SPB Git

spb/wp30_uqo

Public
4.1 KB · 199 lines json
1{2  "config": {3    "sender": "qwen1b5",4    "receiver": "qwen3b",5    "families": [6      "synthetic",7      "exec"8    ],9    "max_tokens_per_doc": 8192,10    "ridge": 1.0,11    "n_dev_eval": 412  },13  "train_docs": [14    "syn_01000_2000",15    "syn_01001_8000",16    "syn_01002_16000",17    "syn_01003_2000",18    "syn_01004_8000",19    "syn_01005_16000",20    "syn_01006_2000",21    "syn_01007_8000",22    "syn_01008_16000",23    "syn_01009_2000",24    "syn_01010_8000",25    "syn_01011_16000",26    "syn_01012_2000",27    "syn_01013_8000",28    "syn_01014_16000",29    "syn_01015_2000",30    "syn_01016_8000",31    "syn_01017_16000",32    "syn_01018_2000",33    "syn_01019_8000",34    "syn_01020_16000",35    "syn_01021_2000",36    "syn_01022_8000",37    "syn_01023_16000",38    "syn_01024_2000",39    "syn_01025_8000",40    "syn_01026_16000",41    "syn_01027_2000",42    "syn_01028_8000",43    "syn_01029_16000",44    "syn_01030_2000",45    "syn_01031_8000",46    "syn_01032_16000",47    "syn_01033_2000",48    "syn_01034_8000",49    "syn_01035_16000",50    "syn_01036_2000",51    "syn_01037_8000",52    "syn_01038_16000",53    "syn_01039_2000",54    "exec_00001_2000",55    "exec_00002_2000",56    "exec_00003_2000",57    "exec_00004_2000",58    "exec_00005_2000",59    "exec_00006_2000",60    "exec_00007_2000",61    "exec_00008_2000",62    "exec_00009_2000",63    "exec_00010_2000",64    "exec_00011_2000",65    "exec_00012_2000",66    "exec_00013_2000",67    "exec_00014_2000",68    "exec_00015_2000",69    "exec_00016_2000",70    "exec_00017_2000",71    "exec_00018_2000",72    "exec_00019_2000",73    "exec_00020_2000"74  ],75  "train_seconds": 288.5171139240265,76  "dev_reconstruction": [77    {78      "doc": "syn_02000_2000",79      "layer": 0,80      "rel_err_k": 0.014429992996156216,81      "rel_err_v": 0.2174794822931289782    },83    {84      "doc": "syn_02000_2000",85      "layer": 18,86      "rel_err_k": 0.19839154183864594,87      "rel_err_v": 0.588057160377502488    },89    {90      "doc": "syn_02000_2000",91      "layer": 35,92      "rel_err_k": 0.19203418493270874,93      "rel_err_v": 0.548503696918487594    },95    {96      "doc": "syn_02001_8000",97      "layer": 0,98      "rel_err_k": 0.008613131940364838,99      "rel_err_v": 0.12830892205238342100    },101    {102      "doc": "syn_02001_8000",103      "layer": 18,104      "rel_err_k": 0.17894120514392853,105      "rel_err_v": 0.5188227891921997106    },107    {108      "doc": "syn_02001_8000",109      "layer": 35,110      "rel_err_k": 0.13947930932044983,111      "rel_err_v": 0.4167284369468689112    },113    {114      "doc": "syn_02002_16000",115      "layer": 0,116      "rel_err_k": 0.008165479637682438,117      "rel_err_v": 0.12064959853887558118    },119    {120      "doc": "syn_02002_16000",121      "layer": 18,122      "rel_err_k": 0.17874450981616974,123      "rel_err_v": 0.5191535353660583124    },125    {126      "doc": "syn_02002_16000",127      "layer": 35,128      "rel_err_k": 0.12985366582870483,129      "rel_err_v": 0.38665786385536194130    },131    {132      "doc": "syn_02003_2000",133      "layer": 0,134      "rel_err_k": 0.012657775543630123,135      "rel_err_v": 0.1916254460811615136    },137    {138      "doc": "syn_02003_2000",139      "layer": 18,140      "rel_err_k": 0.19114117324352264,141      "rel_err_v": 0.5788665413856506142    },143    {144      "doc": "syn_02003_2000",145      "layer": 35,146      "rel_err_k": 0.18840292096138,147      "rel_err_v": 0.5358448624610901148    }149  ],150  "device": "laptop-m5max",151  "sender": "mlx-community/Qwen2.5-1.5B-Instruct-bf16",152  "sender_revision": "4ae77cb209f06199b8df1c94e21ff341332a3a89",153  "receiver": "mlx-community/Qwen2.5-3B-Instruct-bf16",154  "receiver_revision": "bec1ad658e380044bbfe06d4a2268801d3379a11",155  "layer_map": [156    0,157    1,158    2,159    2,160    3,161    4,162    5,163    5,164    6,165    7,166    8,167    8,168    9,169    10,170    11,171    12,172    12,173    13,174    14,175    15,176    15,177    16,178    17,179    18,180    19,181    19,182    20,183    21,184    22,185    22,186    23,187    24,188    25,189    25,190    26,191    27192  ],193  "ridge": 1.0,194  "n_train_tokens": 248569,195  "params": 9455616,196  "checkpoint_sha256": "3802150edf61206f81a5af7f5294981242387bcd2f10291beeac5f93d4340d92",197  "created_utc": "2026-10-30T04:08:47.078307Z",198  "loss": "ridge least squares on un-rotated K and V (reconstruction only)"199}
Files · d5e182fa7d7d5118aa92e61c47c40ed848da0ecf
Menu

Theme