Tele-AI/telechat-7B

1{2  "metadata": {3    "total_size": 144000368644  },5  "weight_map": {6    "h.0.input_layernorm.weight": "pytorch_model_00002-of-00032.bin",7    "h.0.mlp.down_proj.bias": "pytorch_model_00002-of-00032.bin",8    "h.0.mlp.down_proj.weight": "pytorch_model_00002-of-00032.bin",9    "h.0.mlp.gate_proj.weight": "pytorch_model_00002-of-00032.bin",10    "h.0.mlp.up_proj.weight": "pytorch_model_00002-of-00032.bin",11    "h.0.post_attention_layernorm.weight": "pytorch_model_00002-of-00032.bin",12    "h.0.self_attention.dense.bias": "pytorch_model_00002-of-00032.bin",13    "h.0.self_attention.dense.weight": "pytorch_model_00002-of-00032.bin",14    "h.0.self_attention.key_value.weight": "pytorch_model_00002-of-00032.bin",15    "h.0.self_attention.query.weight": "pytorch_model_00002-of-00032.bin",16    "h.1.input_layernorm.weight": "pytorch_model_00003-of-00032.bin",17    "h.1.mlp.down_proj.bias": "pytorch_model_00003-of-00032.bin",18    "h.1.mlp.down_proj.weight": "pytorch_model_00003-of-00032.bin",19    "h.1.mlp.gate_proj.weight": "pytorch_model_00003-of-00032.bin",20    "h.1.mlp.up_proj.weight": "pytorch_model_00003-of-00032.bin",21    "h.1.post_attention_layernorm.weight": "pytorch_model_00003-of-00032.bin",22    "h.1.self_attention.dense.bias": "pytorch_model_00003-of-00032.bin",23    "h.1.self_attention.dense.weight": "pytorch_model_00003-of-00032.bin",24    "h.1.self_attention.key_value.weight": "pytorch_model_00003-of-00032.bin",25    "h.1.self_attention.query.weight": "pytorch_model_00003-of-00032.bin",26    "h.10.input_layernorm.weight": "pytorch_model_00012-of-00032.bin",27    "h.10.mlp.down_proj.bias": "pytorch_model_00012-of-00032.bin",28    "h.10.mlp.down_proj.weight": "pytorch_model_00012-of-00032.bin",29    "h.10.mlp.gate_proj.weight": "pytorch_model_00012-of-00032.bin",30    "h.10.mlp.up_proj.weight": "pytorch_model_00012-of-00032.bin",31    "h.10.post_attention_layernorm.weight": "pytorch_model_00012-of-00032.bin",32    "h.10.self_attention.dense.bias": "pytorch_model_00012-of-00032.bin",33    "h.10.self_attention.dense.weight": "pytorch_model_00012-of-00032.bin",34    "h.10.self_attention.key_value.weight": "pytorch_model_00012-of-00032.bin",35    "h.10.self_attention.query.weight": "pytorch_model_00012-of-00032.bin",36    "h.11.input_layernorm.weight": "pytorch_model_00013-of-00032.bin",37    "h.11.mlp.down_proj.bias": "pytorch_model_00013-of-00032.bin",38    "h.11.mlp.down_proj.weight": "pytorch_model_00013-of-00032.bin",39    "h.11.mlp.gate_proj.weight": "pytorch_model_00013-of-00032.bin",40    "h.11.mlp.up_proj.weight": "pytorch_model_00013-of-00032.bin",41    "h.11.post_attention_layernorm.weight": "pytorch_model_00013-of-00032.bin",42    "h.11.self_attention.dense.bias": "pytorch_model_00013-of-00032.bin",43    "h.11.self_attention.dense.weight": "pytorch_model_00013-of-00032.bin",44    "h.11.self_attention.key_value.weight": "pytorch_model_00013-of-00032.bin",45    "h.11.self_attention.query.weight": "pytorch_model_00013-of-00032.bin",46    "h.12.input_layernorm.weight": "pytorch_model_00014-of-00032.bin",47    "h.12.mlp.down_proj.bias": "pytorch_model_00014-of-00032.bin",48    "h.12.mlp.down_proj.weight": "pytorch_model_00014-of-00032.bin",49    "h.12.mlp.gate_proj.weight": "pytorch_model_00014-of-00032.bin",50    "h.12.mlp.up_proj.weight": "pytorch_model_00014-of-00032.bin",51    "h.12.post_attention_layernorm.weight": "pytorch_model_00014-of-00032.bin",52    "h.12.self_attention.dense.bias": "pytorch_model_00014-of-00032.bin",53    "h.12.self_attention.dense.weight": "pytorch_model_00014-of-00032.bin",54    "h.12.self_attention.key_value.weight": "pytorch_model_00014-of-00032.bin",55    "h.12.self_attention.query.weight": "pytorch_model_00014-of-00032.bin",56    "h.13.input_layernorm.weight": "pytorch_model_00015-of-00032.bin",57    "h.13.mlp.down_proj.bias": "pytorch_model_00015-of-00032.bin",58    "h.13.mlp.down_proj.weight": "pytorch_model_00015-of-00032.bin",59    "h.13.mlp.gate_proj.weight": "pytorch_model_00015-of-00032.bin",60    "h.13.mlp.up_proj.weight": "pytorch_model_00015-of-00032.bin",61    "h.13.post_attention_layernorm.weight": "pytorch_model_00015-of-00032.bin",62    "h.13.self_attention.dense.bias": "pytorch_model_00015-of-00032.bin",63    "h.13.self_attention.dense.weight": "pytorch_model_00015-of-00032.bin",64    "h.13.self_attention.key_value.weight": "pytorch_model_00015-of-00032.bin",65    "h.13.self_attention.query.weight": "pytorch_model_00015-of-00032.bin",66    "h.14.input_layernorm.weight": "pytorch_model_00016-of-00032.bin",67    "h.14.mlp.down_proj.bias": "pytorch_model_00016-of-00032.bin",68    "h.14.mlp.down_proj.weight": "pytorch_model_00016-of-00032.bin",69    "h.14.mlp.gate_proj.weight": "pytorch_model_00016-of-00032.bin",70    "h.14.mlp.up_proj.weight": "pytorch_model_00016-of-00032.bin",71    "h.14.post_attention_layernorm.weight": "pytorch_model_00016-of-00032.bin",72    "h.14.self_attention.dense.bias": "pytorch_model_00016-of-00032.bin",73    "h.14.self_attention.dense.weight": "pytorch_model_00016-of-00032.bin",74    "h.14.self_attention.key_value.weight": "pytorch_model_00016-of-00032.bin",75    "h.14.self_attention.query.weight": "pytorch_model_00016-of-00032.bin",76    "h.15.input_layernorm.weight": "pytorch_model_00017-of-00032.bin",77    "h.15.mlp.down_proj.bias": "pytorch_model_00017-of-00032.bin",78    "h.15.mlp.down_proj.weight": "pytorch_model_00017-of-00032.bin",79    "h.15.mlp.gate_proj.weight": "pytorch_model_00017-of-00032.bin",80    "h.15.mlp.up_proj.weight": "pytorch_model_00017-of-00032.bin",81    "h.15.post_attention_layernorm.weight": "pytorch_model_00017-of-00032.bin",82    "h.15.self_attention.dense.bias": "pytorch_model_00017-of-00032.bin",83    "h.15.self_attention.dense.weight": "pytorch_model_00017-of-00032.bin",84    "h.15.self_attention.key_value.weight": "pytorch_model_00017-of-00032.bin",85    "h.15.self_attention.query.weight": "pytorch_model_00017-of-00032.bin",86    "h.16.input_layernorm.weight": "pytorch_model_00018-of-00032.bin",87    "h.16.mlp.down_proj.bias": "pytorch_model_00018-of-00032.bin",88    "h.16.mlp.down_proj.weight": "pytorch_model_00018-of-00032.bin",89    "h.16.mlp.gate_proj.weight": "pytorch_model_00018-of-00032.bin",90    "h.16.mlp.up_proj.weight": "pytorch_model_00018-of-00032.bin",91    "h.16.post_attention_layernorm.weight": "pytorch_model_00018-of-00032.bin",92    "h.16.self_attention.dense.bias": "pytorch_model_00018-of-00032.bin",93    "h.16.self_attention.dense.weight": "pytorch_model_00018-of-00032.bin",94    "h.16.self_attention.key_value.weight": "pytorch_model_00018-of-00032.bin",95    "h.16.self_attention.query.weight": "pytorch_model_00018-of-00032.bin",96    "h.17.input_layernorm.weight": "pytorch_model_00019-of-00032.bin",97    "h.17.mlp.down_proj.bias": "pytorch_model_00019-of-00032.bin",98    "h.17.mlp.down_proj.weight": "pytorch_model_00019-of-00032.bin",99    "h.17.mlp.gate_proj.weight": "pytorch_model_00019-of-00032.bin",100    "h.17.mlp.up_proj.weight": "pytorch_model_00019-of-00032.bin",101    "h.17.post_attention_layernorm.weight": "pytorch_model_00019-of-00032.bin",102    "h.17.self_attention.dense.bias": "pytorch_model_00019-of-00032.bin",103    "h.17.self_attention.dense.weight": "pytorch_model_00019-of-00032.bin",104    "h.17.self_attention.key_value.weight": "pytorch_model_00019-of-00032.bin",105    "h.17.self_attention.query.weight": "pytorch_model_00019-of-00032.bin",106    "h.18.input_layernorm.weight": "pytorch_model_00020-of-00032.bin",107    "h.18.mlp.down_proj.bias": "pytorch_model_00020-of-00032.bin",108    "h.18.mlp.down_proj.weight": "pytorch_model_00020-of-00032.bin",109    "h.18.mlp.gate_proj.weight": "pytorch_model_00020-of-00032.bin",110    "h.18.mlp.up_proj.weight": "pytorch_model_00020-of-00032.bin",111    "h.18.post_attention_layernorm.weight": "pytorch_model_00020-of-00032.bin",112    "h.18.self_attention.dense.bias": "pytorch_model_00020-of-00032.bin",113    "h.18.self_attention.dense.weight": "pytorch_model_00020-of-00032.bin",114    "h.18.self_attention.key_value.weight": "pytorch_model_00020-of-00032.bin",115    "h.18.self_attention.query.weight": "pytorch_model_00020-of-00032.bin",116    "h.19.input_layernorm.weight": "pytorch_model_00021-of-00032.bin",117    "h.19.mlp.down_proj.bias": "pytorch_model_00021-of-00032.bin",118    "h.19.mlp.down_proj.weight": "pytorch_model_00021-of-00032.bin",119    "h.19.mlp.gate_proj.weight": "pytorch_model_00021-of-00032.bin",120    "h.19.mlp.up_proj.weight": "pytorch_model_00021-of-00032.bin",121    "h.19.post_attention_layernorm.weight": "pytorch_model_00021-of-00032.bin",122    "h.19.self_attention.dense.bias": "pytorch_model_00021-of-00032.bin",123    "h.19.self_attention.dense.weight": "pytorch_model_00021-of-00032.bin",124    "h.19.self_attention.key_value.weight": "pytorch_model_00021-of-00032.bin",125    "h.19.self_attention.query.weight": "pytorch_model_00021-of-00032.bin",126    "h.2.input_layernorm.weight": "pytorch_model_00004-of-00032.bin",127    "h.2.mlp.down_proj.bias": "pytorch_model_00004-of-00032.bin",128    "h.2.mlp.down_proj.weight": "pytorch_model_00004-of-00032.bin",129    "h.2.mlp.gate_proj.weight": "pytorch_model_00004-of-00032.bin",130    "h.2.mlp.up_proj.weight": "pytorch_model_00004-of-00032.bin",131    "h.2.post_attention_layernorm.weight": "pytorch_model_00004-of-00032.bin",132    "h.2.self_attention.dense.bias": "pytorch_model_00004-of-00032.bin",133    "h.2.self_attention.dense.weight": "pytorch_model_00004-of-00032.bin",134    "h.2.self_attention.key_value.weight": "pytorch_model_00004-of-00032.bin",135    "h.2.self_attention.query.weight": "pytorch_model_00004-of-00032.bin",136    "h.20.input_layernorm.weight": "pytorch_model_00022-of-00032.bin",137    "h.20.mlp.down_proj.bias": "pytorch_model_00022-of-00032.bin",138    "h.20.mlp.down_proj.weight": "pytorch_model_00022-of-00032.bin",139    "h.20.mlp.gate_proj.weight": "pytorch_model_00022-of-00032.bin",140    "h.20.mlp.up_proj.weight": "pytorch_model_00022-of-00032.bin",141    "h.20.post_attention_layernorm.weight": "pytorch_model_00022-of-00032.bin",142    "h.20.self_attention.dense.bias": "pytorch_model_00022-of-00032.bin",143    "h.20.self_attention.dense.weight": "pytorch_model_00022-of-00032.bin",144    "h.20.self_attention.key_value.weight": "pytorch_model_00022-of-00032.bin",145    "h.20.self_attention.query.weight": "pytorch_model_00022-of-00032.bin",146    "h.21.input_layernorm.weight": "pytorch_model_00023-of-00032.bin",147    "h.21.mlp.down_proj.bias": "pytorch_model_00023-of-00032.bin",148    "h.21.mlp.down_proj.weight": "pytorch_model_00023-of-00032.bin",149    "h.21.mlp.gate_proj.weight": "pytorch_model_00023-of-00032.bin",150    "h.21.mlp.up_proj.weight": "pytorch_model_00023-of-00032.bin",151    "h.21.post_attention_layernorm.weight": "pytorch_model_00023-of-00032.bin",152    "h.21.self_attention.dense.bias": "pytorch_model_00023-of-00032.bin",153    "h.21.self_attention.dense.weight": "pytorch_model_00023-of-00032.bin",154    "h.21.self_attention.key_value.weight": "pytorch_model_00023-of-00032.bin",155    "h.21.self_attention.query.weight": "pytorch_model_00023-of-00032.bin",156    "h.22.input_layernorm.weight": "pytorch_model_00024-of-00032.bin",157    "h.22.mlp.down_proj.bias": "pytorch_model_00024-of-00032.bin",158    "h.22.mlp.down_proj.weight": "pytorch_model_00024-of-00032.bin",159    "h.22.mlp.gate_proj.weight": "pytorch_model_00024-of-00032.bin",160    "h.22.mlp.up_proj.weight": "pytorch_model_00024-of-00032.bin",161    "h.22.post_attention_layernorm.weight": "pytorch_model_00024-of-00032.bin",162    "h.22.self_attention.dense.bias": "pytorch_model_00024-of-00032.bin",163    "h.22.self_attention.dense.weight": "pytorch_model_00024-of-00032.bin",164    "h.22.self_attention.key_value.weight": "pytorch_model_00024-of-00032.bin",165    "h.22.self_attention.query.weight": "pytorch_model_00024-of-00032.bin",166    "h.23.input_layernorm.weight": "pytorch_model_00025-of-00032.bin",167    "h.23.mlp.down_proj.bias": "pytorch_model_00025-of-00032.bin",168    "h.23.mlp.down_proj.weight": "pytorch_model_00025-of-00032.bin",169    "h.23.mlp.gate_proj.weight": "pytorch_model_00025-of-00032.bin",170    "h.23.mlp.up_proj.weight": "pytorch_model_00025-of-00032.bin",171    "h.23.post_attention_layernorm.weight": "pytorch_model_00025-of-00032.bin",172    "h.23.self_attention.dense.bias": "pytorch_model_00025-of-00032.bin",173    "h.23.self_attention.dense.weight": "pytorch_model_00025-of-00032.bin",174    "h.23.self_attention.key_value.weight": "pytorch_model_00025-of-00032.bin",175    "h.23.self_attention.query.weight": "pytorch_model_00025-of-00032.bin",176    "h.24.input_layernorm.weight": "pytorch_model_00026-of-00032.bin",177    "h.24.mlp.down_proj.bias": "pytorch_model_00026-of-00032.bin",178    "h.24.mlp.down_proj.weight": "pytorch_model_00026-of-00032.bin",179    "h.24.mlp.gate_proj.weight": "pytorch_model_00026-of-00032.bin",180    "h.24.mlp.up_proj.weight": "pytorch_model_00026-of-00032.bin",181    "h.24.post_attention_layernorm.weight": "pytorch_model_00026-of-00032.bin",182    "h.24.self_attention.dense.bias": "pytorch_model_00026-of-00032.bin",183    "h.24.self_attention.dense.weight": "pytorch_model_00026-of-00032.bin",184    "h.24.self_attention.key_value.weight": "pytorch_model_00026-of-00032.bin",185    "h.24.self_attention.query.weight": "pytorch_model_00026-of-00032.bin",186    "h.25.input_layernorm.weight": "pytorch_model_00027-of-00032.bin",187    "h.25.mlp.down_proj.bias": "pytorch_model_00027-of-00032.bin",188    "h.25.mlp.down_proj.weight": "pytorch_model_00027-of-00032.bin",189    "h.25.mlp.gate_proj.weight": "pytorch_model_00027-of-00032.bin",190    "h.25.mlp.up_proj.weight": "pytorch_model_00027-of-00032.bin",191    "h.25.post_attention_layernorm.weight": "pytorch_model_00027-of-00032.bin",192    "h.25.self_attention.dense.bias": "pytorch_model_00027-of-00032.bin",193    "h.25.self_attention.dense.weight": "pytorch_model_00027-of-00032.bin",194    "h.25.self_attention.key_value.weight": "pytorch_model_00027-of-00032.bin",195    "h.25.self_attention.query.weight": "pytorch_model_00027-of-00032.bin",196    "h.26.input_layernorm.weight": "pytorch_model_00028-of-00032.bin",197    "h.26.mlp.down_proj.bias": "pytorch_model_00028-of-00032.bin",198    "h.26.mlp.down_proj.weight": "pytorch_model_00028-of-00032.bin",199    "h.26.mlp.gate_proj.weight": "pytorch_model_00028-of-00032.bin",200    "h.26.mlp.up_proj.weight": "pytorch_model_00028-of-00032.bin",201    "h.26.post_attention_layernorm.weight": "pytorch_model_00028-of-00032.bin",202    "h.26.self_attention.dense.bias": "pytorch_model_00028-of-00032.bin",203    "h.26.self_attention.dense.weight": "pytorch_model_00028-of-00032.bin",204    "h.26.self_attention.key_value.weight": "pytorch_model_00028-of-00032.bin",205    "h.26.self_attention.query.weight": "pytorch_model_00028-of-00032.bin",206    "h.27.input_layernorm.weight": "pytorch_model_00029-of-00032.bin",207    "h.27.mlp.down_proj.bias": "pytorch_model_00029-of-00032.bin",208    "h.27.mlp.down_proj.weight": "pytorch_model_00029-of-00032.bin",209    "h.27.mlp.gate_proj.weight": "pytorch_model_00029-of-00032.bin",210    "h.27.mlp.up_proj.weight": "pytorch_model_00029-of-00032.bin",211    "h.27.post_attention_layernorm.weight": "pytorch_model_00029-of-00032.bin",212    "h.27.self_attention.dense.bias": "pytorch_model_00029-of-00032.bin",213    "h.27.self_attention.dense.weight": "pytorch_model_00029-of-00032.bin",214    "h.27.self_attention.key_value.weight": "pytorch_model_00029-of-00032.bin",215    "h.27.self_attention.query.weight": "pytorch_model_00029-of-00032.bin",216    "h.28.input_layernorm.weight": "pytorch_model_00030-of-00032.bin",217    "h.28.mlp.down_proj.bias": "pytorch_model_00030-of-00032.bin",218    "h.28.mlp.down_proj.weight": "pytorch_model_00030-of-00032.bin",219    "h.28.mlp.gate_proj.weight": "pytorch_model_00030-of-00032.bin",220    "h.28.mlp.up_proj.weight": "pytorch_model_00030-of-00032.bin",221    "h.28.post_attention_layernorm.weight": "pytorch_model_00030-of-00032.bin",222    "h.28.self_attention.dense.bias": "pytorch_model_00030-of-00032.bin",223    "h.28.self_attention.dense.weight": "pytorch_model_00030-of-00032.bin",224    "h.28.self_attention.key_value.weight": "pytorch_model_00030-of-00032.bin",225    "h.28.self_attention.query.weight": "pytorch_model_00030-of-00032.bin",226    "h.29.input_layernorm.weight": "pytorch_model_00031-of-00032.bin",227    "h.29.mlp.down_proj.bias": "pytorch_model_00031-of-00032.bin",228    "h.29.mlp.down_proj.weight": "pytorch_model_00031-of-00032.bin",229    "h.29.mlp.gate_proj.weight": "pytorch_model_00031-of-00032.bin",230    "h.29.mlp.up_proj.weight": "pytorch_model_00031-of-00032.bin",231    "h.29.post_attention_layernorm.weight": "pytorch_model_00031-of-00032.bin",232    "h.29.self_attention.dense.bias": "pytorch_model_00031-of-00032.bin",233    "h.29.self_attention.dense.weight": "pytorch_model_00031-of-00032.bin",234    "h.29.self_attention.key_value.weight": "pytorch_model_00031-of-00032.bin",235    "h.29.self_attention.query.weight": "pytorch_model_00031-of-00032.bin",236    "h.3.input_layernorm.weight": "pytorch_model_00005-of-00032.bin",237    "h.3.mlp.down_proj.bias": "pytorch_model_00005-of-00032.bin",238    "h.3.mlp.down_proj.weight": "pytorch_model_00005-of-00032.bin",239    "h.3.mlp.gate_proj.weight": "pytorch_model_00005-of-00032.bin",240    "h.3.mlp.up_proj.weight": "pytorch_model_00005-of-00032.bin",241    "h.3.post_attention_layernorm.weight": "pytorch_model_00005-of-00032.bin",242    "h.3.self_attention.dense.bias": "pytorch_model_00005-of-00032.bin",243    "h.3.self_attention.dense.weight": "pytorch_model_00005-of-00032.bin",244    "h.3.self_attention.key_value.weight": "pytorch_model_00005-of-00032.bin",245    "h.3.self_attention.query.weight": "pytorch_model_00005-of-00032.bin",246    "h.4.input_layernorm.weight": "pytorch_model_00006-of-00032.bin",247    "h.4.mlp.down_proj.bias": "pytorch_model_00006-of-00032.bin",248    "h.4.mlp.down_proj.weight": "pytorch_model_00006-of-00032.bin",249    "h.4.mlp.gate_proj.weight": "pytorch_model_00006-of-00032.bin",250    "h.4.mlp.up_proj.weight": "pytorch_model_00006-of-00032.bin",251    "h.4.post_attention_layernorm.weight": "pytorch_model_00006-of-00032.bin",252    "h.4.self_attention.dense.bias": "pytorch_model_00006-of-00032.bin",253    "h.4.self_attention.dense.weight": "pytorch_model_00006-of-00032.bin",254    "h.4.self_attention.key_value.weight": "pytorch_model_00006-of-00032.bin",255    "h.4.self_attention.query.weight": "pytorch_model_00006-of-00032.bin",256    "h.5.input_layernorm.weight": "pytorch_model_00007-of-00032.bin",257    "h.5.mlp.down_proj.bias": "pytorch_model_00007-of-00032.bin",258    "h.5.mlp.down_proj.weight": "pytorch_model_00007-of-00032.bin",259    "h.5.mlp.gate_proj.weight": "pytorch_model_00007-of-00032.bin",260    "h.5.mlp.up_proj.weight": "pytorch_model_00007-of-00032.bin",261    "h.5.post_attention_layernorm.weight": "pytorch_model_00007-of-00032.bin",262    "h.5.self_attention.dense.bias": "pytorch_model_00007-of-00032.bin",263    "h.5.self_attention.dense.weight": "pytorch_model_00007-of-00032.bin",264    "h.5.self_attention.key_value.weight": "pytorch_model_00007-of-00032.bin",265    "h.5.self_attention.query.weight": "pytorch_model_00007-of-00032.bin",266    "h.6.input_layernorm.weight": "pytorch_model_00008-of-00032.bin",267    "h.6.mlp.down_proj.bias": "pytorch_model_00008-of-00032.bin",268    "h.6.mlp.down_proj.weight": "pytorch_model_00008-of-00032.bin",269    "h.6.mlp.gate_proj.weight": "pytorch_model_00008-of-00032.bin",270    "h.6.mlp.up_proj.weight": "pytorch_model_00008-of-00032.bin",271    "h.6.post_attention_layernorm.weight": "pytorch_model_00008-of-00032.bin",272    "h.6.self_attention.dense.bias": "pytorch_model_00008-of-00032.bin",273    "h.6.self_attention.dense.weight": "pytorch_model_00008-of-00032.bin",274    "h.6.self_attention.key_value.weight": "pytorch_model_00008-of-00032.bin",275    "h.6.self_attention.query.weight": "pytorch_model_00008-of-00032.bin",276    "h.7.input_layernorm.weight": "pytorch_model_00009-of-00032.bin",277    "h.7.mlp.down_proj.bias": "pytorch_model_00009-of-00032.bin",278    "h.7.mlp.down_proj.weight": "pytorch_model_00009-of-00032.bin",279    "h.7.mlp.gate_proj.weight": "pytorch_model_00009-of-00032.bin",280    "h.7.mlp.up_proj.weight": "pytorch_model_00009-of-00032.bin",281    "h.7.post_attention_layernorm.weight": "pytorch_model_00009-of-00032.bin",282    "h.7.self_attention.dense.bias": "pytorch_model_00009-of-00032.bin",283    "h.7.self_attention.dense.weight": "pytorch_model_00009-of-00032.bin",284    "h.7.self_attention.key_value.weight": "pytorch_model_00009-of-00032.bin",285    "h.7.self_attention.query.weight": "pytorch_model_00009-of-00032.bin",286    "h.8.input_layernorm.weight": "pytorch_model_00010-of-00032.bin",287    "h.8.mlp.down_proj.bias": "pytorch_model_00010-of-00032.bin",288    "h.8.mlp.down_proj.weight": "pytorch_model_00010-of-00032.bin",289    "h.8.mlp.gate_proj.weight": "pytorch_model_00010-of-00032.bin",290    "h.8.mlp.up_proj.weight": "pytorch_model_00010-of-00032.bin",291    "h.8.post_attention_layernorm.weight": "pytorch_model_00010-of-00032.bin",292    "h.8.self_attention.dense.bias": "pytorch_model_00010-of-00032.bin",293    "h.8.self_attention.dense.weight": "pytorch_model_00010-of-00032.bin",294    "h.8.self_attention.key_value.weight": "pytorch_model_00010-of-00032.bin",295    "h.8.self_attention.query.weight": "pytorch_model_00010-of-00032.bin",296    "h.9.input_layernorm.weight": "pytorch_model_00011-of-00032.bin",297    "h.9.mlp.down_proj.bias": "pytorch_model_00011-of-00032.bin",298    "h.9.mlp.down_proj.weight": "pytorch_model_00011-of-00032.bin",299    "h.9.mlp.gate_proj.weight": "pytorch_model_00011-of-00032.bin",300    "h.9.mlp.up_proj.weight": "pytorch_model_00011-of-00032.bin",301    "h.9.post_attention_layernorm.weight": "pytorch_model_00011-of-00032.bin",302    "h.9.self_attention.dense.bias": "pytorch_model_00011-of-00032.bin",303    "h.9.self_attention.dense.weight": "pytorch_model_00011-of-00032.bin",304    "h.9.self_attention.key_value.weight": "pytorch_model_00011-of-00032.bin",305    "h.9.self_attention.query.weight": "pytorch_model_00011-of-00032.bin",306    "ln_f.weight": "pytorch_model_00032-of-00032.bin",307    "word_embeddings.weight": "pytorch_model_00001-of-00032.bin"308  }309}310