, 8, 160, 128]) -> torch.Size([1, 1, 160, 1024]) layer.1.v_cache: torch.Size([1, 8, 160, 128]) -> torch.Size([1, 1, 160, 1024]) layer.2.k_cache: torch.Size([1, 8, 160, 128]) -> torch.Size([1, 1, 160, 1024]) layer.2.v_cache: torch.Size([1, 8, 160, 128]) -> torch.Size([1, 1, 160, 1024]) layer.3.k_cache: torch.Size([1, 8, 160, 128]) -> torch.Size([1, 1, 160, 1024]) layer.3.v_cache: torch.Size([1, 8, 160, 128]) -> torch.Size([1, 1, 160, 1024]) layer.4.k_cache: torch.Size([1, 8, 160, 128]) -> torch.Size([1, 1, 160, 1024]) layer.4.v_cache: torch.Size([1, 8, 160, 128]) -> torch.Size([1, 1, 160, 1024]) layer.4.output: torch.Size([1, 160, 4096]) -> torch.Size([1, 1, 160, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,016B, BPFP=0.0496 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,468B, BPFP=0.2670 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,192B, BPFP=0.2535 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,756B, BPFP=0.2811 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,772B, BPFP=0.2818 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,264B, BPFP=0.3059 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,364B, BPFP=0.3107 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,020B, BPFP=0.3428 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,624B, BPFP=0.2258 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,784B, BPFP=0.3312 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,576B, BPFP=0.0437 ⌛️ [2/4] FRONTEND: Frontend time: 1.975s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 160, 128]) layer.0.v_cache: torch.Size([1, 8, 160, 128]) layer.1.k_cache: torch.Size([1, 8, 160, 128]) layer.1.v_cache: torch.Size([1, 8, 160, 128]) layer.2.k_cache: torch.Size([1, 8, 160, 128]) layer.2.v_cache: torch.Size([1, 8, 160, 128]) layer.3.k_cache: torch.Size([1, 8, 160, 128]) layer.3.v_cache: torch.Size([1, 8, 160, 128]) layer.4.k_cache: torch.Size([1, 8, 160, 128]) layer.4.v_cache: torch.Size([1, 8, 160, 128]) layer.4.output: torch.Size([1, 160, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.577s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 160, 128]) layer.0.v_cache: torch.Size([1, 8, 160, 128]) layer.1.k_cache: torch.Size([1, 8, 160, 128]) layer.1.v_cache: torch.Size([1, 8, 160, 128]) layer.2.k_cache: torch.Size([1, 8, 160, 128]) layer.2.v_cache: torch.Size([1, 8, 160, 128]) layer.3.k_cache: torch.Size([1, 8, 160, 128]) layer.3.v_cache: torch.Size([1, 8, 160, 128]) layer.4.k_cache: torch.Size([1, 8, 160, 128]) layer.4.v_cache: torch.Size([1, 8, 160, 128]) layer.4.output: torch.Size([1, 160, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02677556 26.97220154 layer.0.v_cache 0.00000027 0.00061306 layer.1.k_cache 0.00313261 3.51624374 layer.1.v_cache 0.00000088 0.00268609 layer.2.k_cache 0.00115055 1.48037071 layer.2.v_cache 0.00000113 0.00391896 layer.3.k_cache 0.00131762 1.76438141 layer.3.v_cache 0.00000244 0.00634132 layer.4.k_cache 0.00341674 3.30965538 layer.4.v_cache 0.00000319 0.01108496 layer.4.output 0.00016336 0.17343007 ------------------------------------------------------------------------------------- TOTAL 0.00260389 2.69722982 (elements=2,293,760) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2293760 Total Bytes 57836 BPFP 0.2017 bits/point EBPFP 0.4034 equivalent bits/point MSE 2.697230 ---------------------- -------------------------------------------------------- Time: 3.562s Load: 0.010s, Pack+Encode: 1.975s, Decode+Unpack: 1.577s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 160, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 160, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6972 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample81-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample81-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample82-layer4-item1.zst (85/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample82-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 163, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 163, 128) Output shape: (1, 163, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.output: torch.Size([1, 163, 4096]) -> torch.Size([1, 1, 163, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,004B, BPFP=0.0481 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,328B, BPFP=0.2554 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,048B, BPFP=0.2419 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,944B, BPFP=0.2849 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,868B, BPFP=0.2812 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,672B, BPFP=0.3198 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,036B, BPFP=0.3372 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,020B, BPFP=0.3365 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,204B, BPFP=0.2494 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,084B, BPFP=0.3395 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,132B, BPFP=0.0495 ⌛️ [2/4] FRONTEND: Frontend time: 1.873s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) layer.4.output: torch.Size([1, 163, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.594s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) layer.4.output: torch.Size([1, 163, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02751794 26.03978444 layer.0.v_cache 0.00000027 0.00061427 layer.1.k_cache 0.00304081 3.34228478 layer.1.v_cache 0.00000086 0.00256936 layer.2.k_cache 0.00115337 1.49848367 layer.2.v_cache 0.00000112 0.00371819 layer.3.k_cache 0.00134529 1.78609518 layer.3.v_cache 0.00000213 0.00618448 layer.4.k_cache 0.00339313 3.46520172 layer.4.v_cache 0.00000308 0.01059681 layer.4.output 0.00017108 0.19006653 ------------------------------------------------------------------------------------- TOTAL 0.00265302 2.63684279 (elements=2,336,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2336768 Total Bytes 60340 BPFP 0.2066 bits/point EBPFP 0.4132 equivalent bits/point MSE 2.636843 ---------------------- -------------------------------------------------------- Time: 3.478s Load: 0.011s, Pack+Encode: 1.873s, Decode+Unpack: 1.594s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 163, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6368 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample82-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample82-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample84-layer4-item1.zst (86/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample84-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 173, 128) Output shape: (1, 173, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) -> torch.Size([1, 1, 173, 1024]) layer.4.output: torch.Size([1, 173, 4096]) -> torch.Size([1, 1, 173, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,052B, BPFP=0.0475 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,384B, BPFP=0.2431 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,632B, BPFP=0.2543 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,564B, BPFP=0.2964 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,700B, BPFP=0.2574 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,400B, BPFP=0.3342 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,700B, BPFP=0.3026 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,460B, BPFP=0.3369 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,972B, BPFP=0.2245 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,524B, BPFP=0.3398 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,928B, BPFP=0.0443 ⌛️ [2/4] FRONTEND: Frontend time: 1.942s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) layer.4.output: torch.Size([1, 173, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.427s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 173, 128]) layer.0.v_cache: torch.Size([1, 8, 173, 128]) layer.1.k_cache: torch.Size([1, 8, 173, 128]) layer.1.v_cache: torch.Size([1, 8, 173, 128]) layer.2.k_cache: torch.Size([1, 8, 173, 128]) layer.2.v_cache: torch.Size([1, 8, 173, 128]) layer.3.k_cache: torch.Size([1, 8, 173, 128]) layer.3.v_cache: torch.Size([1, 8, 173, 128]) layer.4.k_cache: torch.Size([1, 8, 173, 128]) layer.4.v_cache: torch.Size([1, 8, 173, 128]) layer.4.output: torch.Size([1, 173, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02651478 26.76528066 layer.0.v_cache 0.00000027 0.00057533 layer.1.k_cache 0.00306594 3.34487897 layer.1.v_cache 0.00000080 0.00250849 layer.2.k_cache 0.00117582 1.45110184 layer.2.v_cache 0.00000111 0.00377238 layer.3.k_cache 0.00132699 1.74381234 layer.3.v_cache 0.00000224 0.00610126 layer.4.k_cache 0.00342058 3.46120316 layer.4.v_cache 0.00000347 0.01022764 layer.4.output 0.00021934 0.17134392 ------------------------------------------------------------------------------------- TOTAL 0.00259924 2.67677413 (elements=2,480,128) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2480128 Total Bytes 62316 BPFP 0.2010 bits/point EBPFP 0.4020 equivalent bits/point MSE 2.676774 ---------------------- -------------------------------------------------------- Time: 3.379s Load: 0.009s, Pack+Encode: 1.942s, Decode+Unpack: 1.427s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 173, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 173, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6768 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample84-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample84-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample85-layer4-item1.zst (87/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample85-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 170, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 170, 128) Output shape: (1, 170, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.0.v_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.1.k_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.1.v_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.2.k_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.2.v_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.3.k_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.3.v_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.4.k_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.4.v_cache: torch.Size([1, 8, 170, 128]) -> torch.Size([1, 1, 170, 1024]) layer.4.output: torch.Size([1, 170, 4096]) -> torch.Size([1, 1, 170, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,032B, BPFP=0.0474 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,464B, BPFP=0.2511 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,588B, BPFP=0.2568 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,072B, BPFP=0.2790 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,476B, BPFP=0.2517 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,216B, BPFP=0.3316 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,444B, BPFP=0.2961 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,968B, BPFP=0.3202 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,124B, BPFP=0.2355 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,212B, BPFP=0.3314 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,072B, BPFP=0.0468 ⌛️ [2/4] FRONTEND: Frontend time: 1.904s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 170, 128]) layer.0.v_cache: torch.Size([1, 8, 170, 128]) layer.1.k_cache: torch.Size([1, 8, 170, 128]) layer.1.v_cache: torch.Size([1, 8, 170, 128]) layer.2.k_cache: torch.Size([1, 8, 170, 128]) layer.2.v_cache: torch.Size([1, 8, 170, 128]) layer.3.k_cache: torch.Size([1, 8, 170, 128]) layer.3.v_cache: torch.Size([1, 8, 170, 128]) layer.4.k_cache: torch.Size([1, 8, 170, 128]) layer.4.v_cache: torch.Size([1, 8, 170, 128]) layer.4.output: torch.Size([1, 170, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.589s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 170, 128]) layer.0.v_cache: torch.Size([1, 8, 170, 128]) layer.1.k_cache: torch.Size([1, 8, 170, 128]) layer.1.v_cache: torch.Size([1, 8, 170, 128]) layer.2.k_cache: torch.Size([1, 8, 170, 128]) layer.2.v_cache: torch.Size([1, 8, 170, 128]) layer.3.k_cache: torch.Size([1, 8, 170, 128]) layer.3.v_cache: torch.Size([1, 8, 170, 128]) layer.4.k_cache: torch.Size([1, 8, 170, 128]) layer.4.v_cache: torch.Size([1, 8, 170, 128]) layer.4.output: torch.Size([1, 170, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02725268 27.40157399 layer.0.v_cache 0.00000026 0.00061787 layer.1.k_cache 0.00300898 3.32053941 layer.1.v_cache 0.00000090 0.00268794 layer.2.k_cache 0.00114627 1.43802598 layer.2.v_cache 0.00000112 0.00380951 layer.3.k_cache 0.00135579 1.73251289 layer.3.v_cache 0.00000210 0.00614270 layer.4.k_cache 0.00341688 3.40615522 layer.4.v_cache 0.00000298 0.01018067 layer.4.output 0.00021049 0.20049946 ------------------------------------------------------------------------------------- TOTAL 0.00264500 2.72316028 (elements=2,437,120) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2437120 Total Bytes 60668 BPFP 0.1991 bits/point EBPFP 0.3983 equivalent bits/point MSE 2.723160 ---------------------- -------------------------------------------------------- Time: 3.504s Load: 0.010s, Pack+Encode: 1.904s, Decode+Unpack: 1.589s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 170, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 170, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7232 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample85-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample85-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample86-layer4-item1.zst (88/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample86-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 163, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 163, 128) Output shape: (1, 163, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) -> torch.Size([1, 1, 163, 1024]) layer.4.output: torch.Size([1, 163, 4096]) -> torch.Size([1, 1, 163, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,000B, BPFP=0.0479 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,360B, BPFP=0.2569 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,620B, BPFP=0.2694 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,232B, BPFP=0.2987 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,832B, BPFP=0.2795 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,640B, BPFP=0.3183 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,956B, BPFP=0.3334 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,776B, BPFP=0.3248 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,284B, BPFP=0.2533 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,584B, BPFP=0.3156 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,724B, BPFP=0.0446 ⌛️ [2/4] FRONTEND: Frontend time: 1.888s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) layer.4.output: torch.Size([1, 163, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.474s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 163, 128]) layer.0.v_cache: torch.Size([1, 8, 163, 128]) layer.1.k_cache: torch.Size([1, 8, 163, 128]) layer.1.v_cache: torch.Size([1, 8, 163, 128]) layer.2.k_cache: torch.Size([1, 8, 163, 128]) layer.2.v_cache: torch.Size([1, 8, 163, 128]) layer.3.k_cache: torch.Size([1, 8, 163, 128]) layer.3.v_cache: torch.Size([1, 8, 163, 128]) layer.4.k_cache: torch.Size([1, 8, 163, 128]) layer.4.v_cache: torch.Size([1, 8, 163, 128]) layer.4.output: torch.Size([1, 163, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02663420 24.27797432 layer.0.v_cache 0.00000027 0.00062233 layer.1.k_cache 0.00304468 3.76418412 layer.1.v_cache 0.00000080 0.00264124 layer.2.k_cache 0.00117203 1.44521233 layer.2.v_cache 0.00000109 0.00376141 layer.3.k_cache 0.00131422 1.71829280 layer.3.v_cache 0.00000209 0.00624852 layer.4.k_cache 0.00347827 3.54402694 layer.4.v_cache 0.00000310 0.01051331 layer.4.output 0.00017721 0.20094358 ------------------------------------------------------------------------------------- TOTAL 0.00259711 2.54123226 (elements=2,336,768) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2336768 Total Bytes 60008 BPFP 0.2054 bits/point EBPFP 0.4109 equivalent bits/point MSE 2.541232 ---------------------- -------------------------------------------------------- Time: 3.371s Load: 0.009s, Pack+Encode: 1.888s, Decode+Unpack: 1.474s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 163, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 163, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.5412 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample86-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample86-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample87-layer4-item1.zst (89/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample87-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 175, 128) Output shape: (1, 175, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) -> torch.Size([1, 1, 175, 1024]) layer.4.output: torch.Size([1, 175, 4096]) -> torch.Size([1, 1, 175, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,036B, BPFP=0.0462 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,980B, BPFP=0.2670 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,488B, BPFP=0.2450 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,412B, BPFP=0.2863 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,516B, BPFP=0.2462 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,368B, BPFP=0.3289 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,544B, BPFP=0.2921 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,644B, BPFP=0.3412 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,968B, BPFP=0.2218 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,440B, BPFP=0.3321 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,144B, BPFP=0.0462 ⌛️ [2/4] FRONTEND: Frontend time: 1.985s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) layer.4.output: torch.Size([1, 175, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.484s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 175, 128]) layer.0.v_cache: torch.Size([1, 8, 175, 128]) layer.1.k_cache: torch.Size([1, 8, 175, 128]) layer.1.v_cache: torch.Size([1, 8, 175, 128]) layer.2.k_cache: torch.Size([1, 8, 175, 128]) layer.2.v_cache: torch.Size([1, 8, 175, 128]) layer.3.k_cache: torch.Size([1, 8, 175, 128]) layer.3.v_cache: torch.Size([1, 8, 175, 128]) layer.4.k_cache: torch.Size([1, 8, 175, 128]) layer.4.v_cache: torch.Size([1, 8, 175, 128]) layer.4.output: torch.Size([1, 175, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02610176 28.52529018 layer.0.v_cache 0.00000027 0.00063047 layer.1.k_cache 0.00311052 3.69551409 layer.1.v_cache 0.00000081 0.00268450 layer.2.k_cache 0.00119380 1.46795306 layer.2.v_cache 0.00000114 0.00383366 layer.3.k_cache 0.00132266 1.76495152 layer.3.v_cache 0.00000212 0.00651862 layer.4.k_cache 0.00346741 3.35172747 layer.4.v_cache 0.00000313 0.01067196 layer.4.output 0.00016538 0.18207414 ------------------------------------------------------------------------------------- TOTAL 0.00256179 2.82557658 (elements=2,508,800) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2508800 Total Bytes 62540 BPFP 0.1994 bits/point EBPFP 0.3989 equivalent bits/point MSE 2.825577 ---------------------- -------------------------------------------------------- Time: 3.480s Load: 0.011s, Pack+Encode: 1.985s, Decode+Unpack: 1.484s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 175, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 175, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8256 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample87-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample87-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample88-layer4-item1.zst (90/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample88-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 172, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 172, 128) Output shape: (1, 172, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 172, 128]) -> torch.Size([1, 1, 172, 1024]) layer.0.v_cache: torch.Size([1, 8, 172, 128]) -> torch.Size([1, 1, 172, 1024]) layer.1.k_cache: torch.Size([1, 8, 172, 128]) -> torch.Size([1, 1, 172, 1024]) layer.1.v_cache: torch.Size([1, 8, 172, 128]) -> torch.Size([1, 1, 172, 1024]) layer.2.k_cache: torch.Size([1, 8, 172, 128]) -> torch.Size([1, 1, 172, 1024]) layer.2.v_cache: torch.Size([1, 8, 172, 128]) -> torch.Size([1, 1, 172, 1024]) layer.3.k_cache: torch.Size([1, 8, 172, 128]) -> torch.Size([1, 1, 172, 1024]) layer.3.v_cache: torch.Size([1, 8, 172, 128]) -> torch.Size([1, 1, 172, 1024]) layer.4.k_cache: torch.Size([1, 8, 172, 128]) -> torch.Size([1, 1, 172, 1024]) layer.4.v_cache: torch.Size([1, 8, 172, 128]) -> torch.Size([1, 1, 172, 1024]) layer.4.output: torch.Size([1, 172, 4096]) -> torch.Size([1, 1, 172, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,008B, BPFP=0.0458 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,956B, BPFP=0.2705 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,428B, BPFP=0.2465 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,492B, BPFP=0.2949 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,596B, BPFP=0.2542 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,396B, BPFP=0.3359 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,632B, BPFP=0.3012 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,040B, BPFP=0.3198 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,044B, BPFP=0.2291 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,056B, BPFP=0.3205 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,972B, BPFP=0.0451 ⌛️ [2/4] FRONTEND: Frontend time: 1.865s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 172, 128]) layer.0.v_cache: torch.Size([1, 8, 172, 128]) layer.1.k_cache: torch.Size([1, 8, 172, 128]) layer.1.v_cache: torch.Size([1, 8, 172, 128]) layer.2.k_cache: torch.Size([1, 8, 172, 128]) layer.2.v_cache: torch.Size([1, 8, 172, 128]) layer.3.k_cache: torch.Size([1, 8, 172, 128]) layer.3.v_cache: torch.Size([1, 8, 172, 128]) layer.4.k_cache: torch.Size([1, 8, 172, 128]) layer.4.v_cache: torch.Size([1, 8, 172, 128]) layer.4.output: torch.Size([1, 172, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.696s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 172, 128]) layer.0.v_cache: torch.Size([1, 8, 172, 128]) layer.1.k_cache: torch.Size([1, 8, 172, 128]) layer.1.v_cache: torch.Size([1, 8, 172, 128]) layer.2.k_cache: torch.Size([1, 8, 172, 128]) layer.2.v_cache: torch.Size([1, 8, 172, 128]) layer.3.k_cache: torch.Size([1, 8, 172, 128]) layer.3.v_cache: torch.Size([1, 8, 172, 128]) layer.4.k_cache: torch.Size([1, 8, 172, 128]) layer.4.v_cache: torch.Size([1, 8, 172, 128]) layer.4.output: torch.Size([1, 172, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02694283 26.57216059 layer.0.v_cache 0.00000026 0.00062773 layer.1.k_cache 0.00313726 3.12232829 layer.1.v_cache 0.00000082 0.00254902 layer.2.k_cache 0.00118743 1.45513472 layer.2.v_cache 0.00000107 0.00357243 layer.3.k_cache 0.00136375 1.70950672 layer.3.v_cache 0.00000202 0.00591524 layer.4.k_cache 0.00345272 3.51252818 layer.4.v_cache 0.00000296 0.01016206 layer.4.output 0.00020827 0.19752742 ------------------------------------------------------------------------------------- TOTAL 0.00263744 2.65604248 (elements=2,465,792) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2465792 Total Bytes 61620 BPFP 0.1999 bits/point EBPFP 0.3998 equivalent bits/point MSE 2.656042 ---------------------- -------------------------------------------------------- Time: 3.571s Load: 0.010s, Pack+Encode: 1.865s, Decode+Unpack: 1.696s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 172, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 172, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6560 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample88-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample88-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample89-layer4-item1.zst (91/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample89-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 169, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 169, 128) Output shape: (1, 169, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.output: torch.Size([1, 169, 4096]) -> torch.Size([1, 1, 169, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,032B, BPFP=0.0477 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,580B, BPFP=0.2580 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,596B, BPFP=0.2587 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,180B, BPFP=0.2857 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,604B, BPFP=0.2591 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,816B, BPFP=0.3151 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,828B, BPFP=0.3156 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,156B, BPFP=0.3308 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,100B, BPFP=0.2358 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,260B, BPFP=0.3356 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,480B, BPFP=0.0402 ⌛️ [2/4] FRONTEND: Frontend time: 1.935s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) layer.4.output: torch.Size([1, 169, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.426s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) layer.4.output: torch.Size([1, 169, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02715943 28.29868366 layer.0.v_cache 0.00000026 0.00060479 layer.1.k_cache 0.00301211 3.69271219 layer.1.v_cache 0.00000078 0.00247345 layer.2.k_cache 0.00115479 1.44362880 layer.2.v_cache 0.00000109 0.00353387 layer.3.k_cache 0.00130335 1.75390950 layer.3.v_cache 0.00000203 0.00590745 layer.4.k_cache 0.00348027 3.55869921 layer.4.v_cache 0.00000300 0.00986027 layer.4.output 0.00019495 0.19058239 ------------------------------------------------------------------------------------- TOTAL 0.00263549 2.82373877 (elements=2,422,784) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2422784 Total Bytes 60632 BPFP 0.2002 bits/point EBPFP 0.4004 equivalent bits/point MSE 2.823739 ---------------------- -------------------------------------------------------- Time: 3.370s Load: 0.009s, Pack+Encode: 1.935s, Decode+Unpack: 1.426s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 169, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8237 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample89-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample89-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample9-layer4-item1.zst (92/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample9-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 216, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.011s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 216, 128) Output shape: (1, 216, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 216, 128]) -> torch.Size([1, 1, 216, 1024]) layer.0.v_cache: torch.Size([1, 8, 216, 128]) -> torch.Size([1, 1, 216, 1024]) layer.1.k_cache: torch.Size([1, 8, 216, 128]) -> torch.Size([1, 1, 216, 1024]) layer.1.v_cache: torch.Size([1, 8, 216, 128]) -> torch.Size([1, 1, 216, 1024]) layer.2.k_cache: torch.Size([1, 8, 216, 128]) -> torch.Size([1, 1, 216, 1024]) layer.2.v_cache: torch.Size([1, 8, 216, 128]) -> torch.Size([1, 1, 216, 1024]) layer.3.k_cache: torch.Size([1, 8, 216, 128]) -> torch.Size([1, 1, 216, 1024]) layer.3.v_cache: torch.Size([1, 8, 216, 128]) -> torch.Size([1, 1, 216, 1024]) layer.4.k_cache: torch.Size([1, 8, 216, 128]) -> torch.Size([1, 1, 216, 1024]) layer.4.v_cache: torch.Size([1, 8, 216, 128]) -> torch.Size([1, 1, 216, 1024]) layer.4.output: torch.Size([1, 216, 4096]) -> torch.Size([1, 1, 216, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,208B, BPFP=0.0437 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 7,152B, BPFP=0.2587 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 6,272B, BPFP=0.2269 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 8,172B, BPFP=0.2956 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 7,644B, BPFP=0.2765 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 9,316B, BPFP=0.3370 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 8,328B, BPFP=0.3012 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 9,212B, BPFP=0.3332 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 6,364B, BPFP=0.2302 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 9,552B, BPFP=0.3455 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 5,196B, BPFP=0.0470 ⌛️ [2/4] FRONTEND: Frontend time: 2.351s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 216, 128]) layer.0.v_cache: torch.Size([1, 8, 216, 128]) layer.1.k_cache: torch.Size([1, 8, 216, 128]) layer.1.v_cache: torch.Size([1, 8, 216, 128]) layer.2.k_cache: torch.Size([1, 8, 216, 128]) layer.2.v_cache: torch.Size([1, 8, 216, 128]) layer.3.k_cache: torch.Size([1, 8, 216, 128]) layer.3.v_cache: torch.Size([1, 8, 216, 128]) layer.4.k_cache: torch.Size([1, 8, 216, 128]) layer.4.v_cache: torch.Size([1, 8, 216, 128]) layer.4.output: torch.Size([1, 216, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.948s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 216, 128]) layer.0.v_cache: torch.Size([1, 8, 216, 128]) layer.1.k_cache: torch.Size([1, 8, 216, 128]) layer.1.v_cache: torch.Size([1, 8, 216, 128]) layer.2.k_cache: torch.Size([1, 8, 216, 128]) layer.2.v_cache: torch.Size([1, 8, 216, 128]) layer.3.k_cache: torch.Size([1, 8, 216, 128]) layer.3.v_cache: torch.Size([1, 8, 216, 128]) layer.4.k_cache: torch.Size([1, 8, 216, 128]) layer.4.v_cache: torch.Size([1, 8, 216, 128]) layer.4.output: torch.Size([1, 216, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02680264 29.43734176 layer.0.v_cache 0.00000027 0.00062246 layer.1.k_cache 0.00301343 3.12320511 layer.1.v_cache 0.00000084 0.00248935 layer.2.k_cache 0.00116753 1.46921299 layer.2.v_cache 0.00000113 0.00368556 layer.3.k_cache 0.00132686 1.70744719 layer.3.v_cache 0.00000211 0.00620803 layer.4.k_cache 0.00344202 3.51207140 layer.4.v_cache 0.00000310 0.01049255 layer.4.output 0.00021250 0.20437680 ------------------------------------------------------------------------------------- TOTAL 0.00261500 2.86359169 (elements=3,096,576) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 3096576 Total Bytes 78416 BPFP 0.2026 bits/point EBPFP 0.4052 equivalent bits/point MSE 2.863592 ---------------------- -------------------------------------------------------- Time: 4.311s Load: 0.011s, Pack+Encode: 2.351s, Decode+Unpack: 1.948s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 216, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 216, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8636 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample9-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample9-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample90-layer4-item1.zst (93/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample90-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 174, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 174, 128) Output shape: (1, 174, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.0.v_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.1.k_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.1.v_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.2.k_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.2.v_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.3.k_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.3.v_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.4.k_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.4.v_cache: torch.Size([1, 8, 174, 128]) -> torch.Size([1, 1, 174, 1024]) layer.4.output: torch.Size([1, 174, 4096]) -> torch.Size([1, 1, 174, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,008B, BPFP=0.0453 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,960B, BPFP=0.2676 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,460B, BPFP=0.2452 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,344B, BPFP=0.2848 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,420B, BPFP=0.2434 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,920B, BPFP=0.3107 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,508B, BPFP=0.2922 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,208B, BPFP=0.3236 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,708B, BPFP=0.2114 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,476B, BPFP=0.3357 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,772B, BPFP=0.0423 ⌛️ [2/4] FRONTEND: Frontend time: 1.872s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 174, 128]) layer.0.v_cache: torch.Size([1, 8, 174, 128]) layer.1.k_cache: torch.Size([1, 8, 174, 128]) layer.1.v_cache: torch.Size([1, 8, 174, 128]) layer.2.k_cache: torch.Size([1, 8, 174, 128]) layer.2.v_cache: torch.Size([1, 8, 174, 128]) layer.3.k_cache: torch.Size([1, 8, 174, 128]) layer.3.v_cache: torch.Size([1, 8, 174, 128]) layer.4.k_cache: torch.Size([1, 8, 174, 128]) layer.4.v_cache: torch.Size([1, 8, 174, 128]) layer.4.output: torch.Size([1, 174, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.604s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 174, 128]) layer.0.v_cache: torch.Size([1, 8, 174, 128]) layer.1.k_cache: torch.Size([1, 8, 174, 128]) layer.1.v_cache: torch.Size([1, 8, 174, 128]) layer.2.k_cache: torch.Size([1, 8, 174, 128]) layer.2.v_cache: torch.Size([1, 8, 174, 128]) layer.3.k_cache: torch.Size([1, 8, 174, 128]) layer.3.v_cache: torch.Size([1, 8, 174, 128]) layer.4.k_cache: torch.Size([1, 8, 174, 128]) layer.4.v_cache: torch.Size([1, 8, 174, 128]) layer.4.output: torch.Size([1, 174, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02737866 28.10909438 layer.0.v_cache 0.00000027 0.00060558 layer.1.k_cache 0.00309889 3.58423764 layer.1.v_cache 0.00000078 0.00245354 layer.2.k_cache 0.00114661 1.47743558 layer.2.v_cache 0.00000107 0.00358020 layer.3.k_cache 0.00139116 1.77741934 layer.3.v_cache 0.00000205 0.00607746 layer.4.k_cache 0.00338123 3.41890217 layer.4.v_cache 0.00000306 0.01058484 layer.4.output 0.00019207 0.18558430 ------------------------------------------------------------------------------------- TOTAL 0.00265515 2.79519485 (elements=2,494,464) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2494464 Total Bytes 60784 BPFP 0.1949 bits/point EBPFP 0.3899 equivalent bits/point MSE 2.795195 ---------------------- -------------------------------------------------------- Time: 3.485s Load: 0.009s, Pack+Encode: 1.872s, Decode+Unpack: 1.604s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 174, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 174, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7952 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample90-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample90-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample91-layer4-item1.zst (94/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample91-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 169, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 169, 128) Output shape: (1, 169, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) -> torch.Size([1, 1, 169, 1024]) layer.4.output: torch.Size([1, 169, 4096]) -> torch.Size([1, 1, 169, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,028B, BPFP=0.0475 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,764B, BPFP=0.2665 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,432B, BPFP=0.2511 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,504B, BPFP=0.3007 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,552B, BPFP=0.2567 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,648B, BPFP=0.3536 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,888B, BPFP=0.3184 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,252B, BPFP=0.3352 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,276B, BPFP=0.2439 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,316B, BPFP=0.3382 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,512B, BPFP=0.0406 ⌛️ [2/4] FRONTEND: Frontend time: 1.939s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) layer.4.output: torch.Size([1, 169, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.469s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 169, 128]) layer.0.v_cache: torch.Size([1, 8, 169, 128]) layer.1.k_cache: torch.Size([1, 8, 169, 128]) layer.1.v_cache: torch.Size([1, 8, 169, 128]) layer.2.k_cache: torch.Size([1, 8, 169, 128]) layer.2.v_cache: torch.Size([1, 8, 169, 128]) layer.3.k_cache: torch.Size([1, 8, 169, 128]) layer.3.v_cache: torch.Size([1, 8, 169, 128]) layer.4.k_cache: torch.Size([1, 8, 169, 128]) layer.4.v_cache: torch.Size([1, 8, 169, 128]) layer.4.output: torch.Size([1, 169, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02703965 27.06396773 layer.0.v_cache 0.00000028 0.00059900 layer.1.k_cache 0.00299766 3.42622475 layer.1.v_cache 0.00000081 0.00244713 layer.2.k_cache 0.00115006 1.45112673 layer.2.v_cache 0.00000108 0.00350201 layer.3.k_cache 0.00135242 1.72387154 layer.3.v_cache 0.00000202 0.00582330 layer.4.k_cache 0.00346273 3.45572279 layer.4.v_cache 0.00000299 0.01002094 layer.4.output 0.00018472 0.19647537 ------------------------------------------------------------------------------------- TOTAL 0.00262490 2.70922910 (elements=2,422,784) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2422784 Total Bytes 62172 BPFP 0.2053 bits/point EBPFP 0.4106 equivalent bits/point MSE 2.709229 ---------------------- -------------------------------------------------------- Time: 3.416s Load: 0.009s, Pack+Encode: 1.939s, Decode+Unpack: 1.469s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 169, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 169, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7092 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample91-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample91-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample92-layer4-item1.zst (95/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample92-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 171, 128) Output shape: (1, 171, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.output: torch.Size([1, 171, 4096]) -> torch.Size([1, 1, 171, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,020B, BPFP=0.0466 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,652B, BPFP=0.2582 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,696B, BPFP=0.2602 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,168B, BPFP=0.2818 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,672B, BPFP=0.2591 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,308B, BPFP=0.3339 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,712B, BPFP=0.3067 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,400B, BPFP=0.3381 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,276B, BPFP=0.2410 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,048B, BPFP=0.3220 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,876B, BPFP=0.0443 ⌛️ [2/4] FRONTEND: Frontend time: 1.927s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.596s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02767900 30.29920219 layer.0.v_cache 0.00000027 0.00062958 layer.1.k_cache 0.00300022 3.36470826 layer.1.v_cache 0.00000079 0.00248685 layer.2.k_cache 0.00116824 1.45572497 layer.2.v_cache 0.00000115 0.00369253 layer.3.k_cache 0.00134906 1.74716133 layer.3.v_cache 0.00000206 0.00598885 layer.4.k_cache 0.00347571 3.50300964 layer.4.v_cache 0.00000302 0.01035367 layer.4.output 0.00024629 0.20140515 ------------------------------------------------------------------------------------- TOTAL 0.00269033 2.94275561 (elements=2,451,456) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2451456 Total Bytes 61828 BPFP 0.2018 bits/point EBPFP 0.4035 equivalent bits/point MSE 2.942756 ---------------------- -------------------------------------------------------- Time: 3.532s Load: 0.010s, Pack+Encode: 1.927s, Decode+Unpack: 1.596s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.9428 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample92-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample92-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample94-layer4-item1.zst (96/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample94-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 171, 128) Output shape: (1, 171, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) -> torch.Size([1, 1, 171, 1024]) layer.4.output: torch.Size([1, 171, 4096]) -> torch.Size([1, 1, 171, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,040B, BPFP=0.0475 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,596B, BPFP=0.2557 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,632B, BPFP=0.2573 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,296B, BPFP=0.2876 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,784B, BPFP=0.2643 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 7,520B, BPFP=0.3436 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,796B, BPFP=0.3105 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,344B, BPFP=0.3355 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,156B, BPFP=0.2356 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,688B, BPFP=0.3512 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,668B, BPFP=0.0419 ⌛️ [2/4] FRONTEND: Frontend time: 1.864s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.484s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 171, 128]) layer.0.v_cache: torch.Size([1, 8, 171, 128]) layer.1.k_cache: torch.Size([1, 8, 171, 128]) layer.1.v_cache: torch.Size([1, 8, 171, 128]) layer.2.k_cache: torch.Size([1, 8, 171, 128]) layer.2.v_cache: torch.Size([1, 8, 171, 128]) layer.3.k_cache: torch.Size([1, 8, 171, 128]) layer.3.v_cache: torch.Size([1, 8, 171, 128]) layer.4.k_cache: torch.Size([1, 8, 171, 128]) layer.4.v_cache: torch.Size([1, 8, 171, 128]) layer.4.output: torch.Size([1, 171, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02680470 27.53050758 layer.0.v_cache 0.00000027 0.00060199 layer.1.k_cache 0.00310069 3.33756510 layer.1.v_cache 0.00000077 0.00244956 layer.2.k_cache 0.00116404 1.44301976 layer.2.v_cache 0.00000106 0.00348728 layer.3.k_cache 0.00136682 1.73090474 layer.3.v_cache 0.00000197 0.00580269 layer.4.k_cache 0.00339793 3.40310615 layer.4.v_cache 0.00000297 0.00980841 layer.4.output 0.00018087 0.17833589 ------------------------------------------------------------------------------------- TOTAL 0.00261177 2.72718549 (elements=2,451,456) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2451456 Total Bytes 62520 BPFP 0.2040 bits/point EBPFP 0.4081 equivalent bits/point MSE 2.727185 ---------------------- -------------------------------------------------------- Time: 3.357s Load: 0.009s, Pack+Encode: 1.864s, Decode+Unpack: 1.484s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 171, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 171, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.7272 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample94-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample94-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample96-layer4-item1.zst (97/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample96-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 156, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 156, 128) Output shape: (1, 156, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.0.v_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.1.k_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.1.v_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.2.k_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.2.v_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.3.k_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.3.v_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.4.k_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.4.v_cache: torch.Size([1, 8, 156, 128]) -> torch.Size([1, 1, 156, 1024]) layer.4.output: torch.Size([1, 156, 4096]) -> torch.Size([1, 1, 156, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,024B, BPFP=0.0513 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 6,016B, BPFP=0.3013 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,072B, BPFP=0.2540 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,096B, BPFP=0.3053 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,424B, BPFP=0.2716 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,948B, BPFP=0.3480 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,336B, BPFP=0.3173 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,876B, BPFP=0.3444 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,056B, BPFP=0.2532 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,980B, BPFP=0.3496 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,900B, BPFP=0.0488 ⌛️ [2/4] FRONTEND: Frontend time: 1.958s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 156, 128]) layer.0.v_cache: torch.Size([1, 8, 156, 128]) layer.1.k_cache: torch.Size([1, 8, 156, 128]) layer.1.v_cache: torch.Size([1, 8, 156, 128]) layer.2.k_cache: torch.Size([1, 8, 156, 128]) layer.2.v_cache: torch.Size([1, 8, 156, 128]) layer.3.k_cache: torch.Size([1, 8, 156, 128]) layer.3.v_cache: torch.Size([1, 8, 156, 128]) layer.4.k_cache: torch.Size([1, 8, 156, 128]) layer.4.v_cache: torch.Size([1, 8, 156, 128]) layer.4.output: torch.Size([1, 156, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.473s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 156, 128]) layer.0.v_cache: torch.Size([1, 8, 156, 128]) layer.1.k_cache: torch.Size([1, 8, 156, 128]) layer.1.v_cache: torch.Size([1, 8, 156, 128]) layer.2.k_cache: torch.Size([1, 8, 156, 128]) layer.2.v_cache: torch.Size([1, 8, 156, 128]) layer.3.k_cache: torch.Size([1, 8, 156, 128]) layer.3.v_cache: torch.Size([1, 8, 156, 128]) layer.4.k_cache: torch.Size([1, 8, 156, 128]) layer.4.v_cache: torch.Size([1, 8, 156, 128]) layer.4.output: torch.Size([1, 156, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02775914 30.82945838 layer.0.v_cache 0.00000026 0.00061363 layer.1.k_cache 0.00308099 3.91151663 layer.1.v_cache 0.00000085 0.00252100 layer.2.k_cache 0.00113398 1.48158773 layer.2.v_cache 0.00000135 0.00363030 layer.3.k_cache 0.00135748 1.79930682 layer.3.v_cache 0.00000209 0.00588115 layer.4.k_cache 0.00343364 3.82121512 layer.4.v_cache 0.00000289 0.00981675 layer.4.output 0.00020185 0.21891824 ------------------------------------------------------------------------------------- TOTAL 0.00268429 3.05294432 (elements=2,236,416) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2236416 Total Bytes 59728 BPFP 0.2137 bits/point EBPFP 0.4273 equivalent bits/point MSE 3.052944 ---------------------- -------------------------------------------------------- Time: 3.441s Load: 0.009s, Pack+Encode: 1.958s, Decode+Unpack: 1.473s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 156, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 156, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 3.0529 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample96-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample96-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample97-layer4-item1.zst (98/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample97-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 162, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 162, 128) Output shape: (1, 162, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.0.v_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.1.k_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.1.v_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.2.k_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.2.v_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.3.k_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.3.v_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.4.k_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.4.v_cache: torch.Size([1, 8, 162, 128]) -> torch.Size([1, 1, 162, 1024]) layer.4.output: torch.Size([1, 162, 4096]) -> torch.Size([1, 1, 162, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,012B, BPFP=0.0488 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,552B, BPFP=0.2677 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,768B, BPFP=0.2782 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,156B, BPFP=0.2969 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 6,024B, BPFP=0.2905 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,596B, BPFP=0.3181 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,088B, BPFP=0.3418 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,232B, BPFP=0.3488 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,004B, BPFP=0.2413 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,124B, BPFP=0.3436 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 4,084B, BPFP=0.0492 ⌛️ [2/4] FRONTEND: Frontend time: 1.850s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 162, 128]) layer.0.v_cache: torch.Size([1, 8, 162, 128]) layer.1.k_cache: torch.Size([1, 8, 162, 128]) layer.1.v_cache: torch.Size([1, 8, 162, 128]) layer.2.k_cache: torch.Size([1, 8, 162, 128]) layer.2.v_cache: torch.Size([1, 8, 162, 128]) layer.3.k_cache: torch.Size([1, 8, 162, 128]) layer.3.v_cache: torch.Size([1, 8, 162, 128]) layer.4.k_cache: torch.Size([1, 8, 162, 128]) layer.4.v_cache: torch.Size([1, 8, 162, 128]) layer.4.output: torch.Size([1, 162, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.676s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 162, 128]) layer.0.v_cache: torch.Size([1, 8, 162, 128]) layer.1.k_cache: torch.Size([1, 8, 162, 128]) layer.1.v_cache: torch.Size([1, 8, 162, 128]) layer.2.k_cache: torch.Size([1, 8, 162, 128]) layer.2.v_cache: torch.Size([1, 8, 162, 128]) layer.3.k_cache: torch.Size([1, 8, 162, 128]) layer.3.v_cache: torch.Size([1, 8, 162, 128]) layer.4.k_cache: torch.Size([1, 8, 162, 128]) layer.4.v_cache: torch.Size([1, 8, 162, 128]) layer.4.output: torch.Size([1, 162, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02742675 28.46523859 layer.0.v_cache 0.00000027 0.00061768 layer.1.k_cache 0.00301362 3.63222739 layer.1.v_cache 0.00000083 0.00266029 layer.2.k_cache 0.00118087 1.45343319 layer.2.v_cache 0.00000111 0.00383282 layer.3.k_cache 0.00131722 1.78746428 layer.3.v_cache 0.00000209 0.00619925 layer.4.k_cache 0.00343243 3.39535409 layer.4.v_cache 0.00000314 0.01077265 layer.4.output 0.00017655 0.19276383 ------------------------------------------------------------------------------------- TOTAL 0.00264889 2.82348969 (elements=2,322,432) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2322432 Total Bytes 61640 BPFP 0.2123 bits/point EBPFP 0.4247 equivalent bits/point MSE 2.823490 ---------------------- -------------------------------------------------------- Time: 3.534s Load: 0.009s, Pack+Encode: 1.850s, Decode+Unpack: 1.676s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 162, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 162, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.8235 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample97-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample97-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample98-layer4-item1.zst (99/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample98-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 157, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.010s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 157, 128) Output shape: (1, 157, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.0.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.1.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.1.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.2.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.2.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.3.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.3.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.4.k_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.4.v_cache: torch.Size([1, 8, 157, 128]) -> torch.Size([1, 1, 157, 1024]) layer.4.output: torch.Size([1, 157, 4096]) -> torch.Size([1, 1, 157, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,020B, BPFP=0.0508 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,096B, BPFP=0.2536 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 4,804B, BPFP=0.2391 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 6,204B, BPFP=0.3087 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,760B, BPFP=0.2866 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,352B, BPFP=0.3161 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 6,296B, BPFP=0.3133 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 7,340B, BPFP=0.3652 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 4,456B, BPFP=0.2217 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 6,940B, BPFP=0.3453 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,920B, BPFP=0.0488 ⌛️ [2/4] FRONTEND: Frontend time: 1.910s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 157, 128]) layer.0.v_cache: torch.Size([1, 8, 157, 128]) layer.1.k_cache: torch.Size([1, 8, 157, 128]) layer.1.v_cache: torch.Size([1, 8, 157, 128]) layer.2.k_cache: torch.Size([1, 8, 157, 128]) layer.2.v_cache: torch.Size([1, 8, 157, 128]) layer.3.k_cache: torch.Size([1, 8, 157, 128]) layer.3.v_cache: torch.Size([1, 8, 157, 128]) layer.4.k_cache: torch.Size([1, 8, 157, 128]) layer.4.v_cache: torch.Size([1, 8, 157, 128]) layer.4.output: torch.Size([1, 157, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.459s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 157, 128]) layer.0.v_cache: torch.Size([1, 8, 157, 128]) layer.1.k_cache: torch.Size([1, 8, 157, 128]) layer.1.v_cache: torch.Size([1, 8, 157, 128]) layer.2.k_cache: torch.Size([1, 8, 157, 128]) layer.2.v_cache: torch.Size([1, 8, 157, 128]) layer.3.k_cache: torch.Size([1, 8, 157, 128]) layer.3.v_cache: torch.Size([1, 8, 157, 128]) layer.4.k_cache: torch.Size([1, 8, 157, 128]) layer.4.v_cache: torch.Size([1, 8, 157, 128]) layer.4.output: torch.Size([1, 157, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02749603 29.80754255 layer.0.v_cache 0.00000027 0.00061403 layer.1.k_cache 0.00316339 3.56130768 layer.1.v_cache 0.00000087 0.00257100 layer.2.k_cache 0.00115124 1.47347478 layer.2.v_cache 0.00000113 0.00388760 layer.3.k_cache 0.00133258 1.76104717 layer.3.v_cache 0.00000205 0.00623441 layer.4.k_cache 0.00350615 3.23501548 layer.4.v_cache 0.00000300 0.01037464 layer.4.output 0.00014089 0.18619059 ------------------------------------------------------------------------------------- TOTAL 0.00265859 2.90048798 (elements=2,250,752) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2250752 Total Bytes 58188 BPFP 0.2068 bits/point EBPFP 0.4136 equivalent bits/point MSE 2.900488 ---------------------- -------------------------------------------------------- Time: 3.380s Load: 0.010s, Pack+Encode: 1.910s, Decode+Unpack: 1.459s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 157, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 157, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.9005 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample98-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample98-layer4-item1.zst 💪 Processing: ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample99-layer4-item1.zst (100/100) [1/4] FRONTEND: Loading features from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample99-layer4-item1.zst... Original data structure: root: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 165, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu ⌛️ [1/4] FRONTEND: Load time: 0.009s ------------------------------------------------------------ Qwen Features Summary ------------------------------------------------------------ Number of layers: 5 Layer indices: [0, 1, 2, 3, 4] Last layer index: 4 Cache shape: (1, 8, 165, 128) Output shape: (1, 165, 4096) Data type: torch.bfloat16 ------------------------------------------------------------ [2/4] FRONTEND: Pack + Encode (strategy: individual)... IndividualPacker: layer.0.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.0.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.1.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.1.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.2.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.2.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.3.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.3.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.4.k_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.4.v_cache: torch.Size([1, 8, 165, 128]) -> torch.Size([1, 1, 165, 1024]) layer.4.output: torch.Size([1, 165, 4096]) -> torch.Size([1, 1, 165, 4096]) Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache layer.0.k_cache: 1,016B, BPFP=0.0481 Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache layer.0.v_cache: 5,624B, BPFP=0.2663 Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache layer.1.k_cache: 5,396B, BPFP=0.2555 Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache layer.1.v_cache: 5,984B, BPFP=0.2833 Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache layer.2.k_cache: 5,660B, BPFP=0.2680 Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache layer.2.v_cache: 6,556B, BPFP=0.3104 Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache layer.3.k_cache: 7,304B, BPFP=0.3458 Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache layer.3.v_cache: 6,756B, BPFP=0.3199 Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache layer.4.k_cache: 5,116B, BPFP=0.2422 Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache layer.4.v_cache: 7,008B, BPFP=0.3318 Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output layer.4.output: 3,620B, BPFP=0.0429 ⌛️ [2/4] FRONTEND: Frontend time: 1.938s (Pack+Encode) [3/4] BACKEND: Decode + Unpack... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 165, 128]) layer.0.v_cache: torch.Size([1, 8, 165, 128]) layer.1.k_cache: torch.Size([1, 8, 165, 128]) layer.1.v_cache: torch.Size([1, 8, 165, 128]) layer.2.k_cache: torch.Size([1, 8, 165, 128]) layer.2.v_cache: torch.Size([1, 8, 165, 128]) layer.3.k_cache: torch.Size([1, 8, 165, 128]) layer.3.v_cache: torch.Size([1, 8, 165, 128]) layer.4.k_cache: torch.Size([1, 8, 165, 128]) layer.4.v_cache: torch.Size([1, 8, 165, 128]) layer.4.output: torch.Size([1, 165, 4096]) ⌛️ [3/4] BACKEND: Backend time: 1.587s [4/4] METRICS: Computing MSE Breakdown... Using per-key quantization points (layer.0.k_cache: torch.Size([256])) for layer.0.k_cache Using per-key quantization points (layer.0.v_cache: torch.Size([256])) for layer.0.v_cache Using per-key quantization points (layer.1.k_cache: torch.Size([256])) for layer.1.k_cache Using per-key quantization points (layer.1.v_cache: torch.Size([256])) for layer.1.v_cache Using per-key quantization points (layer.2.k_cache: torch.Size([256])) for layer.2.k_cache Using per-key quantization points (layer.2.v_cache: torch.Size([256])) for layer.2.v_cache Using per-key quantization points (layer.3.k_cache: torch.Size([256])) for layer.3.k_cache Using per-key quantization points (layer.3.v_cache: torch.Size([256])) for layer.3.v_cache Using per-key quantization points (layer.4.k_cache: torch.Size([256])) for layer.4.k_cache Using per-key quantization points (layer.4.v_cache: torch.Size([256])) for layer.4.v_cache Using per-key quantization points (layer.4.output: torch.Size([256])) for layer.4.output IndividualUnPacker: layer.0.k_cache: torch.Size([1, 8, 165, 128]) layer.0.v_cache: torch.Size([1, 8, 165, 128]) layer.1.k_cache: torch.Size([1, 8, 165, 128]) layer.1.v_cache: torch.Size([1, 8, 165, 128]) layer.2.k_cache: torch.Size([1, 8, 165, 128]) layer.2.v_cache: torch.Size([1, 8, 165, 128]) layer.3.k_cache: torch.Size([1, 8, 165, 128]) layer.3.v_cache: torch.Size([1, 8, 165, 128]) layer.4.k_cache: torch.Size([1, 8, 165, 128]) layer.4.v_cache: torch.Size([1, 8, 165, 128]) layer.4.output: torch.Size([1, 165, 4096]) Per-key MSE Breakdown (Quant=Truncation+Quantization, Codec=Compression): Key Quant-MSE Total-MSE ------------------------------------------------------------------------------------- layer.0.k_cache 0.02734190 25.75327296 layer.0.v_cache 0.00000026 0.00061128 layer.1.k_cache 0.00308234 3.45133353 layer.1.v_cache 0.00000078 0.00251190 layer.2.k_cache 0.00115028 1.45061202 layer.2.v_cache 0.00000111 0.00379122 layer.3.k_cache 0.00133313 1.76425337 layer.3.v_cache 0.00000202 0.00601228 layer.4.k_cache 0.00348877 3.41703473 layer.4.v_cache 0.00000299 0.01014055 layer.4.output 0.00017595 0.18001833 ------------------------------------------------------------------------------------- TOTAL 0.00265053 2.61283194 (elements=2,365,440) ---------------------- -------------------------------------------------------- SAMPLE-WISE STATISTICS ---------------------- -------------------------------------------------------- Handler qwen Strategy individual Architecture elic-featurecoding ---------------------- -------------------------------------------------------- Total Elements 2365440 Total Bytes 60040 BPFP 0.2031 bits/point EBPFP 0.4061 equivalent bits/point MSE 2.612832 ---------------------- -------------------------------------------------------- Time: 3.533s Load: 0.009s, Pack+Encode: 1.938s, Decode+Unpack: 1.587s ---------------------- -------------------------------------------------------- Restored Feature Format: [dict] with 3 keys key['output']: [Tensor] shape=torch.Size([1, 165, 4096]), dtype=torch.bfloat16, device=cpu key['key']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu key['value']: [list] with 5 items item[0]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[1]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[2]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[3]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu item[4]: [Tensor] shape=torch.Size([1, 8, 165, 128]), dtype=torch.bfloat16, device=cpu 💾 Converting with 2.6128 MSE: from ../datasets/Qwen3-500features-L5wCache/Qwen3-500features-L5wCache/qwen/qwen3-8b/fc_arc_challenge/sample99-layer4-item1.zst to output-fixed/qwen/lambda0.001/elic-featurecoding-8bit-individual/fc_arc_challenge/sample99-layer4-item1.zst ------------------------ ---------------------------- TOTAL PROCESSING SUMMARY ------------------------ ---------------------------- Total files 100 Avg BPFP 0.1995 bits/point Avg EBPFP 0.3990 equivalent bits/point Avg MSE 2.724538 Avg Time 3.608s ------------------------ ----------------------------