SPB Git

spb/wp30_uqo

Public
4.2 KB · 202 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": 10.0,11    "neighbours": 1,12    "n_dev_eval": 4,13    "output": "mapper_v2_qwen1b5_to_qwen3b.npz"14  },15  "train_docs": [16    "syn_01000_2000",17    "syn_01001_8000",18    "syn_01002_16000",19    "syn_01003_2000",20    "syn_01004_8000",21    "syn_01005_16000",22    "syn_01006_2000",23    "syn_01007_8000",24    "syn_01008_16000",25    "syn_01009_2000",26    "syn_01010_8000",27    "syn_01011_16000",28    "syn_01012_2000",29    "syn_01013_8000",30    "syn_01014_16000",31    "syn_01015_2000",32    "syn_01016_8000",33    "syn_01017_16000",34    "syn_01018_2000",35    "syn_01019_8000",36    "syn_01020_16000",37    "syn_01021_2000",38    "syn_01022_8000",39    "syn_01023_16000",40    "syn_01024_2000",41    "syn_01025_8000",42    "syn_01026_16000",43    "syn_01027_2000",44    "syn_01028_8000",45    "syn_01029_16000",46    "syn_01030_2000",47    "syn_01031_8000",48    "syn_01032_16000",49    "syn_01033_2000",50    "syn_01034_8000",51    "syn_01035_16000",52    "syn_01036_2000",53    "syn_01037_8000",54    "syn_01038_16000",55    "syn_01039_2000",56    "exec_00001_2000",57    "exec_00002_2000",58    "exec_00003_2000",59    "exec_00004_2000",60    "exec_00005_2000",61    "exec_00006_2000",62    "exec_00007_2000",63    "exec_00008_2000",64    "exec_00009_2000",65    "exec_00010_2000",66    "exec_00011_2000",67    "exec_00012_2000",68    "exec_00013_2000",69    "exec_00014_2000",70    "exec_00015_2000",71    "exec_00016_2000",72    "exec_00017_2000",73    "exec_00018_2000",74    "exec_00019_2000",75    "exec_00020_2000"76  ],77  "train_seconds": 483.06637501716614,78  "dev_reconstruction": [79    {80      "doc": "syn_02000_2000",81      "layer": 0,82      "rel_err_k": 0.0117700370028615,83      "rel_err_v": 0.1769662797451019384    },85    {86      "doc": "syn_02000_2000",87      "layer": 18,88      "rel_err_k": 0.1631951928138733,89      "rel_err_v": 0.4864163696765899790    },91    {92      "doc": "syn_02000_2000",93      "layer": 35,94      "rel_err_k": 0.1578485369682312,95      "rel_err_v": 0.449933677911758496    },97    {98      "doc": "syn_02001_8000",99      "layer": 0,100      "rel_err_k": 0.006607668939977884,101      "rel_err_v": 0.09916035085916519102    },103    {104      "doc": "syn_02001_8000",105      "layer": 18,106      "rel_err_k": 0.1464068740606308,107      "rel_err_v": 0.42671793699264526108    },109    {110      "doc": "syn_02001_8000",111      "layer": 35,112      "rel_err_k": 0.11391212046146393,113      "rel_err_v": 0.33776769042015076114    },115    {116      "doc": "syn_02002_16000",117      "layer": 0,118      "rel_err_k": 0.0067402757704257965,119      "rel_err_v": 0.0981205627322197120    },121    {122      "doc": "syn_02002_16000",123      "layer": 18,124      "rel_err_k": 0.14760245382785797,125      "rel_err_v": 0.4306700825691223126    },127    {128      "doc": "syn_02002_16000",129      "layer": 35,130      "rel_err_k": 0.1058705523610115,131      "rel_err_v": 0.31111809611320496132    },133    {134      "doc": "syn_02003_2000",135      "layer": 0,136      "rel_err_k": 0.009620909579098225,137      "rel_err_v": 0.14274893701076508138    },139    {140      "doc": "syn_02003_2000",141      "layer": 18,142      "rel_err_k": 0.15372033417224884,143      "rel_err_v": 0.4719652533531189144    },145    {146      "doc": "syn_02003_2000",147      "layer": 35,148      "rel_err_k": 0.15298303961753845,149      "rel_err_v": 0.43625742197036743150    }151  ],152  "device": "laptop-m5max",153  "sender": "mlx-community/Qwen2.5-1.5B-Instruct-bf16",154  "sender_revision": "4ae77cb209f06199b8df1c94e21ff341332a3a89",155  "receiver": "mlx-community/Qwen2.5-3B-Instruct-bf16",156  "receiver_revision": "bec1ad658e380044bbfe06d4a2268801d3379a11",157  "layer_map": [158    0,159    1,160    2,161    2,162    3,163    4,164    5,165    5,166    6,167    7,168    8,169    8,170    9,171    10,172    11,173    12,174    12,175    13,176    14,177    15,178    15,179    16,180    17,181    18,182    19,183    19,184    20,185    21,186    22,187    22,188    23,189    24,190    25,191    25,192    26,193    27194  ],195  "ridge": 10.0,196  "n_train_tokens": 248569,197  "params": 28329984,198  "neighbours": 1,199  "checkpoint_sha256": "3147cd7ba42e0bf00a590a1cfadaa53fb0d190623af5864e8b3a92771a288921",200  "created_utc": "2026-10-30T04:19:15.145448Z",201  "loss": "ridge least squares on un-rotated K and V (reconstruction only)"202}
Files · 2ccc6b7e52385b2438c0f3e62e88226009fb4f9f
Menu

Theme