diff --git a/config.json b/config.json index c7d952732a640a58a369bdff5ab6bcf6908cd1c4..6f78cba2babdf9a9d78541d7dd2527610b604c7b 100644 --- a/config.json +++ b/config.json @@ -1,5 +1,5 @@ { - "_name_or_path": "/project/lt200059-openth/.cache/huggingface/hub/models--daryl149--llama-2-7b-chat-hf/snapshots/d8654a4f69178a0c9260cf730241ebac2e72b923", + "_name_or_path": "/home/kviriyay/lt200059-openth/.cache/huggingface/hub/models--ChanonUtupon--openthaigpt-merge-lora-llama-2-7B/snapshots/fd2d65dace47949d7682e0285145dafc6af260b8", "architectures": [ "LlamaForCausalLM" ], @@ -9,15 +9,18 @@ "hidden_size": 4096, "initializer_range": 0.02, "intermediate_size": 11008, - "max_position_embeddings": 2048, + "max_position_embeddings": 4096, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 32, + "num_key_value_heads": 32, "pad_token_id": 0, - "rms_norm_eps": 1e-06, + "pretraining_tp": 1, + "rms_norm_eps": 1e-05, + "rope_scaling": null, "tie_word_embeddings": false, "torch_dtype": "float16", "transformers_version": "4.29.0", "use_cache": true, - "vocab_size": 32000 + "vocab_size": 56554 } diff --git a/generation_config.json b/generation_config.json index 66f8de0fce8e8ba487f5c8bb6ad72712d7e5ed0c..5526f6c0b4d365557fe8e3eaeef737553a5cf3a3 100644 --- a/generation_config.json +++ b/generation_config.json @@ -2,6 +2,9 @@ "_from_model_config": true, "bos_token_id": 1, "eos_token_id": 2, + "max_length": 4096, "pad_token_id": 0, + "temperature": 0.9, + "top_p": 0.6, "transformers_version": "4.29.0" } diff --git a/pytorch_model-00001-of-00039.bin b/pytorch_model-00001-of-00039.bin deleted file mode 100644 index 66f3f7dd211f20c46c50e6ab6dcaa8bb51454387..0000000000000000000000000000000000000000 --- a/pytorch_model-00001-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:fe38b8674f434765ec072bf666f63b97bfa24d7484d753ca627ec2100bd6a8c6 -size 396364479 diff --git a/pytorch_model-00001-of-00041.bin b/pytorch_model-00001-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..829c76205d4e2e222867550a893579920218094b --- /dev/null +++ b/pytorch_model-00001-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a84f3830738588b03a50e6a43d9267feee9ec16978a0565920d8d027531a418d +size 537 diff --git a/pytorch_model-00002-of-00039.bin b/pytorch_model-00002-of-00039.bin deleted file mode 100644 index 5ee1de4f5f774bdc2e925cb5fbe06687b3cd535a..0000000000000000000000000000000000000000 --- a/pytorch_model-00002-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:567dd141b24f9fe38f0715cdb44ddc5040260032817ee0764fbb1a91c579a247 -size 371215393 diff --git a/pytorch_model-00002-of-00041.bin b/pytorch_model-00002-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..c8bb789d20ff9d7b85d0dfefa94c9fa8c8c68ea3 --- /dev/null +++ b/pytorch_model-00002-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a14e289efe623dda34911224d278827211e44fdd7616465d3f9a645431970ede +size 463291306 diff --git a/pytorch_model-00003-of-00039.bin b/pytorch_model-00003-of-00039.bin deleted file mode 100644 index 7121c017fe93369fa4e48788f1f8c8fb833f782a..0000000000000000000000000000000000000000 --- a/pytorch_model-00003-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:21897bd09a2fdd438ba7eb6e5e3f73f2ccba1995b44c2e6894ddb60cec6f51a5 -size 371215986 diff --git a/pytorch_model-00003-of-00041.bin b/pytorch_model-00003-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..7433bda229cb7fda903f2b0c5bb60242d0a47675 --- /dev/null +++ b/pytorch_model-00003-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e45f3c518c61076c1b58b29128154df283fdfbd3fe04bbf2e274076d0baa7898 +size 314575888 diff --git a/pytorch_model-00004-of-00039.bin b/pytorch_model-00004-of-00039.bin deleted file mode 100644 index 139bcf93165b71d820cbe4b5e1405fcb7a6a5318..0000000000000000000000000000000000000000 --- a/pytorch_model-00004-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:ffb9232b5e4ebfc290ac4509a497f0832f5affa673771a2ef05feb67e5c215aa -size 371215986 diff --git a/pytorch_model-00004-of-00041.bin b/pytorch_model-00004-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..5fadf1f8893df73033ed3985c96eb5e70fc6ff1a --- /dev/null +++ b/pytorch_model-00004-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:94a7e0aa2ba7131ce7c047374d9616b3914d4ff1444463437f6224da421f2a02 +size 314592882 diff --git a/pytorch_model-00005-of-00039.bin b/pytorch_model-00005-of-00039.bin deleted file mode 100644 index 14bb57cc76492e6cf10aae06e3a8519159da14cc..0000000000000000000000000000000000000000 --- a/pytorch_model-00005-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:230e822392707c874a228d1db02481373a7902706eba1c4770618750b29d9952 -size 371215986 diff --git a/pytorch_model-00005-of-00041.bin b/pytorch_model-00005-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..687022f708d167100955419255d9fa0eb5163157 --- /dev/null +++ b/pytorch_model-00005-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9978b5d3dc5d9cda206c6f9e6b535d9073007bad2f431aee783d668f250ae5e +size 314592882 diff --git a/pytorch_model-00006-of-00039.bin b/pytorch_model-00006-of-00039.bin deleted file mode 100644 index ede34dac1833effe3b31659ce78e23f6cefbf44f..0000000000000000000000000000000000000000 --- a/pytorch_model-00006-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:0ad788b0be08b250bd15a4c904e12717f32cfb8056cf5b555aa55ad23292c5b8 -size 314575888 diff --git a/pytorch_model-00006-of-00041.bin b/pytorch_model-00006-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..246f9635c69577cad1101384cd6dd195e09c652b --- /dev/null +++ b/pytorch_model-00006-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3436e05dfc50009e6a3ad209eff87c94e09873983bf8d259467af9ca72be6d8d +size 371215393 diff --git a/pytorch_model-00007-of-00039.bin b/pytorch_model-00007-of-00039.bin deleted file mode 100644 index 61cefe2ed838c8ca84626489349497bc7f4b052b..0000000000000000000000000000000000000000 --- a/pytorch_model-00007-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:af0f7668496a668337c29541cd6350fe81aa526656a6d1fe0c5aa61f6582df6d -size 314592882 diff --git a/pytorch_model-00007-of-00041.bin b/pytorch_model-00007-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..8e322b90a2a11b33511426a028d592d457602e6f --- /dev/null +++ b/pytorch_model-00007-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ff389907f17bd191a2825fd4391d18b2fe54c5edf5f3d6f8d1befb8664c13ec8 +size 371215986 diff --git a/pytorch_model-00008-of-00039.bin b/pytorch_model-00008-of-00039.bin deleted file mode 100644 index 58be0b4218e54e0db3b6f46effbcc1992d03f177..0000000000000000000000000000000000000000 --- a/pytorch_model-00008-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:891c1e0584197f83d2286ddb4d550b811f8de41bb7c67e62d3908e21945ff4dd -size 314592882 diff --git a/pytorch_model-00008-of-00041.bin b/pytorch_model-00008-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..dc24bb04c399b48777d7a5b50cd8564ebb6fd797 --- /dev/null +++ b/pytorch_model-00008-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58daead9601404784f9f8c91259afebaa57b5b74ee242f7659e11ae67e9c68d0 +size 371215986 diff --git a/pytorch_model-00009-of-00039.bin b/pytorch_model-00009-of-00039.bin deleted file mode 100644 index 5786e4f50cfb7532d5051f8631fd61e8b89e3c20..0000000000000000000000000000000000000000 --- a/pytorch_model-00009-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:818d760dbe9433b5af6d72ee4e338df25a8fe93dac2df631c84069a150ae8f14 -size 371215393 diff --git a/pytorch_model-00009-of-00041.bin b/pytorch_model-00009-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..65c7b8413fa0a79d22ac26331cc92e7d648681e7 --- /dev/null +++ b/pytorch_model-00009-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9f63d2e0cf54e79c5d663b83e6ef0cff04e92743422221daf2253a3895330ff +size 371215986 diff --git a/pytorch_model-00010-of-00039.bin b/pytorch_model-00010-of-00039.bin deleted file mode 100644 index e6e03b03441d62fbee443b4127fb4a47068dc2da..0000000000000000000000000000000000000000 --- a/pytorch_model-00010-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:86a85d53b815f5f9f5c811e49108cfdb03725d8a2b860a0f3f5b98acaffa7dd3 -size 371215986 diff --git a/pytorch_model-00010-of-00041.bin b/pytorch_model-00010-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..ab639b15e86077ee529a6d45b1afbbb92a7750f0 --- /dev/null +++ b/pytorch_model-00010-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57b28b6491ca1923df2c0263a37f1e9bc95497cdb1d508aabd5c255c6592fa42 +size 314575888 diff --git a/pytorch_model-00011-of-00039.bin b/pytorch_model-00011-of-00039.bin deleted file mode 100644 index b0bb7661c7daf096975abb4b74afd12f636b2ffa..0000000000000000000000000000000000000000 --- a/pytorch_model-00011-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:27c4618e25a63198e7511874e1dba4bf46f7952b7ccf6b1f53a4ca3c8ecb5791 -size 371215986 diff --git a/pytorch_model-00011-of-00041.bin b/pytorch_model-00011-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..02612d39eefa9cd16cb56ea905bcfa24564ffa58 --- /dev/null +++ b/pytorch_model-00011-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2412e30df08dea8ff8194d91256f921023a23ab753d8fef58412100f5533942 +size 314592882 diff --git a/pytorch_model-00012-of-00039.bin b/pytorch_model-00012-of-00039.bin deleted file mode 100644 index cc5d186f6092fd5127cc20b18e452461353f1609..0000000000000000000000000000000000000000 --- a/pytorch_model-00012-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:3a4f7f4a6a4274130ee81030f08959d4661e8caec43aadda802046c6840173ad -size 371215986 diff --git a/pytorch_model-00012-of-00041.bin b/pytorch_model-00012-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..b90cc82695a5761902a1915c0e9bd0f20bf1fbb0 --- /dev/null +++ b/pytorch_model-00012-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e8d454b8b739891b82660b071be3dda190adbf606b9ccf99c2901c5ab686366a +size 314592882 diff --git a/pytorch_model-00013-of-00039.bin b/pytorch_model-00013-of-00039.bin deleted file mode 100644 index 50df0d389be11f465fcb3734d8c80004914a6322..0000000000000000000000000000000000000000 --- a/pytorch_model-00013-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:5fa960f004ba3f5dd7cec1c10ece9f5bca02cf5e8f6401eda744ced4683ce41d -size 314575888 diff --git a/pytorch_model-00013-of-00041.bin b/pytorch_model-00013-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..cacc92e2c60d54b0421b9549627fec7a5bfa5727 --- /dev/null +++ b/pytorch_model-00013-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ece794b263ebcb99a7c3f8bb7123e9f69fb6d114bddc3bfb09c0a2d9e46bb504 +size 371215393 diff --git a/pytorch_model-00014-of-00039.bin b/pytorch_model-00014-of-00039.bin deleted file mode 100644 index 9554510ac554f74023ff9058449fc1414b7eaa13..0000000000000000000000000000000000000000 --- a/pytorch_model-00014-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:83afdfd23765d5dd21975843003de193b9c231169d66685a4276af8387003f54 -size 314592882 diff --git a/pytorch_model-00014-of-00041.bin b/pytorch_model-00014-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..f16e571d1ec0d023ee03cc8bed19164d99b72f25 --- /dev/null +++ b/pytorch_model-00014-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:811be5881fbcc063c9b35c273c693f403f17c213db5d83044d4540e1475757e1 +size 371215986 diff --git a/pytorch_model-00015-of-00039.bin b/pytorch_model-00015-of-00039.bin deleted file mode 100644 index 5c18d6076265562a65aa6effcc872328db0d04df..0000000000000000000000000000000000000000 --- a/pytorch_model-00015-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:a30d391f53bda6e1a991dbdde8036b835c39a0cdb57de4de53499b6cacd8d847 -size 314592882 diff --git a/pytorch_model-00015-of-00041.bin b/pytorch_model-00015-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..917e9c261c439ed1cdebd616ecb029650728b160 --- /dev/null +++ b/pytorch_model-00015-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a1cb4c01c8f06b9417c375c965a9fdf0e804ebb570186693bc6e2eb9cc79a591 +size 371215986 diff --git a/pytorch_model-00016-of-00039.bin b/pytorch_model-00016-of-00039.bin deleted file mode 100644 index 35a24d2278f047fbe20f166abd5665f786cfede5..0000000000000000000000000000000000000000 --- a/pytorch_model-00016-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:a3b5d99ace2c5439c03f0e228761ee9d1363c15c95b85545467fd19a69dd9fe6 -size 371215393 diff --git a/pytorch_model-00016-of-00041.bin b/pytorch_model-00016-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..1a8f0913f22a71023952937232c54dae660c15f3 --- /dev/null +++ b/pytorch_model-00016-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:893034f5a4bcf38b48be9411a7e8fc8aebb75322b99e11aa7ed99960faa63cea +size 371215986 diff --git a/pytorch_model-00017-of-00039.bin b/pytorch_model-00017-of-00039.bin deleted file mode 100644 index 92ee97bc37b2557262846064c7f8aae30683e04b..0000000000000000000000000000000000000000 --- a/pytorch_model-00017-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:aafcb10dfa023c7684c239cc41640e3bae7c58df07a897044a3a4b4f35dc264c -size 371215986 diff --git a/pytorch_model-00017-of-00041.bin b/pytorch_model-00017-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..b5ebdabd69f233e5b174c367786ebf5c7a6f146b --- /dev/null +++ b/pytorch_model-00017-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65bb12ed523b37706befcb3acb1feba1afbb71333c6ac934bc234da65ed4f95a +size 314575888 diff --git a/pytorch_model-00018-of-00039.bin b/pytorch_model-00018-of-00039.bin deleted file mode 100644 index 2cf8d27bf317361a6842c2628b074f01d358ffc3..0000000000000000000000000000000000000000 --- a/pytorch_model-00018-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:3bc451038e8e2159b11719177e65d4a17eff1460af59b78bc393c1fe8c7e5335 -size 371215986 diff --git a/pytorch_model-00018-of-00041.bin b/pytorch_model-00018-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..05eb0280ae2a0fa20f03e83359362e4d144debc1 --- /dev/null +++ b/pytorch_model-00018-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bbd37b800ff0ff0e9351c6b183bb5e7a5f629131b006f09e66eeff408bf1358b +size 314592882 diff --git a/pytorch_model-00019-of-00039.bin b/pytorch_model-00019-of-00039.bin deleted file mode 100644 index b5fedfa8f3369803348fbec51474c66d54aae90a..0000000000000000000000000000000000000000 --- a/pytorch_model-00019-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:370b3b71b7abac7b1e9d44d14e996051038df535b245d83eefbafec8f2e9cb2e -size 371215986 diff --git a/pytorch_model-00019-of-00041.bin b/pytorch_model-00019-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..40666805a4f68c37784fa19724d2ff97843ad28a --- /dev/null +++ b/pytorch_model-00019-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b6969d780b1e51a438691119b2c437cc1288763b948e19b4c85e4155d2383dd +size 314592882 diff --git a/pytorch_model-00020-of-00039.bin b/pytorch_model-00020-of-00039.bin deleted file mode 100644 index 9c18e8f36ce106be1254c7ba503b836e445aadbf..0000000000000000000000000000000000000000 --- a/pytorch_model-00020-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:5e9e4cfdd35fd23a8914dcb4a021846c42ee9bd315efa686230d32b9ca9cd881 -size 314575888 diff --git a/pytorch_model-00020-of-00041.bin b/pytorch_model-00020-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..018f45b6b4793f0c570d03b9a905e2a41e861da4 --- /dev/null +++ b/pytorch_model-00020-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82b9ad229c8224e333bd5cc474fd9103562459b8d5e4a4d33bf1c65b2df9793c +size 371215393 diff --git a/pytorch_model-00021-of-00039.bin b/pytorch_model-00021-of-00039.bin deleted file mode 100644 index 63f9cb57e5142991754a8e3996a7905577e41c81..0000000000000000000000000000000000000000 --- a/pytorch_model-00021-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:2bf27fd98853ae003178850ecd17317c72e894c2ff79220744f9a46de4161c02 -size 314592882 diff --git a/pytorch_model-00021-of-00041.bin b/pytorch_model-00021-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..a4bd287fec8c443e00aef367b47e2272385f2ba7 --- /dev/null +++ b/pytorch_model-00021-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:183fc16fbb856e3e7273ec78e2a563273bac9bd616f139f22df7da6604bd3e31 +size 371215986 diff --git a/pytorch_model-00022-of-00039.bin b/pytorch_model-00022-of-00039.bin deleted file mode 100644 index 73b213cb7e66b4a56b84e620dc3e5e09b52fc88f..0000000000000000000000000000000000000000 --- a/pytorch_model-00022-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:a3797ec61650672727c8320ed96ce044b84b1882f67f070b93183f6da51f7e4a -size 314592882 diff --git a/pytorch_model-00022-of-00041.bin b/pytorch_model-00022-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..ea7b5b23a11b1f352d8c489107caef7cc8ebdc49 --- /dev/null +++ b/pytorch_model-00022-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c619eb40545f9169f3397701851fffa6ab61e0fd54b8c09dc159b63dbbc0c0e +size 371215986 diff --git a/pytorch_model-00023-of-00039.bin b/pytorch_model-00023-of-00039.bin deleted file mode 100644 index 29dc288224f5fe0425e0d6d9963b3d2e082dc0b0..0000000000000000000000000000000000000000 --- a/pytorch_model-00023-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:f3e30b0fc19e9fbd5a0bcb3f5c1048a39a0572869c8c06545028c477a840f49d -size 371215393 diff --git a/pytorch_model-00023-of-00041.bin b/pytorch_model-00023-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..290ade576d4137dbbdb691e6786955d5e5601435 --- /dev/null +++ b/pytorch_model-00023-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c937b960b605bcd115093cfbc4462f480ef1adf2cd24d4de7aa67305ebee8c5d +size 371215986 diff --git a/pytorch_model-00024-of-00039.bin b/pytorch_model-00024-of-00039.bin deleted file mode 100644 index 41c3f22493004d317501e5f0acf4114232686f94..0000000000000000000000000000000000000000 --- a/pytorch_model-00024-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:351a7e84d4daf6c5bb8a43206b5e5e105c21a196b3f62c65b76db30582e4a993 -size 371215986 diff --git a/pytorch_model-00024-of-00041.bin b/pytorch_model-00024-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..db1a483c40fab050a64a80efbd39923b13942483 --- /dev/null +++ b/pytorch_model-00024-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:381e233f75ca847aa955b5056fbd8319bef408095d571f97faeb8f6b6e7af448 +size 314575888 diff --git a/pytorch_model-00025-of-00039.bin b/pytorch_model-00025-of-00039.bin deleted file mode 100644 index ad9e77323c7f5e0d5b08ea260ac8fb072b5924fd..0000000000000000000000000000000000000000 --- a/pytorch_model-00025-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:8c2acf37331a0ce418909fdd3b868999a1646daed1cbbb4444aaeda655c54245 -size 371215986 diff --git a/pytorch_model-00025-of-00041.bin b/pytorch_model-00025-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..cd0b7f6f21d8b281b3ded10c1f725f1b03443c44 --- /dev/null +++ b/pytorch_model-00025-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f307e8db035760e96fc65d0472d219ef7726c1a34bec0ca3c5af9a3c4f2e393 +size 314592882 diff --git a/pytorch_model-00026-of-00039.bin b/pytorch_model-00026-of-00039.bin deleted file mode 100644 index 44d451b7216ce765eb0c7e6ae47bfab65fa92b13..0000000000000000000000000000000000000000 --- a/pytorch_model-00026-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:e6ac65a1bd86ff080fe237dcfedf08d57857d5ebaeeac27b38bbaaf4328d820e -size 371215986 diff --git a/pytorch_model-00026-of-00041.bin b/pytorch_model-00026-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..d807a2a7aa93a36e1d98f6f7cafb920a00b0570a --- /dev/null +++ b/pytorch_model-00026-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8c86e30701b6596215a6834f3e4c32f1b6e916e2e8ac7c5b9dffbffa3ab4cb6 +size 314592882 diff --git a/pytorch_model-00027-of-00039.bin b/pytorch_model-00027-of-00039.bin deleted file mode 100644 index 998001b54aceb959a564ff92ad3b57a1fc59dc73..0000000000000000000000000000000000000000 --- a/pytorch_model-00027-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:8d83d9fab596e0dff584d64a98870776780009c5b32da84d03b3f7a24a55c2fb -size 314575888 diff --git a/pytorch_model-00027-of-00041.bin b/pytorch_model-00027-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..ad0058e6f112eb5c31817a595e9362c835541d93 --- /dev/null +++ b/pytorch_model-00027-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f928e6197788cc7e18a8b833ca1e3656319454f4738268fc01587c3aa4fece43 +size 371215393 diff --git a/pytorch_model-00028-of-00039.bin b/pytorch_model-00028-of-00039.bin deleted file mode 100644 index e8898ad37bea2ef772a2005bbf555df1d01081e6..0000000000000000000000000000000000000000 --- a/pytorch_model-00028-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:5458cb6d9eeef57b179792f287f199dc4983214f94ae4b5998620348698a76eb -size 314592882 diff --git a/pytorch_model-00028-of-00041.bin b/pytorch_model-00028-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..18ae63de5efddb60652bda70ef211face75ab527 --- /dev/null +++ b/pytorch_model-00028-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0343fbe5a4c64d31d91248d4d947e48c99f755555907bd9098506762bce843f9 +size 371215986 diff --git a/pytorch_model-00029-of-00039.bin b/pytorch_model-00029-of-00039.bin deleted file mode 100644 index 7ce5d131ccca2d4a614b074888b82ebf1973373c..0000000000000000000000000000000000000000 --- a/pytorch_model-00029-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:e7cc2fa23e15a68a83b50827a5c275bbda55c658051100f8d5d3de5cd67f918e -size 314592882 diff --git a/pytorch_model-00029-of-00041.bin b/pytorch_model-00029-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..edb93366afb9bc33b0b517e92636903eabf08bdd --- /dev/null +++ b/pytorch_model-00029-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7654548a3176d790318535c13152482134e1abfd5222fb3cc61e2f386c9d7f2f +size 371215986 diff --git a/pytorch_model-00030-of-00039.bin b/pytorch_model-00030-of-00039.bin deleted file mode 100644 index f7fa85579f8c16de0160bdb743e1eec3bb258f21..0000000000000000000000000000000000000000 --- a/pytorch_model-00030-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:c8975dd0e8fa70b6dccd5269c2f5eb9df8e04a8f24bdf79d3997bb922c90c41d -size 371215393 diff --git a/pytorch_model-00030-of-00041.bin b/pytorch_model-00030-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..1348c835b9454172356a50634ca6642d63247524 --- /dev/null +++ b/pytorch_model-00030-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e04421b5d73cfb6b3f22d4ea130b6587c200701ecaea40d4f75ab6df365c824 +size 371215986 diff --git a/pytorch_model-00031-of-00039.bin b/pytorch_model-00031-of-00039.bin deleted file mode 100644 index 2a94bbad043d08289795e7aababa8976d1728b9a..0000000000000000000000000000000000000000 --- a/pytorch_model-00031-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:b6139c95d7d0bafdf6d8f075cbfd6c6d20c5db3df52b3fbae93ee1bf547c26ed -size 371215986 diff --git a/pytorch_model-00031-of-00041.bin b/pytorch_model-00031-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..4eb01eb78d33af5b61c061c67ac64ebe3f1cba75 --- /dev/null +++ b/pytorch_model-00031-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4f36b98600a73f932694502be99975d88c15e7cbd4cc4bc95aedcea1b226177f +size 314575888 diff --git a/pytorch_model-00032-of-00039.bin b/pytorch_model-00032-of-00039.bin deleted file mode 100644 index 3a89f04a99a79e30d236205be877e6278d8a15b4..0000000000000000000000000000000000000000 --- a/pytorch_model-00032-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:03503f0e9b6dd484ca19c0482dcae4d54c06f30cdb6225b4448b20384176562d -size 371215986 diff --git a/pytorch_model-00032-of-00041.bin b/pytorch_model-00032-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..47e5707e45dbaab8c05c96561eeb3920d8dfe48c --- /dev/null +++ b/pytorch_model-00032-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:607467cf4ad4d71677b13fb6beb7f866a31db208d49be4ab42ad8c6f9691fa8c +size 314592882 diff --git a/pytorch_model-00033-of-00039.bin b/pytorch_model-00033-of-00039.bin deleted file mode 100644 index dbc199d28869356bd5312b62fce0df165bdcac90..0000000000000000000000000000000000000000 --- a/pytorch_model-00033-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:74bb29f1d6d928a48e87c2bf3c4ad3629862fead97908494116de9b4036a716a -size 371215986 diff --git a/pytorch_model-00033-of-00041.bin b/pytorch_model-00033-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..6a351e726f66181f8d00ed9c6894d39a8d6322f8 --- /dev/null +++ b/pytorch_model-00033-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b14ce7c4a9f0b3869957f68886c06c564dd7880cb0d153f022ca45c7a065fbe2 +size 314592882 diff --git a/pytorch_model-00034-of-00039.bin b/pytorch_model-00034-of-00039.bin deleted file mode 100644 index 291b3341f6e775e892a352e757ec1e727d95ce4f..0000000000000000000000000000000000000000 --- a/pytorch_model-00034-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:3b30fab70f3099ec5e3afd654fe28b40f66c7b561ab0b3453a2661552ba24d9f -size 314575888 diff --git a/pytorch_model-00034-of-00041.bin b/pytorch_model-00034-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..8c04de68578d986aa1d028fc0c555474ddee690a --- /dev/null +++ b/pytorch_model-00034-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1133450fa7ae5df4578c7587a682efaf4a4907cbe03499efd65639b45863ae9f +size 371215393 diff --git a/pytorch_model-00035-of-00039.bin b/pytorch_model-00035-of-00039.bin deleted file mode 100644 index bb2d40ee253c49dd07bd97eef6c53314c8ce66b9..0000000000000000000000000000000000000000 --- a/pytorch_model-00035-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:87236715920308af65325c02169dd556a051d0fb4c6c67231d9d3e14ae3f40c1 -size 314592882 diff --git a/pytorch_model-00035-of-00041.bin b/pytorch_model-00035-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..cf551b3ef5bf0c2f9c0288335d70469cc361b2ad --- /dev/null +++ b/pytorch_model-00035-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f2d431971ecfb2024f99910426e4776a3489a331483ebaab6385c6a8ee5168c +size 371215986 diff --git a/pytorch_model-00036-of-00039.bin b/pytorch_model-00036-of-00039.bin deleted file mode 100644 index d30859e193db86340181b40e18a608f66509aabf..0000000000000000000000000000000000000000 --- a/pytorch_model-00036-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:8e023f0b43fd93b47cd6c98106264d934601601b4f283e131c387bf97b241ec0 -size 314592882 diff --git a/pytorch_model-00036-of-00041.bin b/pytorch_model-00036-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..7385a932e7ab2963e7a773f3c222c30c1dc4385d --- /dev/null +++ b/pytorch_model-00036-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:29ce28c0a62d182818a7ab2405fa4b8e7cf1fa30b675acb03d014588d84c9c82 +size 371215986 diff --git a/pytorch_model-00037-of-00039.bin b/pytorch_model-00037-of-00039.bin deleted file mode 100644 index b467443172da95cc4ba3f4a6d9318022fc49571f..0000000000000000000000000000000000000000 --- a/pytorch_model-00037-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:5061f2bf0985604a8e7d409d09f0f0947fabad9219e9fe64f3194d85742523eb -size 371215393 diff --git a/pytorch_model-00037-of-00041.bin b/pytorch_model-00037-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..696a5043d13ed90b9554d04272dc348afc56c211 --- /dev/null +++ b/pytorch_model-00037-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26dc6f45c5945f795cdbea5e04b88f1d835ad80a5059abcb3ccdb3697994f862 +size 371215986 diff --git a/pytorch_model-00038-of-00039.bin b/pytorch_model-00038-of-00039.bin deleted file mode 100644 index c650ad1adb291404cebfd7f42ba8a3276a5e7d63..0000000000000000000000000000000000000000 --- a/pytorch_model-00038-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:a7dfc779e32574edac9244021c33512397888415efd19e6d79acd89f8ef70f86 -size 304114977 diff --git a/pytorch_model-00038-of-00041.bin b/pytorch_model-00038-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..415ca8453681f524e97dab9aec8d04c43825be43 --- /dev/null +++ b/pytorch_model-00038-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26cee756c0d66f7bccc21bd4295df5983ab5efad65405a27fe5beefa94f572df +size 314575888 diff --git a/pytorch_model-00039-of-00039.bin b/pytorch_model-00039-of-00039.bin deleted file mode 100644 index 5f3672650fadc3893c06446abffb25e118fe6856..0000000000000000000000000000000000000000 --- a/pytorch_model-00039-of-00039.bin +++ /dev/null @@ -1,3 +0,0 @@ -version https://git-lfs.github.com/spec/v1 -oid sha256:4faa8dc42a67a488687de8b673d36b429524140d1a844df7ef3281c70e3a3750 -size 262144938 diff --git a/pytorch_model-00039-of-00041.bin b/pytorch_model-00039-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..dbe257314fb4cd2c3d99284d2d5415a8f9ce6e9a --- /dev/null +++ b/pytorch_model-00039-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9dd75c62a05ed8e2761766910c57cf7da0bb4227c510a3e121de8d4053b16046 +size 314592882 diff --git a/pytorch_model-00040-of-00041.bin b/pytorch_model-00040-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..746e6f4160d097ae52dfddcf7fd205affaeef020 --- /dev/null +++ b/pytorch_model-00040-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fa94382ffb7c6e2bc819c6def43993bcb44573a02e9d89fcf72486d677d9c002 +size 180381806 diff --git a/pytorch_model-00041-of-00041.bin b/pytorch_model-00041-of-00041.bin new file mode 100644 index 0000000000000000000000000000000000000000..777757bde9886ea860af193908b4fa6d8e5b4dc9 --- /dev/null +++ b/pytorch_model-00041-of-00041.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd18cb36ca8603b1df1817c96c908d80e9fe45908c0e2bd7ace3105b438ee639 +size 463291306 diff --git a/pytorch_model.bin.index.json b/pytorch_model.bin.index.json index 37bef0973307eb5cf479070232c9ba0a2c7bcfbe..2a9284212b93e87cea3eb091342a1fce749429b6 100644 --- a/pytorch_model.bin.index.json +++ b/pytorch_model.bin.index.json @@ -1,330 +1,330 @@ { "metadata": { - "total_size": 13476839424 + "total_size": 13879132160 }, "weight_map": { - "lm_head.weight": "pytorch_model-00039-of-00039.bin", - "model.embed_tokens.weight": "pytorch_model-00001-of-00039.bin", - "model.layers.0.input_layernorm.weight": "pytorch_model-00002-of-00039.bin", - "model.layers.0.mlp.down_proj.weight": "pytorch_model-00002-of-00039.bin", - "model.layers.0.mlp.gate_proj.weight": "pytorch_model-00002-of-00039.bin", - "model.layers.0.mlp.up_proj.weight": "pytorch_model-00002-of-00039.bin", - "model.layers.0.post_attention_layernorm.weight": "pytorch_model-00002-of-00039.bin", - "model.layers.0.self_attn.k_proj.weight": "pytorch_model-00001-of-00039.bin", - "model.layers.0.self_attn.o_proj.weight": "pytorch_model-00001-of-00039.bin", - "model.layers.0.self_attn.q_proj.weight": "pytorch_model-00001-of-00039.bin", - "model.layers.0.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00039.bin", - "model.layers.0.self_attn.v_proj.weight": "pytorch_model-00001-of-00039.bin", - "model.layers.1.input_layernorm.weight": "pytorch_model-00003-of-00039.bin", - "model.layers.1.mlp.down_proj.weight": "pytorch_model-00003-of-00039.bin", - "model.layers.1.mlp.gate_proj.weight": "pytorch_model-00003-of-00039.bin", - "model.layers.1.mlp.up_proj.weight": "pytorch_model-00003-of-00039.bin", - "model.layers.1.post_attention_layernorm.weight": "pytorch_model-00003-of-00039.bin", - "model.layers.1.self_attn.k_proj.weight": "pytorch_model-00002-of-00039.bin", - "model.layers.1.self_attn.o_proj.weight": "pytorch_model-00003-of-00039.bin", - "model.layers.1.self_attn.q_proj.weight": "pytorch_model-00002-of-00039.bin", - "model.layers.1.self_attn.rotary_emb.inv_freq": "pytorch_model-00003-of-00039.bin", - "model.layers.1.self_attn.v_proj.weight": "pytorch_model-00002-of-00039.bin", - "model.layers.10.input_layernorm.weight": "pytorch_model-00014-of-00039.bin", - "model.layers.10.mlp.down_proj.weight": "pytorch_model-00013-of-00039.bin", - "model.layers.10.mlp.gate_proj.weight": "pytorch_model-00013-of-00039.bin", - "model.layers.10.mlp.up_proj.weight": "pytorch_model-00014-of-00039.bin", - "model.layers.10.post_attention_layernorm.weight": "pytorch_model-00014-of-00039.bin", - "model.layers.10.self_attn.k_proj.weight": "pytorch_model-00013-of-00039.bin", - "model.layers.10.self_attn.o_proj.weight": "pytorch_model-00013-of-00039.bin", - "model.layers.10.self_attn.q_proj.weight": "pytorch_model-00013-of-00039.bin", - "model.layers.10.self_attn.rotary_emb.inv_freq": "pytorch_model-00013-of-00039.bin", - "model.layers.10.self_attn.v_proj.weight": "pytorch_model-00013-of-00039.bin", - "model.layers.11.input_layernorm.weight": "pytorch_model-00015-of-00039.bin", - "model.layers.11.mlp.down_proj.weight": "pytorch_model-00015-of-00039.bin", - "model.layers.11.mlp.gate_proj.weight": "pytorch_model-00014-of-00039.bin", - "model.layers.11.mlp.up_proj.weight": "pytorch_model-00015-of-00039.bin", - "model.layers.11.post_attention_layernorm.weight": "pytorch_model-00015-of-00039.bin", - "model.layers.11.self_attn.k_proj.weight": "pytorch_model-00014-of-00039.bin", - "model.layers.11.self_attn.o_proj.weight": "pytorch_model-00014-of-00039.bin", - "model.layers.11.self_attn.q_proj.weight": "pytorch_model-00014-of-00039.bin", - "model.layers.11.self_attn.rotary_emb.inv_freq": "pytorch_model-00014-of-00039.bin", - "model.layers.11.self_attn.v_proj.weight": "pytorch_model-00014-of-00039.bin", - "model.layers.12.input_layernorm.weight": "pytorch_model-00016-of-00039.bin", - "model.layers.12.mlp.down_proj.weight": "pytorch_model-00016-of-00039.bin", - "model.layers.12.mlp.gate_proj.weight": "pytorch_model-00016-of-00039.bin", - "model.layers.12.mlp.up_proj.weight": "pytorch_model-00016-of-00039.bin", - "model.layers.12.post_attention_layernorm.weight": "pytorch_model-00016-of-00039.bin", - "model.layers.12.self_attn.k_proj.weight": "pytorch_model-00015-of-00039.bin", - "model.layers.12.self_attn.o_proj.weight": "pytorch_model-00015-of-00039.bin", - "model.layers.12.self_attn.q_proj.weight": "pytorch_model-00015-of-00039.bin", - "model.layers.12.self_attn.rotary_emb.inv_freq": "pytorch_model-00015-of-00039.bin", - "model.layers.12.self_attn.v_proj.weight": "pytorch_model-00015-of-00039.bin", - "model.layers.13.input_layernorm.weight": "pytorch_model-00017-of-00039.bin", - "model.layers.13.mlp.down_proj.weight": "pytorch_model-00017-of-00039.bin", - "model.layers.13.mlp.gate_proj.weight": "pytorch_model-00017-of-00039.bin", - "model.layers.13.mlp.up_proj.weight": "pytorch_model-00017-of-00039.bin", - "model.layers.13.post_attention_layernorm.weight": "pytorch_model-00017-of-00039.bin", - "model.layers.13.self_attn.k_proj.weight": "pytorch_model-00016-of-00039.bin", - "model.layers.13.self_attn.o_proj.weight": "pytorch_model-00017-of-00039.bin", - "model.layers.13.self_attn.q_proj.weight": "pytorch_model-00016-of-00039.bin", - "model.layers.13.self_attn.rotary_emb.inv_freq": "pytorch_model-00017-of-00039.bin", - "model.layers.13.self_attn.v_proj.weight": "pytorch_model-00016-of-00039.bin", - "model.layers.14.input_layernorm.weight": "pytorch_model-00018-of-00039.bin", - "model.layers.14.mlp.down_proj.weight": "pytorch_model-00018-of-00039.bin", - "model.layers.14.mlp.gate_proj.weight": "pytorch_model-00018-of-00039.bin", - "model.layers.14.mlp.up_proj.weight": "pytorch_model-00018-of-00039.bin", - "model.layers.14.post_attention_layernorm.weight": "pytorch_model-00018-of-00039.bin", - "model.layers.14.self_attn.k_proj.weight": "pytorch_model-00017-of-00039.bin", - "model.layers.14.self_attn.o_proj.weight": "pytorch_model-00018-of-00039.bin", - "model.layers.14.self_attn.q_proj.weight": "pytorch_model-00017-of-00039.bin", - "model.layers.14.self_attn.rotary_emb.inv_freq": "pytorch_model-00018-of-00039.bin", - "model.layers.14.self_attn.v_proj.weight": "pytorch_model-00018-of-00039.bin", - "model.layers.15.input_layernorm.weight": "pytorch_model-00019-of-00039.bin", - "model.layers.15.mlp.down_proj.weight": "pytorch_model-00019-of-00039.bin", - "model.layers.15.mlp.gate_proj.weight": "pytorch_model-00019-of-00039.bin", - "model.layers.15.mlp.up_proj.weight": "pytorch_model-00019-of-00039.bin", - "model.layers.15.post_attention_layernorm.weight": "pytorch_model-00019-of-00039.bin", - "model.layers.15.self_attn.k_proj.weight": "pytorch_model-00019-of-00039.bin", - "model.layers.15.self_attn.o_proj.weight": "pytorch_model-00019-of-00039.bin", - "model.layers.15.self_attn.q_proj.weight": "pytorch_model-00018-of-00039.bin", - "model.layers.15.self_attn.rotary_emb.inv_freq": "pytorch_model-00019-of-00039.bin", - "model.layers.15.self_attn.v_proj.weight": "pytorch_model-00019-of-00039.bin", - "model.layers.16.input_layernorm.weight": "pytorch_model-00021-of-00039.bin", - "model.layers.16.mlp.down_proj.weight": "pytorch_model-00020-of-00039.bin", - "model.layers.16.mlp.gate_proj.weight": "pytorch_model-00020-of-00039.bin", - "model.layers.16.mlp.up_proj.weight": "pytorch_model-00021-of-00039.bin", - "model.layers.16.post_attention_layernorm.weight": "pytorch_model-00021-of-00039.bin", - "model.layers.16.self_attn.k_proj.weight": "pytorch_model-00020-of-00039.bin", - "model.layers.16.self_attn.o_proj.weight": "pytorch_model-00020-of-00039.bin", - "model.layers.16.self_attn.q_proj.weight": "pytorch_model-00020-of-00039.bin", - "model.layers.16.self_attn.rotary_emb.inv_freq": "pytorch_model-00020-of-00039.bin", - "model.layers.16.self_attn.v_proj.weight": "pytorch_model-00020-of-00039.bin", - "model.layers.17.input_layernorm.weight": "pytorch_model-00022-of-00039.bin", - "model.layers.17.mlp.down_proj.weight": "pytorch_model-00022-of-00039.bin", - "model.layers.17.mlp.gate_proj.weight": "pytorch_model-00021-of-00039.bin", - "model.layers.17.mlp.up_proj.weight": "pytorch_model-00022-of-00039.bin", - "model.layers.17.post_attention_layernorm.weight": "pytorch_model-00022-of-00039.bin", - "model.layers.17.self_attn.k_proj.weight": "pytorch_model-00021-of-00039.bin", - "model.layers.17.self_attn.o_proj.weight": "pytorch_model-00021-of-00039.bin", - "model.layers.17.self_attn.q_proj.weight": "pytorch_model-00021-of-00039.bin", - "model.layers.17.self_attn.rotary_emb.inv_freq": "pytorch_model-00021-of-00039.bin", - "model.layers.17.self_attn.v_proj.weight": "pytorch_model-00021-of-00039.bin", - "model.layers.18.input_layernorm.weight": "pytorch_model-00023-of-00039.bin", - "model.layers.18.mlp.down_proj.weight": "pytorch_model-00023-of-00039.bin", - "model.layers.18.mlp.gate_proj.weight": "pytorch_model-00023-of-00039.bin", - "model.layers.18.mlp.up_proj.weight": "pytorch_model-00023-of-00039.bin", - "model.layers.18.post_attention_layernorm.weight": "pytorch_model-00023-of-00039.bin", - "model.layers.18.self_attn.k_proj.weight": "pytorch_model-00022-of-00039.bin", - "model.layers.18.self_attn.o_proj.weight": "pytorch_model-00022-of-00039.bin", - "model.layers.18.self_attn.q_proj.weight": "pytorch_model-00022-of-00039.bin", - "model.layers.18.self_attn.rotary_emb.inv_freq": "pytorch_model-00022-of-00039.bin", - "model.layers.18.self_attn.v_proj.weight": "pytorch_model-00022-of-00039.bin", - "model.layers.19.input_layernorm.weight": "pytorch_model-00024-of-00039.bin", - "model.layers.19.mlp.down_proj.weight": "pytorch_model-00024-of-00039.bin", - "model.layers.19.mlp.gate_proj.weight": "pytorch_model-00024-of-00039.bin", - "model.layers.19.mlp.up_proj.weight": "pytorch_model-00024-of-00039.bin", - "model.layers.19.post_attention_layernorm.weight": "pytorch_model-00024-of-00039.bin", - "model.layers.19.self_attn.k_proj.weight": "pytorch_model-00023-of-00039.bin", - "model.layers.19.self_attn.o_proj.weight": "pytorch_model-00024-of-00039.bin", - "model.layers.19.self_attn.q_proj.weight": "pytorch_model-00023-of-00039.bin", - "model.layers.19.self_attn.rotary_emb.inv_freq": "pytorch_model-00024-of-00039.bin", - "model.layers.19.self_attn.v_proj.weight": "pytorch_model-00023-of-00039.bin", - "model.layers.2.input_layernorm.weight": "pytorch_model-00004-of-00039.bin", - "model.layers.2.mlp.down_proj.weight": "pytorch_model-00004-of-00039.bin", - "model.layers.2.mlp.gate_proj.weight": "pytorch_model-00004-of-00039.bin", - "model.layers.2.mlp.up_proj.weight": "pytorch_model-00004-of-00039.bin", - "model.layers.2.post_attention_layernorm.weight": "pytorch_model-00004-of-00039.bin", - "model.layers.2.self_attn.k_proj.weight": "pytorch_model-00003-of-00039.bin", - "model.layers.2.self_attn.o_proj.weight": "pytorch_model-00004-of-00039.bin", - "model.layers.2.self_attn.q_proj.weight": "pytorch_model-00003-of-00039.bin", - "model.layers.2.self_attn.rotary_emb.inv_freq": "pytorch_model-00004-of-00039.bin", - "model.layers.2.self_attn.v_proj.weight": "pytorch_model-00004-of-00039.bin", - "model.layers.20.input_layernorm.weight": "pytorch_model-00025-of-00039.bin", - "model.layers.20.mlp.down_proj.weight": "pytorch_model-00025-of-00039.bin", - "model.layers.20.mlp.gate_proj.weight": "pytorch_model-00025-of-00039.bin", - "model.layers.20.mlp.up_proj.weight": "pytorch_model-00025-of-00039.bin", - "model.layers.20.post_attention_layernorm.weight": "pytorch_model-00025-of-00039.bin", - "model.layers.20.self_attn.k_proj.weight": "pytorch_model-00024-of-00039.bin", - "model.layers.20.self_attn.o_proj.weight": "pytorch_model-00025-of-00039.bin", - "model.layers.20.self_attn.q_proj.weight": "pytorch_model-00024-of-00039.bin", - "model.layers.20.self_attn.rotary_emb.inv_freq": "pytorch_model-00025-of-00039.bin", - "model.layers.20.self_attn.v_proj.weight": "pytorch_model-00025-of-00039.bin", - "model.layers.21.input_layernorm.weight": "pytorch_model-00026-of-00039.bin", - "model.layers.21.mlp.down_proj.weight": "pytorch_model-00026-of-00039.bin", - "model.layers.21.mlp.gate_proj.weight": "pytorch_model-00026-of-00039.bin", - "model.layers.21.mlp.up_proj.weight": "pytorch_model-00026-of-00039.bin", - "model.layers.21.post_attention_layernorm.weight": "pytorch_model-00026-of-00039.bin", - "model.layers.21.self_attn.k_proj.weight": "pytorch_model-00026-of-00039.bin", - "model.layers.21.self_attn.o_proj.weight": "pytorch_model-00026-of-00039.bin", - "model.layers.21.self_attn.q_proj.weight": "pytorch_model-00025-of-00039.bin", - "model.layers.21.self_attn.rotary_emb.inv_freq": "pytorch_model-00026-of-00039.bin", - "model.layers.21.self_attn.v_proj.weight": "pytorch_model-00026-of-00039.bin", - "model.layers.22.input_layernorm.weight": "pytorch_model-00028-of-00039.bin", - "model.layers.22.mlp.down_proj.weight": "pytorch_model-00027-of-00039.bin", - "model.layers.22.mlp.gate_proj.weight": "pytorch_model-00027-of-00039.bin", - "model.layers.22.mlp.up_proj.weight": "pytorch_model-00028-of-00039.bin", - "model.layers.22.post_attention_layernorm.weight": "pytorch_model-00028-of-00039.bin", - "model.layers.22.self_attn.k_proj.weight": "pytorch_model-00027-of-00039.bin", - "model.layers.22.self_attn.o_proj.weight": "pytorch_model-00027-of-00039.bin", - "model.layers.22.self_attn.q_proj.weight": "pytorch_model-00027-of-00039.bin", - "model.layers.22.self_attn.rotary_emb.inv_freq": "pytorch_model-00027-of-00039.bin", - "model.layers.22.self_attn.v_proj.weight": "pytorch_model-00027-of-00039.bin", - "model.layers.23.input_layernorm.weight": "pytorch_model-00029-of-00039.bin", - "model.layers.23.mlp.down_proj.weight": "pytorch_model-00029-of-00039.bin", - "model.layers.23.mlp.gate_proj.weight": "pytorch_model-00028-of-00039.bin", - "model.layers.23.mlp.up_proj.weight": "pytorch_model-00029-of-00039.bin", - "model.layers.23.post_attention_layernorm.weight": "pytorch_model-00029-of-00039.bin", - "model.layers.23.self_attn.k_proj.weight": "pytorch_model-00028-of-00039.bin", - "model.layers.23.self_attn.o_proj.weight": "pytorch_model-00028-of-00039.bin", - "model.layers.23.self_attn.q_proj.weight": "pytorch_model-00028-of-00039.bin", - "model.layers.23.self_attn.rotary_emb.inv_freq": "pytorch_model-00028-of-00039.bin", - "model.layers.23.self_attn.v_proj.weight": "pytorch_model-00028-of-00039.bin", - "model.layers.24.input_layernorm.weight": "pytorch_model-00030-of-00039.bin", - "model.layers.24.mlp.down_proj.weight": "pytorch_model-00030-of-00039.bin", - "model.layers.24.mlp.gate_proj.weight": "pytorch_model-00030-of-00039.bin", - "model.layers.24.mlp.up_proj.weight": "pytorch_model-00030-of-00039.bin", - "model.layers.24.post_attention_layernorm.weight": "pytorch_model-00030-of-00039.bin", - "model.layers.24.self_attn.k_proj.weight": "pytorch_model-00029-of-00039.bin", - "model.layers.24.self_attn.o_proj.weight": "pytorch_model-00029-of-00039.bin", - "model.layers.24.self_attn.q_proj.weight": "pytorch_model-00029-of-00039.bin", - "model.layers.24.self_attn.rotary_emb.inv_freq": "pytorch_model-00029-of-00039.bin", - "model.layers.24.self_attn.v_proj.weight": "pytorch_model-00029-of-00039.bin", - "model.layers.25.input_layernorm.weight": "pytorch_model-00031-of-00039.bin", - "model.layers.25.mlp.down_proj.weight": "pytorch_model-00031-of-00039.bin", - "model.layers.25.mlp.gate_proj.weight": "pytorch_model-00031-of-00039.bin", - "model.layers.25.mlp.up_proj.weight": "pytorch_model-00031-of-00039.bin", - "model.layers.25.post_attention_layernorm.weight": "pytorch_model-00031-of-00039.bin", - "model.layers.25.self_attn.k_proj.weight": "pytorch_model-00030-of-00039.bin", - "model.layers.25.self_attn.o_proj.weight": "pytorch_model-00031-of-00039.bin", - "model.layers.25.self_attn.q_proj.weight": "pytorch_model-00030-of-00039.bin", - "model.layers.25.self_attn.rotary_emb.inv_freq": "pytorch_model-00031-of-00039.bin", - "model.layers.25.self_attn.v_proj.weight": "pytorch_model-00030-of-00039.bin", - "model.layers.26.input_layernorm.weight": "pytorch_model-00032-of-00039.bin", - "model.layers.26.mlp.down_proj.weight": "pytorch_model-00032-of-00039.bin", - "model.layers.26.mlp.gate_proj.weight": "pytorch_model-00032-of-00039.bin", - "model.layers.26.mlp.up_proj.weight": "pytorch_model-00032-of-00039.bin", - "model.layers.26.post_attention_layernorm.weight": "pytorch_model-00032-of-00039.bin", - "model.layers.26.self_attn.k_proj.weight": "pytorch_model-00031-of-00039.bin", - "model.layers.26.self_attn.o_proj.weight": "pytorch_model-00032-of-00039.bin", - "model.layers.26.self_attn.q_proj.weight": "pytorch_model-00031-of-00039.bin", - "model.layers.26.self_attn.rotary_emb.inv_freq": "pytorch_model-00032-of-00039.bin", - "model.layers.26.self_attn.v_proj.weight": "pytorch_model-00032-of-00039.bin", - "model.layers.27.input_layernorm.weight": "pytorch_model-00033-of-00039.bin", - "model.layers.27.mlp.down_proj.weight": "pytorch_model-00033-of-00039.bin", - "model.layers.27.mlp.gate_proj.weight": "pytorch_model-00033-of-00039.bin", - "model.layers.27.mlp.up_proj.weight": "pytorch_model-00033-of-00039.bin", - "model.layers.27.post_attention_layernorm.weight": "pytorch_model-00033-of-00039.bin", - "model.layers.27.self_attn.k_proj.weight": "pytorch_model-00033-of-00039.bin", - "model.layers.27.self_attn.o_proj.weight": "pytorch_model-00033-of-00039.bin", - "model.layers.27.self_attn.q_proj.weight": "pytorch_model-00032-of-00039.bin", - "model.layers.27.self_attn.rotary_emb.inv_freq": "pytorch_model-00033-of-00039.bin", - "model.layers.27.self_attn.v_proj.weight": "pytorch_model-00033-of-00039.bin", - "model.layers.28.input_layernorm.weight": "pytorch_model-00035-of-00039.bin", - "model.layers.28.mlp.down_proj.weight": "pytorch_model-00034-of-00039.bin", - "model.layers.28.mlp.gate_proj.weight": "pytorch_model-00034-of-00039.bin", - "model.layers.28.mlp.up_proj.weight": "pytorch_model-00035-of-00039.bin", - "model.layers.28.post_attention_layernorm.weight": "pytorch_model-00035-of-00039.bin", - "model.layers.28.self_attn.k_proj.weight": "pytorch_model-00034-of-00039.bin", - "model.layers.28.self_attn.o_proj.weight": "pytorch_model-00034-of-00039.bin", - "model.layers.28.self_attn.q_proj.weight": "pytorch_model-00034-of-00039.bin", - "model.layers.28.self_attn.rotary_emb.inv_freq": "pytorch_model-00034-of-00039.bin", - "model.layers.28.self_attn.v_proj.weight": "pytorch_model-00034-of-00039.bin", - "model.layers.29.input_layernorm.weight": "pytorch_model-00036-of-00039.bin", - "model.layers.29.mlp.down_proj.weight": "pytorch_model-00036-of-00039.bin", - "model.layers.29.mlp.gate_proj.weight": "pytorch_model-00035-of-00039.bin", - "model.layers.29.mlp.up_proj.weight": "pytorch_model-00036-of-00039.bin", - "model.layers.29.post_attention_layernorm.weight": "pytorch_model-00036-of-00039.bin", - "model.layers.29.self_attn.k_proj.weight": "pytorch_model-00035-of-00039.bin", - "model.layers.29.self_attn.o_proj.weight": "pytorch_model-00035-of-00039.bin", - "model.layers.29.self_attn.q_proj.weight": "pytorch_model-00035-of-00039.bin", - "model.layers.29.self_attn.rotary_emb.inv_freq": "pytorch_model-00035-of-00039.bin", - "model.layers.29.self_attn.v_proj.weight": "pytorch_model-00035-of-00039.bin", - "model.layers.3.input_layernorm.weight": "pytorch_model-00005-of-00039.bin", - "model.layers.3.mlp.down_proj.weight": "pytorch_model-00005-of-00039.bin", - "model.layers.3.mlp.gate_proj.weight": "pytorch_model-00005-of-00039.bin", - "model.layers.3.mlp.up_proj.weight": "pytorch_model-00005-of-00039.bin", - "model.layers.3.post_attention_layernorm.weight": "pytorch_model-00005-of-00039.bin", - "model.layers.3.self_attn.k_proj.weight": "pytorch_model-00005-of-00039.bin", - "model.layers.3.self_attn.o_proj.weight": "pytorch_model-00005-of-00039.bin", - "model.layers.3.self_attn.q_proj.weight": "pytorch_model-00004-of-00039.bin", - "model.layers.3.self_attn.rotary_emb.inv_freq": "pytorch_model-00005-of-00039.bin", - "model.layers.3.self_attn.v_proj.weight": "pytorch_model-00005-of-00039.bin", - "model.layers.30.input_layernorm.weight": "pytorch_model-00037-of-00039.bin", - "model.layers.30.mlp.down_proj.weight": "pytorch_model-00037-of-00039.bin", - "model.layers.30.mlp.gate_proj.weight": "pytorch_model-00037-of-00039.bin", - "model.layers.30.mlp.up_proj.weight": "pytorch_model-00037-of-00039.bin", - "model.layers.30.post_attention_layernorm.weight": "pytorch_model-00037-of-00039.bin", - "model.layers.30.self_attn.k_proj.weight": "pytorch_model-00036-of-00039.bin", - "model.layers.30.self_attn.o_proj.weight": "pytorch_model-00036-of-00039.bin", - "model.layers.30.self_attn.q_proj.weight": "pytorch_model-00036-of-00039.bin", - "model.layers.30.self_attn.rotary_emb.inv_freq": "pytorch_model-00036-of-00039.bin", - "model.layers.30.self_attn.v_proj.weight": "pytorch_model-00036-of-00039.bin", - "model.layers.31.input_layernorm.weight": "pytorch_model-00038-of-00039.bin", - "model.layers.31.mlp.down_proj.weight": "pytorch_model-00038-of-00039.bin", - "model.layers.31.mlp.gate_proj.weight": "pytorch_model-00038-of-00039.bin", - "model.layers.31.mlp.up_proj.weight": "pytorch_model-00038-of-00039.bin", - "model.layers.31.post_attention_layernorm.weight": "pytorch_model-00038-of-00039.bin", - "model.layers.31.self_attn.k_proj.weight": "pytorch_model-00037-of-00039.bin", - "model.layers.31.self_attn.o_proj.weight": "pytorch_model-00038-of-00039.bin", - "model.layers.31.self_attn.q_proj.weight": "pytorch_model-00037-of-00039.bin", - "model.layers.31.self_attn.rotary_emb.inv_freq": "pytorch_model-00038-of-00039.bin", - "model.layers.31.self_attn.v_proj.weight": "pytorch_model-00037-of-00039.bin", - "model.layers.4.input_layernorm.weight": "pytorch_model-00007-of-00039.bin", - "model.layers.4.mlp.down_proj.weight": "pytorch_model-00006-of-00039.bin", - "model.layers.4.mlp.gate_proj.weight": "pytorch_model-00006-of-00039.bin", - "model.layers.4.mlp.up_proj.weight": "pytorch_model-00007-of-00039.bin", - "model.layers.4.post_attention_layernorm.weight": "pytorch_model-00007-of-00039.bin", - "model.layers.4.self_attn.k_proj.weight": "pytorch_model-00006-of-00039.bin", - "model.layers.4.self_attn.o_proj.weight": "pytorch_model-00006-of-00039.bin", - "model.layers.4.self_attn.q_proj.weight": "pytorch_model-00006-of-00039.bin", - "model.layers.4.self_attn.rotary_emb.inv_freq": "pytorch_model-00006-of-00039.bin", - "model.layers.4.self_attn.v_proj.weight": "pytorch_model-00006-of-00039.bin", - "model.layers.5.input_layernorm.weight": "pytorch_model-00008-of-00039.bin", - "model.layers.5.mlp.down_proj.weight": "pytorch_model-00008-of-00039.bin", - "model.layers.5.mlp.gate_proj.weight": "pytorch_model-00007-of-00039.bin", - "model.layers.5.mlp.up_proj.weight": "pytorch_model-00008-of-00039.bin", - "model.layers.5.post_attention_layernorm.weight": "pytorch_model-00008-of-00039.bin", - "model.layers.5.self_attn.k_proj.weight": "pytorch_model-00007-of-00039.bin", - "model.layers.5.self_attn.o_proj.weight": "pytorch_model-00007-of-00039.bin", - "model.layers.5.self_attn.q_proj.weight": "pytorch_model-00007-of-00039.bin", - "model.layers.5.self_attn.rotary_emb.inv_freq": "pytorch_model-00007-of-00039.bin", - "model.layers.5.self_attn.v_proj.weight": "pytorch_model-00007-of-00039.bin", - "model.layers.6.input_layernorm.weight": "pytorch_model-00009-of-00039.bin", - "model.layers.6.mlp.down_proj.weight": "pytorch_model-00009-of-00039.bin", - "model.layers.6.mlp.gate_proj.weight": "pytorch_model-00009-of-00039.bin", - "model.layers.6.mlp.up_proj.weight": "pytorch_model-00009-of-00039.bin", - "model.layers.6.post_attention_layernorm.weight": "pytorch_model-00009-of-00039.bin", - "model.layers.6.self_attn.k_proj.weight": "pytorch_model-00008-of-00039.bin", - "model.layers.6.self_attn.o_proj.weight": "pytorch_model-00008-of-00039.bin", - "model.layers.6.self_attn.q_proj.weight": "pytorch_model-00008-of-00039.bin", - "model.layers.6.self_attn.rotary_emb.inv_freq": "pytorch_model-00008-of-00039.bin", - "model.layers.6.self_attn.v_proj.weight": "pytorch_model-00008-of-00039.bin", - "model.layers.7.input_layernorm.weight": "pytorch_model-00010-of-00039.bin", - "model.layers.7.mlp.down_proj.weight": "pytorch_model-00010-of-00039.bin", - "model.layers.7.mlp.gate_proj.weight": "pytorch_model-00010-of-00039.bin", - "model.layers.7.mlp.up_proj.weight": "pytorch_model-00010-of-00039.bin", - "model.layers.7.post_attention_layernorm.weight": "pytorch_model-00010-of-00039.bin", - "model.layers.7.self_attn.k_proj.weight": "pytorch_model-00009-of-00039.bin", - "model.layers.7.self_attn.o_proj.weight": "pytorch_model-00010-of-00039.bin", - "model.layers.7.self_attn.q_proj.weight": "pytorch_model-00009-of-00039.bin", - "model.layers.7.self_attn.rotary_emb.inv_freq": "pytorch_model-00010-of-00039.bin", - "model.layers.7.self_attn.v_proj.weight": "pytorch_model-00009-of-00039.bin", - "model.layers.8.input_layernorm.weight": "pytorch_model-00011-of-00039.bin", - "model.layers.8.mlp.down_proj.weight": "pytorch_model-00011-of-00039.bin", - "model.layers.8.mlp.gate_proj.weight": "pytorch_model-00011-of-00039.bin", - "model.layers.8.mlp.up_proj.weight": "pytorch_model-00011-of-00039.bin", - "model.layers.8.post_attention_layernorm.weight": "pytorch_model-00011-of-00039.bin", - "model.layers.8.self_attn.k_proj.weight": "pytorch_model-00010-of-00039.bin", - "model.layers.8.self_attn.o_proj.weight": "pytorch_model-00011-of-00039.bin", - "model.layers.8.self_attn.q_proj.weight": "pytorch_model-00010-of-00039.bin", - "model.layers.8.self_attn.rotary_emb.inv_freq": "pytorch_model-00011-of-00039.bin", - "model.layers.8.self_attn.v_proj.weight": "pytorch_model-00011-of-00039.bin", - "model.layers.9.input_layernorm.weight": "pytorch_model-00012-of-00039.bin", - "model.layers.9.mlp.down_proj.weight": "pytorch_model-00012-of-00039.bin", - "model.layers.9.mlp.gate_proj.weight": "pytorch_model-00012-of-00039.bin", - "model.layers.9.mlp.up_proj.weight": "pytorch_model-00012-of-00039.bin", - "model.layers.9.post_attention_layernorm.weight": "pytorch_model-00012-of-00039.bin", - "model.layers.9.self_attn.k_proj.weight": "pytorch_model-00012-of-00039.bin", - "model.layers.9.self_attn.o_proj.weight": "pytorch_model-00012-of-00039.bin", - "model.layers.9.self_attn.q_proj.weight": "pytorch_model-00011-of-00039.bin", - "model.layers.9.self_attn.rotary_emb.inv_freq": "pytorch_model-00012-of-00039.bin", - "model.layers.9.self_attn.v_proj.weight": "pytorch_model-00012-of-00039.bin", - "model.norm.weight": "pytorch_model-00038-of-00039.bin" + "lm_head.weight": "pytorch_model-00041-of-00041.bin", + "model.embed_tokens.weight": "pytorch_model-00002-of-00041.bin", + "model.layers.0.input_layernorm.weight": "pytorch_model-00004-of-00041.bin", + "model.layers.0.mlp.down_proj.weight": "pytorch_model-00003-of-00041.bin", + "model.layers.0.mlp.gate_proj.weight": "pytorch_model-00003-of-00041.bin", + "model.layers.0.mlp.up_proj.weight": "pytorch_model-00004-of-00041.bin", + "model.layers.0.post_attention_layernorm.weight": "pytorch_model-00004-of-00041.bin", + "model.layers.0.self_attn.k_proj.weight": "pytorch_model-00003-of-00041.bin", + "model.layers.0.self_attn.o_proj.weight": "pytorch_model-00003-of-00041.bin", + "model.layers.0.self_attn.q_proj.weight": "pytorch_model-00003-of-00041.bin", + "model.layers.0.self_attn.rotary_emb.inv_freq": "pytorch_model-00003-of-00041.bin", + "model.layers.0.self_attn.v_proj.weight": "pytorch_model-00003-of-00041.bin", + "model.layers.1.input_layernorm.weight": "pytorch_model-00005-of-00041.bin", + "model.layers.1.mlp.down_proj.weight": "pytorch_model-00005-of-00041.bin", + "model.layers.1.mlp.gate_proj.weight": "pytorch_model-00004-of-00041.bin", + "model.layers.1.mlp.up_proj.weight": "pytorch_model-00005-of-00041.bin", + "model.layers.1.post_attention_layernorm.weight": "pytorch_model-00005-of-00041.bin", + "model.layers.1.self_attn.k_proj.weight": "pytorch_model-00004-of-00041.bin", + "model.layers.1.self_attn.o_proj.weight": "pytorch_model-00004-of-00041.bin", + "model.layers.1.self_attn.q_proj.weight": "pytorch_model-00004-of-00041.bin", + "model.layers.1.self_attn.rotary_emb.inv_freq": "pytorch_model-00004-of-00041.bin", + "model.layers.1.self_attn.v_proj.weight": "pytorch_model-00004-of-00041.bin", + "model.layers.10.input_layernorm.weight": "pytorch_model-00015-of-00041.bin", + "model.layers.10.mlp.down_proj.weight": "pytorch_model-00015-of-00041.bin", + "model.layers.10.mlp.gate_proj.weight": "pytorch_model-00015-of-00041.bin", + "model.layers.10.mlp.up_proj.weight": "pytorch_model-00015-of-00041.bin", + "model.layers.10.post_attention_layernorm.weight": "pytorch_model-00015-of-00041.bin", + "model.layers.10.self_attn.k_proj.weight": "pytorch_model-00014-of-00041.bin", + "model.layers.10.self_attn.o_proj.weight": "pytorch_model-00015-of-00041.bin", + "model.layers.10.self_attn.q_proj.weight": "pytorch_model-00014-of-00041.bin", + "model.layers.10.self_attn.rotary_emb.inv_freq": "pytorch_model-00015-of-00041.bin", + "model.layers.10.self_attn.v_proj.weight": "pytorch_model-00015-of-00041.bin", + "model.layers.11.input_layernorm.weight": "pytorch_model-00016-of-00041.bin", + "model.layers.11.mlp.down_proj.weight": "pytorch_model-00016-of-00041.bin", + "model.layers.11.mlp.gate_proj.weight": "pytorch_model-00016-of-00041.bin", + "model.layers.11.mlp.up_proj.weight": "pytorch_model-00016-of-00041.bin", + "model.layers.11.post_attention_layernorm.weight": "pytorch_model-00016-of-00041.bin", + "model.layers.11.self_attn.k_proj.weight": "pytorch_model-00016-of-00041.bin", + "model.layers.11.self_attn.o_proj.weight": "pytorch_model-00016-of-00041.bin", + "model.layers.11.self_attn.q_proj.weight": "pytorch_model-00015-of-00041.bin", + "model.layers.11.self_attn.rotary_emb.inv_freq": "pytorch_model-00016-of-00041.bin", + "model.layers.11.self_attn.v_proj.weight": "pytorch_model-00016-of-00041.bin", + "model.layers.12.input_layernorm.weight": "pytorch_model-00018-of-00041.bin", + "model.layers.12.mlp.down_proj.weight": "pytorch_model-00017-of-00041.bin", + "model.layers.12.mlp.gate_proj.weight": "pytorch_model-00017-of-00041.bin", + "model.layers.12.mlp.up_proj.weight": "pytorch_model-00018-of-00041.bin", + "model.layers.12.post_attention_layernorm.weight": "pytorch_model-00018-of-00041.bin", + "model.layers.12.self_attn.k_proj.weight": "pytorch_model-00017-of-00041.bin", + "model.layers.12.self_attn.o_proj.weight": "pytorch_model-00017-of-00041.bin", + "model.layers.12.self_attn.q_proj.weight": "pytorch_model-00017-of-00041.bin", + "model.layers.12.self_attn.rotary_emb.inv_freq": "pytorch_model-00017-of-00041.bin", + "model.layers.12.self_attn.v_proj.weight": "pytorch_model-00017-of-00041.bin", + "model.layers.13.input_layernorm.weight": "pytorch_model-00019-of-00041.bin", + "model.layers.13.mlp.down_proj.weight": "pytorch_model-00019-of-00041.bin", + "model.layers.13.mlp.gate_proj.weight": "pytorch_model-00018-of-00041.bin", + "model.layers.13.mlp.up_proj.weight": "pytorch_model-00019-of-00041.bin", + "model.layers.13.post_attention_layernorm.weight": "pytorch_model-00019-of-00041.bin", + "model.layers.13.self_attn.k_proj.weight": "pytorch_model-00018-of-00041.bin", + "model.layers.13.self_attn.o_proj.weight": "pytorch_model-00018-of-00041.bin", + "model.layers.13.self_attn.q_proj.weight": "pytorch_model-00018-of-00041.bin", + "model.layers.13.self_attn.rotary_emb.inv_freq": "pytorch_model-00018-of-00041.bin", + "model.layers.13.self_attn.v_proj.weight": "pytorch_model-00018-of-00041.bin", + "model.layers.14.input_layernorm.weight": "pytorch_model-00020-of-00041.bin", + "model.layers.14.mlp.down_proj.weight": "pytorch_model-00020-of-00041.bin", + "model.layers.14.mlp.gate_proj.weight": "pytorch_model-00020-of-00041.bin", + "model.layers.14.mlp.up_proj.weight": "pytorch_model-00020-of-00041.bin", + "model.layers.14.post_attention_layernorm.weight": "pytorch_model-00020-of-00041.bin", + "model.layers.14.self_attn.k_proj.weight": "pytorch_model-00019-of-00041.bin", + "model.layers.14.self_attn.o_proj.weight": "pytorch_model-00019-of-00041.bin", + "model.layers.14.self_attn.q_proj.weight": "pytorch_model-00019-of-00041.bin", + "model.layers.14.self_attn.rotary_emb.inv_freq": "pytorch_model-00019-of-00041.bin", + "model.layers.14.self_attn.v_proj.weight": "pytorch_model-00019-of-00041.bin", + "model.layers.15.input_layernorm.weight": "pytorch_model-00021-of-00041.bin", + "model.layers.15.mlp.down_proj.weight": "pytorch_model-00021-of-00041.bin", + "model.layers.15.mlp.gate_proj.weight": "pytorch_model-00021-of-00041.bin", + "model.layers.15.mlp.up_proj.weight": "pytorch_model-00021-of-00041.bin", + "model.layers.15.post_attention_layernorm.weight": "pytorch_model-00021-of-00041.bin", + "model.layers.15.self_attn.k_proj.weight": "pytorch_model-00020-of-00041.bin", + "model.layers.15.self_attn.o_proj.weight": "pytorch_model-00021-of-00041.bin", + "model.layers.15.self_attn.q_proj.weight": "pytorch_model-00020-of-00041.bin", + "model.layers.15.self_attn.rotary_emb.inv_freq": "pytorch_model-00021-of-00041.bin", + "model.layers.15.self_attn.v_proj.weight": "pytorch_model-00020-of-00041.bin", + "model.layers.16.input_layernorm.weight": "pytorch_model-00022-of-00041.bin", + "model.layers.16.mlp.down_proj.weight": "pytorch_model-00022-of-00041.bin", + "model.layers.16.mlp.gate_proj.weight": "pytorch_model-00022-of-00041.bin", + "model.layers.16.mlp.up_proj.weight": "pytorch_model-00022-of-00041.bin", + "model.layers.16.post_attention_layernorm.weight": "pytorch_model-00022-of-00041.bin", + "model.layers.16.self_attn.k_proj.weight": "pytorch_model-00021-of-00041.bin", + "model.layers.16.self_attn.o_proj.weight": "pytorch_model-00022-of-00041.bin", + "model.layers.16.self_attn.q_proj.weight": "pytorch_model-00021-of-00041.bin", + "model.layers.16.self_attn.rotary_emb.inv_freq": "pytorch_model-00022-of-00041.bin", + "model.layers.16.self_attn.v_proj.weight": "pytorch_model-00022-of-00041.bin", + "model.layers.17.input_layernorm.weight": "pytorch_model-00023-of-00041.bin", + "model.layers.17.mlp.down_proj.weight": "pytorch_model-00023-of-00041.bin", + "model.layers.17.mlp.gate_proj.weight": "pytorch_model-00023-of-00041.bin", + "model.layers.17.mlp.up_proj.weight": "pytorch_model-00023-of-00041.bin", + "model.layers.17.post_attention_layernorm.weight": "pytorch_model-00023-of-00041.bin", + "model.layers.17.self_attn.k_proj.weight": "pytorch_model-00023-of-00041.bin", + "model.layers.17.self_attn.o_proj.weight": "pytorch_model-00023-of-00041.bin", + "model.layers.17.self_attn.q_proj.weight": "pytorch_model-00022-of-00041.bin", + "model.layers.17.self_attn.rotary_emb.inv_freq": "pytorch_model-00023-of-00041.bin", + "model.layers.17.self_attn.v_proj.weight": "pytorch_model-00023-of-00041.bin", + "model.layers.18.input_layernorm.weight": "pytorch_model-00025-of-00041.bin", + "model.layers.18.mlp.down_proj.weight": "pytorch_model-00024-of-00041.bin", + "model.layers.18.mlp.gate_proj.weight": "pytorch_model-00024-of-00041.bin", + "model.layers.18.mlp.up_proj.weight": "pytorch_model-00025-of-00041.bin", + "model.layers.18.post_attention_layernorm.weight": "pytorch_model-00025-of-00041.bin", + "model.layers.18.self_attn.k_proj.weight": "pytorch_model-00024-of-00041.bin", + "model.layers.18.self_attn.o_proj.weight": "pytorch_model-00024-of-00041.bin", + "model.layers.18.self_attn.q_proj.weight": "pytorch_model-00024-of-00041.bin", + "model.layers.18.self_attn.rotary_emb.inv_freq": "pytorch_model-00024-of-00041.bin", + "model.layers.18.self_attn.v_proj.weight": "pytorch_model-00024-of-00041.bin", + "model.layers.19.input_layernorm.weight": "pytorch_model-00026-of-00041.bin", + "model.layers.19.mlp.down_proj.weight": "pytorch_model-00026-of-00041.bin", + "model.layers.19.mlp.gate_proj.weight": "pytorch_model-00025-of-00041.bin", + "model.layers.19.mlp.up_proj.weight": "pytorch_model-00026-of-00041.bin", + "model.layers.19.post_attention_layernorm.weight": "pytorch_model-00026-of-00041.bin", + "model.layers.19.self_attn.k_proj.weight": "pytorch_model-00025-of-00041.bin", + "model.layers.19.self_attn.o_proj.weight": "pytorch_model-00025-of-00041.bin", + "model.layers.19.self_attn.q_proj.weight": "pytorch_model-00025-of-00041.bin", + "model.layers.19.self_attn.rotary_emb.inv_freq": "pytorch_model-00025-of-00041.bin", + "model.layers.19.self_attn.v_proj.weight": "pytorch_model-00025-of-00041.bin", + "model.layers.2.input_layernorm.weight": "pytorch_model-00006-of-00041.bin", + "model.layers.2.mlp.down_proj.weight": "pytorch_model-00006-of-00041.bin", + "model.layers.2.mlp.gate_proj.weight": "pytorch_model-00006-of-00041.bin", + "model.layers.2.mlp.up_proj.weight": "pytorch_model-00006-of-00041.bin", + "model.layers.2.post_attention_layernorm.weight": "pytorch_model-00006-of-00041.bin", + "model.layers.2.self_attn.k_proj.weight": "pytorch_model-00005-of-00041.bin", + "model.layers.2.self_attn.o_proj.weight": "pytorch_model-00005-of-00041.bin", + "model.layers.2.self_attn.q_proj.weight": "pytorch_model-00005-of-00041.bin", + "model.layers.2.self_attn.rotary_emb.inv_freq": "pytorch_model-00005-of-00041.bin", + "model.layers.2.self_attn.v_proj.weight": "pytorch_model-00005-of-00041.bin", + "model.layers.20.input_layernorm.weight": "pytorch_model-00027-of-00041.bin", + "model.layers.20.mlp.down_proj.weight": "pytorch_model-00027-of-00041.bin", + "model.layers.20.mlp.gate_proj.weight": "pytorch_model-00027-of-00041.bin", + "model.layers.20.mlp.up_proj.weight": "pytorch_model-00027-of-00041.bin", + "model.layers.20.post_attention_layernorm.weight": "pytorch_model-00027-of-00041.bin", + "model.layers.20.self_attn.k_proj.weight": "pytorch_model-00026-of-00041.bin", + "model.layers.20.self_attn.o_proj.weight": "pytorch_model-00026-of-00041.bin", + "model.layers.20.self_attn.q_proj.weight": "pytorch_model-00026-of-00041.bin", + "model.layers.20.self_attn.rotary_emb.inv_freq": "pytorch_model-00026-of-00041.bin", + "model.layers.20.self_attn.v_proj.weight": "pytorch_model-00026-of-00041.bin", + "model.layers.21.input_layernorm.weight": "pytorch_model-00028-of-00041.bin", + "model.layers.21.mlp.down_proj.weight": "pytorch_model-00028-of-00041.bin", + "model.layers.21.mlp.gate_proj.weight": "pytorch_model-00028-of-00041.bin", + "model.layers.21.mlp.up_proj.weight": "pytorch_model-00028-of-00041.bin", + "model.layers.21.post_attention_layernorm.weight": "pytorch_model-00028-of-00041.bin", + "model.layers.21.self_attn.k_proj.weight": "pytorch_model-00027-of-00041.bin", + "model.layers.21.self_attn.o_proj.weight": "pytorch_model-00028-of-00041.bin", + "model.layers.21.self_attn.q_proj.weight": "pytorch_model-00027-of-00041.bin", + "model.layers.21.self_attn.rotary_emb.inv_freq": "pytorch_model-00028-of-00041.bin", + "model.layers.21.self_attn.v_proj.weight": "pytorch_model-00027-of-00041.bin", + "model.layers.22.input_layernorm.weight": "pytorch_model-00029-of-00041.bin", + "model.layers.22.mlp.down_proj.weight": "pytorch_model-00029-of-00041.bin", + "model.layers.22.mlp.gate_proj.weight": "pytorch_model-00029-of-00041.bin", + "model.layers.22.mlp.up_proj.weight": "pytorch_model-00029-of-00041.bin", + "model.layers.22.post_attention_layernorm.weight": "pytorch_model-00029-of-00041.bin", + "model.layers.22.self_attn.k_proj.weight": "pytorch_model-00028-of-00041.bin", + "model.layers.22.self_attn.o_proj.weight": "pytorch_model-00029-of-00041.bin", + "model.layers.22.self_attn.q_proj.weight": "pytorch_model-00028-of-00041.bin", + "model.layers.22.self_attn.rotary_emb.inv_freq": "pytorch_model-00029-of-00041.bin", + "model.layers.22.self_attn.v_proj.weight": "pytorch_model-00029-of-00041.bin", + "model.layers.23.input_layernorm.weight": "pytorch_model-00030-of-00041.bin", + "model.layers.23.mlp.down_proj.weight": "pytorch_model-00030-of-00041.bin", + "model.layers.23.mlp.gate_proj.weight": "pytorch_model-00030-of-00041.bin", + "model.layers.23.mlp.up_proj.weight": "pytorch_model-00030-of-00041.bin", + "model.layers.23.post_attention_layernorm.weight": "pytorch_model-00030-of-00041.bin", + "model.layers.23.self_attn.k_proj.weight": "pytorch_model-00030-of-00041.bin", + "model.layers.23.self_attn.o_proj.weight": "pytorch_model-00030-of-00041.bin", + "model.layers.23.self_attn.q_proj.weight": "pytorch_model-00029-of-00041.bin", + "model.layers.23.self_attn.rotary_emb.inv_freq": "pytorch_model-00030-of-00041.bin", + "model.layers.23.self_attn.v_proj.weight": "pytorch_model-00030-of-00041.bin", + "model.layers.24.input_layernorm.weight": "pytorch_model-00032-of-00041.bin", + "model.layers.24.mlp.down_proj.weight": "pytorch_model-00031-of-00041.bin", + "model.layers.24.mlp.gate_proj.weight": "pytorch_model-00031-of-00041.bin", + "model.layers.24.mlp.up_proj.weight": "pytorch_model-00032-of-00041.bin", + "model.layers.24.post_attention_layernorm.weight": "pytorch_model-00032-of-00041.bin", + "model.layers.24.self_attn.k_proj.weight": "pytorch_model-00031-of-00041.bin", + "model.layers.24.self_attn.o_proj.weight": "pytorch_model-00031-of-00041.bin", + "model.layers.24.self_attn.q_proj.weight": "pytorch_model-00031-of-00041.bin", + "model.layers.24.self_attn.rotary_emb.inv_freq": "pytorch_model-00031-of-00041.bin", + "model.layers.24.self_attn.v_proj.weight": "pytorch_model-00031-of-00041.bin", + "model.layers.25.input_layernorm.weight": "pytorch_model-00033-of-00041.bin", + "model.layers.25.mlp.down_proj.weight": "pytorch_model-00033-of-00041.bin", + "model.layers.25.mlp.gate_proj.weight": "pytorch_model-00032-of-00041.bin", + "model.layers.25.mlp.up_proj.weight": "pytorch_model-00033-of-00041.bin", + "model.layers.25.post_attention_layernorm.weight": "pytorch_model-00033-of-00041.bin", + "model.layers.25.self_attn.k_proj.weight": "pytorch_model-00032-of-00041.bin", + "model.layers.25.self_attn.o_proj.weight": "pytorch_model-00032-of-00041.bin", + "model.layers.25.self_attn.q_proj.weight": "pytorch_model-00032-of-00041.bin", + "model.layers.25.self_attn.rotary_emb.inv_freq": "pytorch_model-00032-of-00041.bin", + "model.layers.25.self_attn.v_proj.weight": "pytorch_model-00032-of-00041.bin", + "model.layers.26.input_layernorm.weight": "pytorch_model-00034-of-00041.bin", + "model.layers.26.mlp.down_proj.weight": "pytorch_model-00034-of-00041.bin", + "model.layers.26.mlp.gate_proj.weight": "pytorch_model-00034-of-00041.bin", + "model.layers.26.mlp.up_proj.weight": "pytorch_model-00034-of-00041.bin", + "model.layers.26.post_attention_layernorm.weight": "pytorch_model-00034-of-00041.bin", + "model.layers.26.self_attn.k_proj.weight": "pytorch_model-00033-of-00041.bin", + "model.layers.26.self_attn.o_proj.weight": "pytorch_model-00033-of-00041.bin", + "model.layers.26.self_attn.q_proj.weight": "pytorch_model-00033-of-00041.bin", + "model.layers.26.self_attn.rotary_emb.inv_freq": "pytorch_model-00033-of-00041.bin", + "model.layers.26.self_attn.v_proj.weight": "pytorch_model-00033-of-00041.bin", + "model.layers.27.input_layernorm.weight": "pytorch_model-00035-of-00041.bin", + "model.layers.27.mlp.down_proj.weight": "pytorch_model-00035-of-00041.bin", + "model.layers.27.mlp.gate_proj.weight": "pytorch_model-00035-of-00041.bin", + "model.layers.27.mlp.up_proj.weight": "pytorch_model-00035-of-00041.bin", + "model.layers.27.post_attention_layernorm.weight": "pytorch_model-00035-of-00041.bin", + "model.layers.27.self_attn.k_proj.weight": "pytorch_model-00034-of-00041.bin", + "model.layers.27.self_attn.o_proj.weight": "pytorch_model-00035-of-00041.bin", + "model.layers.27.self_attn.q_proj.weight": "pytorch_model-00034-of-00041.bin", + "model.layers.27.self_attn.rotary_emb.inv_freq": "pytorch_model-00035-of-00041.bin", + "model.layers.27.self_attn.v_proj.weight": "pytorch_model-00034-of-00041.bin", + "model.layers.28.input_layernorm.weight": "pytorch_model-00036-of-00041.bin", + "model.layers.28.mlp.down_proj.weight": "pytorch_model-00036-of-00041.bin", + "model.layers.28.mlp.gate_proj.weight": "pytorch_model-00036-of-00041.bin", + "model.layers.28.mlp.up_proj.weight": "pytorch_model-00036-of-00041.bin", + "model.layers.28.post_attention_layernorm.weight": "pytorch_model-00036-of-00041.bin", + "model.layers.28.self_attn.k_proj.weight": "pytorch_model-00035-of-00041.bin", + "model.layers.28.self_attn.o_proj.weight": "pytorch_model-00036-of-00041.bin", + "model.layers.28.self_attn.q_proj.weight": "pytorch_model-00035-of-00041.bin", + "model.layers.28.self_attn.rotary_emb.inv_freq": "pytorch_model-00036-of-00041.bin", + "model.layers.28.self_attn.v_proj.weight": "pytorch_model-00036-of-00041.bin", + "model.layers.29.input_layernorm.weight": "pytorch_model-00037-of-00041.bin", + "model.layers.29.mlp.down_proj.weight": "pytorch_model-00037-of-00041.bin", + "model.layers.29.mlp.gate_proj.weight": "pytorch_model-00037-of-00041.bin", + "model.layers.29.mlp.up_proj.weight": "pytorch_model-00037-of-00041.bin", + "model.layers.29.post_attention_layernorm.weight": "pytorch_model-00037-of-00041.bin", + "model.layers.29.self_attn.k_proj.weight": "pytorch_model-00037-of-00041.bin", + "model.layers.29.self_attn.o_proj.weight": "pytorch_model-00037-of-00041.bin", + "model.layers.29.self_attn.q_proj.weight": "pytorch_model-00036-of-00041.bin", + "model.layers.29.self_attn.rotary_emb.inv_freq": "pytorch_model-00037-of-00041.bin", + "model.layers.29.self_attn.v_proj.weight": "pytorch_model-00037-of-00041.bin", + "model.layers.3.input_layernorm.weight": "pytorch_model-00007-of-00041.bin", + "model.layers.3.mlp.down_proj.weight": "pytorch_model-00007-of-00041.bin", + "model.layers.3.mlp.gate_proj.weight": "pytorch_model-00007-of-00041.bin", + "model.layers.3.mlp.up_proj.weight": "pytorch_model-00007-of-00041.bin", + "model.layers.3.post_attention_layernorm.weight": "pytorch_model-00007-of-00041.bin", + "model.layers.3.self_attn.k_proj.weight": "pytorch_model-00006-of-00041.bin", + "model.layers.3.self_attn.o_proj.weight": "pytorch_model-00007-of-00041.bin", + "model.layers.3.self_attn.q_proj.weight": "pytorch_model-00006-of-00041.bin", + "model.layers.3.self_attn.rotary_emb.inv_freq": "pytorch_model-00007-of-00041.bin", + "model.layers.3.self_attn.v_proj.weight": "pytorch_model-00006-of-00041.bin", + "model.layers.30.input_layernorm.weight": "pytorch_model-00039-of-00041.bin", + "model.layers.30.mlp.down_proj.weight": "pytorch_model-00038-of-00041.bin", + "model.layers.30.mlp.gate_proj.weight": "pytorch_model-00038-of-00041.bin", + "model.layers.30.mlp.up_proj.weight": "pytorch_model-00039-of-00041.bin", + "model.layers.30.post_attention_layernorm.weight": "pytorch_model-00039-of-00041.bin", + "model.layers.30.self_attn.k_proj.weight": "pytorch_model-00038-of-00041.bin", + "model.layers.30.self_attn.o_proj.weight": "pytorch_model-00038-of-00041.bin", + "model.layers.30.self_attn.q_proj.weight": "pytorch_model-00038-of-00041.bin", + "model.layers.30.self_attn.rotary_emb.inv_freq": "pytorch_model-00038-of-00041.bin", + "model.layers.30.self_attn.v_proj.weight": "pytorch_model-00038-of-00041.bin", + "model.layers.31.input_layernorm.weight": "pytorch_model-00040-of-00041.bin", + "model.layers.31.mlp.down_proj.weight": "pytorch_model-00040-of-00041.bin", + "model.layers.31.mlp.gate_proj.weight": "pytorch_model-00039-of-00041.bin", + "model.layers.31.mlp.up_proj.weight": "pytorch_model-00040-of-00041.bin", + "model.layers.31.post_attention_layernorm.weight": "pytorch_model-00040-of-00041.bin", + "model.layers.31.self_attn.k_proj.weight": "pytorch_model-00039-of-00041.bin", + "model.layers.31.self_attn.o_proj.weight": "pytorch_model-00039-of-00041.bin", + "model.layers.31.self_attn.q_proj.weight": "pytorch_model-00039-of-00041.bin", + "model.layers.31.self_attn.rotary_emb.inv_freq": "pytorch_model-00039-of-00041.bin", + "model.layers.31.self_attn.v_proj.weight": "pytorch_model-00039-of-00041.bin", + "model.layers.4.input_layernorm.weight": "pytorch_model-00008-of-00041.bin", + "model.layers.4.mlp.down_proj.weight": "pytorch_model-00008-of-00041.bin", + "model.layers.4.mlp.gate_proj.weight": "pytorch_model-00008-of-00041.bin", + "model.layers.4.mlp.up_proj.weight": "pytorch_model-00008-of-00041.bin", + "model.layers.4.post_attention_layernorm.weight": "pytorch_model-00008-of-00041.bin", + "model.layers.4.self_attn.k_proj.weight": "pytorch_model-00007-of-00041.bin", + "model.layers.4.self_attn.o_proj.weight": "pytorch_model-00008-of-00041.bin", + "model.layers.4.self_attn.q_proj.weight": "pytorch_model-00007-of-00041.bin", + "model.layers.4.self_attn.rotary_emb.inv_freq": "pytorch_model-00008-of-00041.bin", + "model.layers.4.self_attn.v_proj.weight": "pytorch_model-00008-of-00041.bin", + "model.layers.5.input_layernorm.weight": "pytorch_model-00009-of-00041.bin", + "model.layers.5.mlp.down_proj.weight": "pytorch_model-00009-of-00041.bin", + "model.layers.5.mlp.gate_proj.weight": "pytorch_model-00009-of-00041.bin", + "model.layers.5.mlp.up_proj.weight": "pytorch_model-00009-of-00041.bin", + "model.layers.5.post_attention_layernorm.weight": "pytorch_model-00009-of-00041.bin", + "model.layers.5.self_attn.k_proj.weight": "pytorch_model-00009-of-00041.bin", + "model.layers.5.self_attn.o_proj.weight": "pytorch_model-00009-of-00041.bin", + "model.layers.5.self_attn.q_proj.weight": "pytorch_model-00008-of-00041.bin", + "model.layers.5.self_attn.rotary_emb.inv_freq": "pytorch_model-00009-of-00041.bin", + "model.layers.5.self_attn.v_proj.weight": "pytorch_model-00009-of-00041.bin", + "model.layers.6.input_layernorm.weight": "pytorch_model-00011-of-00041.bin", + "model.layers.6.mlp.down_proj.weight": "pytorch_model-00010-of-00041.bin", + "model.layers.6.mlp.gate_proj.weight": "pytorch_model-00010-of-00041.bin", + "model.layers.6.mlp.up_proj.weight": "pytorch_model-00011-of-00041.bin", + "model.layers.6.post_attention_layernorm.weight": "pytorch_model-00011-of-00041.bin", + "model.layers.6.self_attn.k_proj.weight": "pytorch_model-00010-of-00041.bin", + "model.layers.6.self_attn.o_proj.weight": "pytorch_model-00010-of-00041.bin", + "model.layers.6.self_attn.q_proj.weight": "pytorch_model-00010-of-00041.bin", + "model.layers.6.self_attn.rotary_emb.inv_freq": "pytorch_model-00010-of-00041.bin", + "model.layers.6.self_attn.v_proj.weight": "pytorch_model-00010-of-00041.bin", + "model.layers.7.input_layernorm.weight": "pytorch_model-00012-of-00041.bin", + "model.layers.7.mlp.down_proj.weight": "pytorch_model-00012-of-00041.bin", + "model.layers.7.mlp.gate_proj.weight": "pytorch_model-00011-of-00041.bin", + "model.layers.7.mlp.up_proj.weight": "pytorch_model-00012-of-00041.bin", + "model.layers.7.post_attention_layernorm.weight": "pytorch_model-00012-of-00041.bin", + "model.layers.7.self_attn.k_proj.weight": "pytorch_model-00011-of-00041.bin", + "model.layers.7.self_attn.o_proj.weight": "pytorch_model-00011-of-00041.bin", + "model.layers.7.self_attn.q_proj.weight": "pytorch_model-00011-of-00041.bin", + "model.layers.7.self_attn.rotary_emb.inv_freq": "pytorch_model-00011-of-00041.bin", + "model.layers.7.self_attn.v_proj.weight": "pytorch_model-00011-of-00041.bin", + "model.layers.8.input_layernorm.weight": "pytorch_model-00013-of-00041.bin", + "model.layers.8.mlp.down_proj.weight": "pytorch_model-00013-of-00041.bin", + "model.layers.8.mlp.gate_proj.weight": "pytorch_model-00013-of-00041.bin", + "model.layers.8.mlp.up_proj.weight": "pytorch_model-00013-of-00041.bin", + "model.layers.8.post_attention_layernorm.weight": "pytorch_model-00013-of-00041.bin", + "model.layers.8.self_attn.k_proj.weight": "pytorch_model-00012-of-00041.bin", + "model.layers.8.self_attn.o_proj.weight": "pytorch_model-00012-of-00041.bin", + "model.layers.8.self_attn.q_proj.weight": "pytorch_model-00012-of-00041.bin", + "model.layers.8.self_attn.rotary_emb.inv_freq": "pytorch_model-00012-of-00041.bin", + "model.layers.8.self_attn.v_proj.weight": "pytorch_model-00012-of-00041.bin", + "model.layers.9.input_layernorm.weight": "pytorch_model-00014-of-00041.bin", + "model.layers.9.mlp.down_proj.weight": "pytorch_model-00014-of-00041.bin", + "model.layers.9.mlp.gate_proj.weight": "pytorch_model-00014-of-00041.bin", + "model.layers.9.mlp.up_proj.weight": "pytorch_model-00014-of-00041.bin", + "model.layers.9.post_attention_layernorm.weight": "pytorch_model-00014-of-00041.bin", + "model.layers.9.self_attn.k_proj.weight": "pytorch_model-00013-of-00041.bin", + "model.layers.9.self_attn.o_proj.weight": "pytorch_model-00014-of-00041.bin", + "model.layers.9.self_attn.q_proj.weight": "pytorch_model-00013-of-00041.bin", + "model.layers.9.self_attn.rotary_emb.inv_freq": "pytorch_model-00014-of-00041.bin", + "model.layers.9.self_attn.v_proj.weight": "pytorch_model-00013-of-00041.bin", + "model.norm.weight": "pytorch_model-00040-of-00041.bin" } }