diff --git a/3b_layer0_mlp.down_proj_output_pretrain_representation.pth b/3b_layer0_mlp.down_proj_output_pretrain_representation.pth index ed784b25751a7947993fdc5161c6dcc4cd02fb32..8c80ba70ca006128222da2772b669c5ee6fba5f0 100644 --- a/3b_layer0_mlp.down_proj_output_pretrain_representation.pth +++ b/3b_layer0_mlp.down_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:f4803298646634c1a913da27c7520803ec1c0f9c61451ee7e7c0f33eadefc81c -size 123344455 +oid sha256:bd43192778e0dfa5cbe35cbd2968ee4b0305e86c8c99da5e7c3038f949ff079d +size 92585543 diff --git a/3b_layer0_mlp.up_proj_output_pretrain_representation.pth b/3b_layer0_mlp.up_proj_output_pretrain_representation.pth index 661820cfcc5b5c6bbcba8ee027dab8f5578c5570..80a8d44e76ebdcfdf2bbd52c7ed7fada20f17f9f 100644 --- a/3b_layer0_mlp.up_proj_output_pretrain_representation.pth +++ b/3b_layer0_mlp.up_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:03a3e67ca1fa465ae166a65074de8929d69a6f523860d5288a6e4a2af2987346 -size 493372925 +oid sha256:8254c2b117ba886ae1c4c307ce91961e9f0f54e001df982b97c58394251802c7 +size 246892029 diff --git a/3b_layer0_self_attn.v_proj_output_pretrain_representation.pth b/3b_layer0_self_attn.v_proj_output_pretrain_representation.pth index 3d17799b47675bbc4b748512db85f511b6a36886..11ff71673f86ac114de1f2b4633dcac4d44b385d 100644 --- a/3b_layer0_self_attn.v_proj_output_pretrain_representation.pth +++ b/3b_layer0_self_attn.v_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b5b5197997eac5c5ce7d9ae508888d88260ba4dfefc63e31d32800fe7bd947a4 -size 30837334 +oid sha256:bd1a0451084b985ca7677d0d4fdb1772b0b12f3cfb9585e5dcafd56302724d79 +size 30862934 diff --git a/3b_layer10_mlp.up_proj_output_pretrain_representation.pth b/3b_layer10_mlp.up_proj_output_pretrain_representation.pth index b36af22c525a199de99b85f2b9f61372032cd30e..dbf429ce26f45a1831c18ab257b6fbf1941be154 100644 --- a/3b_layer10_mlp.up_proj_output_pretrain_representation.pth +++ b/3b_layer10_mlp.up_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:99f8491a6dd2d275f0a5daa38134b1dbdb08b0784d68a79c919e0e0241f7e099 -size 493372930 +oid sha256:aa549f99d33fd039a769465d12fa68b33d38ca98cd24e55ba4d3130ef16cb0a2 +size 246892034 diff --git a/3b_layer10_self_attn.o_proj_input_pretrain_representation.pth b/3b_layer10_self_attn.o_proj_input_pretrain_representation.pth index c2ef16c7fbce4f77660b5dfa9a5c54560b1f0228..a2e4f9bec6bdbb3b3bcd77822ea478bbc430432b 100644 --- a/3b_layer10_self_attn.o_proj_input_pretrain_representation.pth +++ b/3b_layer10_self_attn.o_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:eef129f6b916a149bbf63901a4e52fbd1044ee72b356973ae124d1e9aa1b8dad -size 123344470 +oid sha256:68d73e2df08ce1d2b87a0f6f93978ebf2d401b64cf3b1c6e3816a2df0a6cb67f +size 92585558 diff --git a/3b_layer11_mlp.down_proj_input_pretrain_representation.pth b/3b_layer11_mlp.down_proj_input_pretrain_representation.pth index b2d1a737fc692ea7a235ba7b82cc17891c4db5c1..e9871fa6e598a3256a8ad44728797a8a27ffa257 100644 --- a/3b_layer11_mlp.down_proj_input_pretrain_representation.pth +++ b/3b_layer11_mlp.down_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:1c081a9f9ee94e22e3dc781ded8e3a2867e2da464084a91a2555b39ff6b1bba6 -size 493372999 +oid sha256:8d3d57ab4a4355ef355551ccd561a1670f6da79b87126bd017db0e0252bd5776 +size 246892103 diff --git a/3b_layer11_self_attn.k_proj_output_pretrain_representation.pth b/3b_layer11_self_attn.k_proj_output_pretrain_representation.pth index c60a08c5642e7321a801436aa17818274683276e..cc8597afe9a0414a3e4b6bd5bdea75a1970f2778 100644 --- a/3b_layer11_self_attn.k_proj_output_pretrain_representation.pth +++ b/3b_layer11_self_attn.k_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:38fcd9aeae7e4aee0edebed67e622befeb8fc42a1489beee2174986fea2d3aaa -size 30837339 +oid sha256:5cc3b75e28cc9727d7106e0750d43422e11c01ea32956fd1d900af61641962df +size 30862939 diff --git a/3b_layer12_mlp.down_proj_output_pretrain_representation.pth b/3b_layer12_mlp.down_proj_output_pretrain_representation.pth index 480855a86c8d8722310c06e0cd6b24a2f789fe96..491a2e5a963537e202223030eda932c0e54f77c5 100644 --- a/3b_layer12_mlp.down_proj_output_pretrain_representation.pth +++ b/3b_layer12_mlp.down_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:a50a9b633f02b643895ead5bc152f10ce0471cd7bc1708f417d752d73bcb38a4 -size 123344460 +oid sha256:32a590efe46958dd39fefd74db326fce9a854db0d82896c9e2fab9fb9a1e0749 +size 92585548 diff --git a/3b_layer12_self_attn.o_proj_input_pretrain_representation.pth b/3b_layer12_self_attn.o_proj_input_pretrain_representation.pth index bc587cc7d99ce586e575e517e9e2aaa205f35e43..815f46cd74617c27f1739afd8c5ef3b8c3b64326 100644 --- a/3b_layer12_self_attn.o_proj_input_pretrain_representation.pth +++ b/3b_layer12_self_attn.o_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:1c0c7d946bd7e89fb44dc284abb8c81d4b28c1554d7be4da978c3dd0eda17125 -size 123344470 +oid sha256:8fdfe3d9cdf8533b32a318ee20296e4a24b435e91fa2aed4bee8537ebf0438c5 +size 92585558 diff --git a/3b_layer13_mlp.gate_proj_input_pretrain_representation.pth b/3b_layer13_mlp.gate_proj_input_pretrain_representation.pth index c03205391c7fb0574991bbd5d38fff16fda65e9b..1519ba38e0ab419a43f628519b9845518b55da46 100644 --- a/3b_layer13_mlp.gate_proj_input_pretrain_representation.pth +++ b/3b_layer13_mlp.gate_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:cc783f030eae14b85d49dd41ec0a5e26dfe880bf342b42947f13b3efc4be137a -size 123344455 +oid sha256:5221f754fe05f8828d45c97b7d5ac4a443bba7ce77a14c18faa008639ef08f0b +size 92585543 diff --git a/3b_layer13_self_attn.k_proj_input_pretrain_representation.pth b/3b_layer13_self_attn.k_proj_input_pretrain_representation.pth index 3722f268899403034308d8a0bff1da7766f7e009..6819cde8877ece449d06ae3421637b4d3825a37a 100644 --- a/3b_layer13_self_attn.k_proj_input_pretrain_representation.pth +++ b/3b_layer13_self_attn.k_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:423f8a859ae5a086652c7ba32672859f98995f1f93e0df399955bee184496b09 -size 123344470 +oid sha256:f7fb65f3c2382eeb4a20caef1c024573b451c9594fddf598406b18eada0d9b08 +size 92585558 diff --git a/3b_layer13_self_attn.k_proj_output_pretrain_representation.pth b/3b_layer13_self_attn.k_proj_output_pretrain_representation.pth index 92d46f9a375ecae0dacb47ca969d55e8752dd274..20044b0904334fc50051d1fa987174a474f1d70f 100644 --- a/3b_layer13_self_attn.k_proj_output_pretrain_representation.pth +++ b/3b_layer13_self_attn.k_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:04f4bc23b7425e93a97f0ef93eb668120143d7468787f96e9e5ccc22ce20b452 -size 30837339 +oid sha256:e06909ecad4f7d6c6a3de4c1f2d8670e395e87f97380ee186124efceb7cfc7a2 +size 30862939 diff --git a/3b_layer13_self_attn.o_proj_output_pretrain_representation.pth b/3b_layer13_self_attn.o_proj_output_pretrain_representation.pth index 8238955d1b6de390688360fa65c49dea8e389ebe..e60d70871602f67baa336e50ec2383ef0b56cf3d 100644 --- a/3b_layer13_self_attn.o_proj_output_pretrain_representation.pth +++ b/3b_layer13_self_attn.o_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e802cece00fe2a8b6c48d2e99441e517cf5ec4815ce2c996f185e2883025bb6d -size 123344475 +oid sha256:39ed07b036e092a8f13ac0edd27f08097558dcd7756bf63237d6b26e869aa8c7 +size 92585563 diff --git a/3b_layer13_self_attn.v_proj_input_pretrain_representation.pth b/3b_layer13_self_attn.v_proj_input_pretrain_representation.pth index 6c0964c0381c0c10ffe18c62f19e7fb7e9ae847b..17d40d5070dc756b78ecf9abc314601ddfad943c 100644 --- a/3b_layer13_self_attn.v_proj_input_pretrain_representation.pth +++ b/3b_layer13_self_attn.v_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e1241ccaed49077569dad53fd40dd6d55f83893527329a8a1fc1e11fd4c2b1df -size 123344470 +oid sha256:a50a3498d82c3dcb1a4a47d056a02f6e63745f38516df614185c5894b9f8ea97 +size 92585558 diff --git a/3b_layer15_mlp.down_proj_input_pretrain_representation.pth b/3b_layer15_mlp.down_proj_input_pretrain_representation.pth index b44f55934c9bade1b865eb7b56677c2b83b659e6..50d4aea5c1e20b208d2cd0ba42f1d83f5446442f 100644 --- a/3b_layer15_mlp.down_proj_input_pretrain_representation.pth +++ b/3b_layer15_mlp.down_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:79b247bc370fa4ddcd4185044fac9ff71dbee040b17573e203339f1da21eab9f -size 493372999 +oid sha256:3d34d2907d4bd26eff34cb3a8abf09fa451acd2d2b591a8f03bddd032725f15a +size 246892103 diff --git a/3b_layer15_self_attn.q_proj_input_pretrain_representation.pth b/3b_layer15_self_attn.q_proj_input_pretrain_representation.pth index d467c0b0839f922cfe426a240d0d9f74d1cd340c..adbaf86ce3ae267c734916f864a00f0b479b8f4d 100644 --- a/3b_layer15_self_attn.q_proj_input_pretrain_representation.pth +++ b/3b_layer15_self_attn.q_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:9c00ee36249e6292dd5423674a322ef1c27b2f0ab3f38b0bfd9605e0a4780e61 -size 123344470 +oid sha256:11b29f185747a9791bad16af467557de1990b6d58618aefa972927ad94e082ea +size 92585558 diff --git a/3b_layer15_self_attn.v_proj_output_pretrain_representation.pth b/3b_layer15_self_attn.v_proj_output_pretrain_representation.pth index de76b15cc5f5cd905e313cc854085463b2ac0e37..419644786ed6f143f6054a69f880b887138341a4 100644 --- a/3b_layer15_self_attn.v_proj_output_pretrain_representation.pth +++ b/3b_layer15_self_attn.v_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:c8e9c5a8abd8b7fec7bbca230aef2c49a33dff33beea297dd10f2786114be2f6 -size 30837339 +oid sha256:d7973aaad3daaeb695a944223ce8654a13a3924a0bbae59f9b7cb3acc4e18cc8 +size 30862939 diff --git a/3b_layer16_self_attn.o_proj_output_pretrain_representation.pth b/3b_layer16_self_attn.o_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..9a9f8ac7da3ccb60cc809c3b728cc1464355a5e0 --- /dev/null +++ b/3b_layer16_self_attn.o_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8b702c51ed70de2e58c5ba3503fbd1fb8d12b9915ce9a186aced994327f4666c +size 92585563 diff --git a/3b_layer16_self_attn.q_proj_input_pretrain_representation.pth b/3b_layer16_self_attn.q_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..70eb6dba7b102a2fdef7da7a1253079d67fb5491 --- /dev/null +++ b/3b_layer16_self_attn.q_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7e65555b1ccd20afb3a13a8b958600dad6b0595653980e83dc80bd033cc858a +size 92585558 diff --git a/3b_layer16_self_attn.q_proj_output_pretrain_representation.pth b/3b_layer16_self_attn.q_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..fc760a7216b971f1f9127f58412f35ccb14f2c9d --- /dev/null +++ b/3b_layer16_self_attn.q_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:11d37618233e3b1f6c870374fcb61f4bb821c8c2a79a235e64a92bb084d44a89 +size 92585563 diff --git a/3b_layer17_mlp.down_proj_input_pretrain_representation.pth b/3b_layer17_mlp.down_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..f78ce3fc9c2cb57f7e86773d6243c7e906c9c006 --- /dev/null +++ b/3b_layer17_mlp.down_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:424bc4812c3a426d6a755040ae1aeb44615b23dcbaa5e1715e54e20de3069e14 +size 246892103 diff --git a/3b_layer17_mlp.gate_proj_output_pretrain_representation.pth b/3b_layer17_mlp.gate_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..803b8926af5836a7026e6bc3ec92b88d77fdfe63 --- /dev/null +++ b/3b_layer17_mlp.gate_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:028aa49d4a0d7be7bdc1137791afe81ed461385ba6fe06dad278bb2138b51da9 +size 246892108 diff --git a/3b_layer17_mlp.up_proj_output_pretrain_representation.pth b/3b_layer17_mlp.up_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..4f456b910f0d35f0e67a691c77f6d9786711080d --- /dev/null +++ b/3b_layer17_mlp.up_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70a93308cfcd528f14f65cc529cbb44edea0e6526cce566249d8e9f818ed92d1 +size 246892034 diff --git a/3b_layer18_mlp.up_proj_input_pretrain_representation.pth b/3b_layer18_mlp.up_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..790f21de79b448884a1130fb028eed4b3f73dc0b --- /dev/null +++ b/3b_layer18_mlp.up_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:62eb41531f6db36b463c992813b431ae748b8c52499b1d6d134b1c7989563417 +size 92585469 diff --git a/3b_layer18_self_attn.q_proj_output_pretrain_representation.pth b/3b_layer18_self_attn.q_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..503ed138eff76a3c9bc9a8887d86b123dd241303 --- /dev/null +++ b/3b_layer18_self_attn.q_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25acfd751295a3921f86353c8a88c1486f5eab799eb8e84b87b4da272f94c7b0 +size 92585563 diff --git a/3b_layer19_mlp.down_proj_input_pretrain_representation.pth b/3b_layer19_mlp.down_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..5f117e103d0dd2d0e22a215d3456125f8b0e1b97 --- /dev/null +++ b/3b_layer19_mlp.down_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:54ea5b0a4e0dffab4b0a432ad2629c3fa0e0bd40105a3acc732d9e443a68f19a +size 246892103 diff --git a/3b_layer19_mlp.gate_proj_output_pretrain_representation.pth b/3b_layer19_mlp.gate_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..90af9bdbfd49d6959e60fc54d4298f08d4077c0b --- /dev/null +++ b/3b_layer19_mlp.gate_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:145d7ade85ffee41fce499025f41a2c6fe95ba7226f99798f62b4229b863927f +size 246892108 diff --git a/3b_layer19_self_attn.v_proj_output_pretrain_representation.pth b/3b_layer19_self_attn.v_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..02183978a110f86918fbd328a11e5dcee1785b5f --- /dev/null +++ b/3b_layer19_self_attn.v_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0156c6aa3209e4193c2703cae1284c10d173118268e38e6a70738a29c59b571 +size 30862939 diff --git a/3b_layer1_mlp.down_proj_output_pretrain_representation.pth b/3b_layer1_mlp.down_proj_output_pretrain_representation.pth index acc6a659b55ccc16957b14e0abe54e2333bb14d8..37951d72d8a1dc747edbe8072512f6a75b0e571c 100644 --- a/3b_layer1_mlp.down_proj_output_pretrain_representation.pth +++ b/3b_layer1_mlp.down_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:55eef398fb372442a79c5e99ce8db1a21b83d04242e4657e83289c85f72e9bf7 -size 123344455 +oid sha256:548d0c08d4723cd7367f1450d4004facb57ba0009af8f78a7fd5931a4fc83e2e +size 92585543 diff --git a/3b_layer1_mlp.gate_proj_input_pretrain_representation.pth b/3b_layer1_mlp.gate_proj_input_pretrain_representation.pth index 8177d2aeb11ba0ff3ab5c49f5b1cff4f40251d9f..7c93b5551b39983682e6eea25032d3e0701c6eb6 100644 --- a/3b_layer1_mlp.gate_proj_input_pretrain_representation.pth +++ b/3b_layer1_mlp.gate_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:711c63e0bf74e71b8676d749c98bc4513fb04ae72e7e677cbd6a85ed2dbae0ff -size 123344386 +oid sha256:e28e94ac38bd17be0cee9014912416cedb7231a7e1466893a08927bf3b58688c +size 92585474 diff --git a/3b_layer20_mlp.down_proj_output_pretrain_representation.pth b/3b_layer20_mlp.down_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..7cde46e5635255adc02f43a520ca8ff6a880c707 --- /dev/null +++ b/3b_layer20_mlp.down_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:156c6b60239353846443ab43a8c98b742858b01b2b8f3bca560604c036d2cd2d +size 92585548 diff --git a/3b_layer20_mlp.gate_proj_output_pretrain_representation.pth b/3b_layer20_mlp.gate_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..855654fd7f64eac9039018cb3fca6bbf5baa4ede --- /dev/null +++ b/3b_layer20_mlp.gate_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5eb2a48f59114f1cbe669ff4b3b2ba5e1b2c2b573f8c64085733406694a93068 +size 246892108 diff --git a/3b_layer20_self_attn.q_proj_output_pretrain_representation.pth b/3b_layer20_self_attn.q_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..69584985d0d43e43714fa7eae9f90d63532ed320 --- /dev/null +++ b/3b_layer20_self_attn.q_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9c21e8c6ba35fb638d38bade538bb62c6a7f8e900254382b82ca993661c046e0 +size 92585563 diff --git a/3b_layer20_self_attn.v_proj_output_pretrain_representation.pth b/3b_layer20_self_attn.v_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..ad0c345aa1ea37b5f9288c08ba340260b4c85d31 --- /dev/null +++ b/3b_layer20_self_attn.v_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e6d4e58b9ebb93845edaeeed81e8b615fa56020902fc598b1a3d47cd0ea9b98 +size 30862939 diff --git a/3b_layer21_mlp.gate_proj_output_pretrain_representation.pth b/3b_layer21_mlp.gate_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..90d2d47531269cacbebcc293347563a065df469b --- /dev/null +++ b/3b_layer21_mlp.gate_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ee89434fe3f55447bfeb805027043ff2956e0b0d05a2b2a5f0c1d195408e2b9 +size 246892108 diff --git a/3b_layer21_mlp.up_proj_output_pretrain_representation.pth b/3b_layer21_mlp.up_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..bab8a1787a391d6dcfd7004b3c892019035b4831 --- /dev/null +++ b/3b_layer21_mlp.up_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31ddaff8610183dfe33189f6a6ec95c5c724a5238a03ac8c186ea4eb0a5fe7f5 +size 246892034 diff --git a/3b_layer21_self_attn.k_proj_input_pretrain_representation.pth b/3b_layer21_self_attn.k_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..f354f30808d13c860f22da330aa647aed6357f69 --- /dev/null +++ b/3b_layer21_self_attn.k_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c9ce3d4237ebccfeb5e5191de75c22242be73f47e16f720decbe350ac1f74ca +size 92585558 diff --git a/3b_layer22_self_attn.k_proj_output_pretrain_representation.pth b/3b_layer22_self_attn.k_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..81157a9f2b114ee68ce419788ff3d4abf6347cc9 --- /dev/null +++ b/3b_layer22_self_attn.k_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e3a67fc08b7baf0f691195ee5e390e567ac7c68079b3a3c81b0bf5e5f490b66c +size 30862939 diff --git a/3b_layer22_self_attn.q_proj_input_pretrain_representation.pth b/3b_layer22_self_attn.q_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..0f26a33d9ff22d0b810780d92eff5bd22c8d2810 --- /dev/null +++ b/3b_layer22_self_attn.q_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:036fad3bfb470d9d4656c28e0d6a8d6602eccfa95fc42f142b652f71bf160e6c +size 92585558 diff --git a/3b_layer24_self_attn.k_proj_input_pretrain_representation.pth b/3b_layer24_self_attn.k_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..ccae2aba9253769e7121ae1394de598662af1ecb --- /dev/null +++ b/3b_layer24_self_attn.k_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f78b40fe94e6e53f527505f5a8bbf5e00fb936946a944bdae473ca7294f5b107 +size 92585558 diff --git a/3b_layer25_mlp.down_proj_input_pretrain_representation.pth b/3b_layer25_mlp.down_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..bd8bd21bff71eeb630a5e96dd828e2e0ef28f2d0 --- /dev/null +++ b/3b_layer25_mlp.down_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0deaf329c9cd84015baaa00ff9478ccf05e82c123536677a2ef8be1e318aec52 +size 246892103 diff --git a/3b_layer25_self_attn.v_proj_input_pretrain_representation.pth b/3b_layer25_self_attn.v_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..141f7fc78e43bdc442dc6e4a224edd09ddbb0efc --- /dev/null +++ b/3b_layer25_self_attn.v_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45f697f3c8f33da0acbb1bafddbd9ead53279669fa81db256c853a26014b9565 +size 92585558 diff --git a/3b_layer26_mlp.down_proj_input_pretrain_representation.pth b/3b_layer26_mlp.down_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..04efa25ba60c54458ddb42c4c28e09b2b2e74ce5 --- /dev/null +++ b/3b_layer26_mlp.down_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:806fc534582e42335394bbc3412be7766c8bb64b7a77f1895b58c6fee77a520a +size 246892103 diff --git a/3b_layer26_mlp.gate_proj_input_pretrain_representation.pth b/3b_layer26_mlp.gate_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..2efa71a99b4c4f82ed25e0c1ac6c942d105e0b43 --- /dev/null +++ b/3b_layer26_mlp.gate_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4c33cd3b77b784f6d35efc3a429494a71d423f739452d61476eff74b7d229fb9 +size 92585543 diff --git a/3b_layer26_mlp.up_proj_input_pretrain_representation.pth b/3b_layer26_mlp.up_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..b1923c1e10c4462420ca5e69dbdee7c24f4bfc32 --- /dev/null +++ b/3b_layer26_mlp.up_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e8130db7c7ed7fb89e3d98a57574d176c305555fb67e40a48bb9688797d47b4 +size 92585469 diff --git a/3b_layer27_mlp.down_proj_input_pretrain_representation.pth b/3b_layer27_mlp.down_proj_input_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..62f0c0346f70720c82af7debd52011f7a71800dd --- /dev/null +++ b/3b_layer27_mlp.down_proj_input_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:10014ff9455695e28ac31ee7330ac9235401bb0b98e01823810c6411114b0198 +size 246892103 diff --git a/3b_layer27_mlp.gate_proj_output_pretrain_representation.pth b/3b_layer27_mlp.gate_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..b180b0d768f11a3e9f1ac5db8f85305ba691aea0 --- /dev/null +++ b/3b_layer27_mlp.gate_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:52e4ffded83aaa8d4d89c71c2c497fc57499578526317b9ed9bc858cd8e3793a +size 246892108 diff --git a/3b_layer27_self_attn.v_proj_output_pretrain_representation.pth b/3b_layer27_self_attn.v_proj_output_pretrain_representation.pth new file mode 100644 index 0000000000000000000000000000000000000000..d838f66397ce4974bf410e89f1a4d598ab8b3b61 --- /dev/null +++ b/3b_layer27_self_attn.v_proj_output_pretrain_representation.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb77e2dedcfd65b8a3340a41d52c8f11e0b794865b5973028914cb03af6a6de0 +size 30862939 diff --git a/3b_layer2_self_attn.k_proj_input_pretrain_representation.pth b/3b_layer2_self_attn.k_proj_input_pretrain_representation.pth index 1a09f60f20c8f712c2f13ba81276bff4753d25a2..aafbd031ade6bff97f5b3ce65817873e0c1d8b71 100644 --- a/3b_layer2_self_attn.k_proj_input_pretrain_representation.pth +++ b/3b_layer2_self_attn.k_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:35c1af2647052c75df469e5af7085b84319258fe260d13de6a2e85c962158485 -size 123344465 +oid sha256:1f0573857e3fbf87a4138ca0927cc9d18133d1b934048bd6b7c58fce73ba49d7 +size 92585553 diff --git a/3b_layer3_mlp.down_proj_output_pretrain_representation.pth b/3b_layer3_mlp.down_proj_output_pretrain_representation.pth index b97ac28f1770f8c26c49e00a80d738b4a9f9e3d1..d766f31bdb826c64f0bd7eaf6f73ae0ba6cb9d0b 100644 --- a/3b_layer3_mlp.down_proj_output_pretrain_representation.pth +++ b/3b_layer3_mlp.down_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e4b029b2d4d9674e8b56cd0520234ebd4799a2dd6906d308c7d7e05de8692131 -size 123344455 +oid sha256:0fa07658ee4e3bb721f472b6faa882ae68c8ed8cd2e2ee2eaff4273926991c0d +size 92585543 diff --git a/3b_layer3_mlp.gate_proj_input_pretrain_representation.pth b/3b_layer3_mlp.gate_proj_input_pretrain_representation.pth index 6705c855575b18d0480ea1872734cb72acd9829d..183786f5202a842b339d60bdf4df561d3f879421 100644 --- a/3b_layer3_mlp.gate_proj_input_pretrain_representation.pth +++ b/3b_layer3_mlp.gate_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:c1ff9e76551e83543481d6eb76189391a8b8b968803815b326f98a84bda6f4d3 -size 123344386 +oid sha256:4ddd74e0567c5197c10dfd5084cd39619bc73a9714fc8996f2806b377470e59a +size 92585474 diff --git a/3b_layer3_mlp.up_proj_output_pretrain_representation.pth b/3b_layer3_mlp.up_proj_output_pretrain_representation.pth index f23f0a29e7da1bf17c0968e843ec304e297d34fc..2b8840e13834fcc0950bcf059933a43e2b3394ee 100644 --- a/3b_layer3_mlp.up_proj_output_pretrain_representation.pth +++ b/3b_layer3_mlp.up_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:7ea5ccfb494f5aa02555a2d916aa013a6711c1f75a8b6d90a70d121e9d33d338 -size 493372925 +oid sha256:c833eca980f6b0fc02c3fbdc0433623ecf7af4a55430f922bc96cf9e23c323af +size 246892029 diff --git a/3b_layer3_self_attn.q_proj_input_pretrain_representation.pth b/3b_layer3_self_attn.q_proj_input_pretrain_representation.pth index a48c3a12aad8ff02917cf3ec8ee963fb392f8fb1..cb9c7add1a3e4f58438f044971e8de2062282d52 100644 --- a/3b_layer3_self_attn.q_proj_input_pretrain_representation.pth +++ b/3b_layer3_self_attn.q_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:ea32928bec7fbf996b169925ab5de8a47c872daebc3cb0acd5e55e1a0c741aaa -size 123344465 +oid sha256:af2b3530eb97eaf521e27d628c40976705dce26ae101bf7119b5acb0164480a8 +size 92585553 diff --git a/3b_layer3_self_attn.q_proj_output_pretrain_representation.pth b/3b_layer3_self_attn.q_proj_output_pretrain_representation.pth index 8f111ec147c3a4dbbf3c1b6c665de7ff608cc17d..200e757b7034f76610a2144f3b69367224aec2f5 100644 --- a/3b_layer3_self_attn.q_proj_output_pretrain_representation.pth +++ b/3b_layer3_self_attn.q_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:23cffd79e010c3bc9f52f7207a39af7cdf9482f95ae19124bea88ecbc0027a4d -size 123344470 +oid sha256:fc23e6bbb8c95af4dd1a0d02bdd8edb2a0d26a0152bce180a099082a4655347e +size 92585558 diff --git a/3b_layer4_mlp.gate_proj_output_pretrain_representation.pth b/3b_layer4_mlp.gate_proj_output_pretrain_representation.pth index 399a35faedbe7c83adac692eae63504d626c18b1..a1f8f22a1b19aadcfab53533788de784139e1b7f 100644 --- a/3b_layer4_mlp.gate_proj_output_pretrain_representation.pth +++ b/3b_layer4_mlp.gate_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:043aedf293c2cf152b541efaf3291c687c108e07e0f52627273e146107c94662 -size 493372999 +oid sha256:33b56345c019fbc96a5cf9728b2f7611e3d376465b70f3e797ec782889979cab +size 246892103 diff --git a/3b_layer4_self_attn.q_proj_input_pretrain_representation.pth b/3b_layer4_self_attn.q_proj_input_pretrain_representation.pth index cfd39cdeffe32d542e4bdd95d273162ff2f0c978..0089d52c2b44d9fddd1eb8a4789df02fa73e7775 100644 --- a/3b_layer4_self_attn.q_proj_input_pretrain_representation.pth +++ b/3b_layer4_self_attn.q_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:aedbee6e28dd3434abec165629805b18d64524fa62d33a1a7fbe70e1cc956ffa -size 123344465 +oid sha256:d05972f2e6f03b70849af7d07f91f5446ff42eab94199d1b19e4629e471e9fd7 +size 92585553 diff --git a/3b_layer6_self_attn.q_proj_input_pretrain_representation.pth b/3b_layer6_self_attn.q_proj_input_pretrain_representation.pth index bff376db00a7fda404f6b3bdb3457251773b22f3..068dfe2de4d6f623af4a5bcce439d2540c13b3de 100644 --- a/3b_layer6_self_attn.q_proj_input_pretrain_representation.pth +++ b/3b_layer6_self_attn.q_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:081bbd5c8de087a039d24b3c33f085ad64f0986aa536e00f8d0440b5d120eddc -size 123344465 +oid sha256:416ecb2a14fddf812112f2ef2e71229ef533647f4f32581657a3bd9513340857 +size 92585553 diff --git a/3b_layer7_mlp.down_proj_output_pretrain_representation.pth b/3b_layer7_mlp.down_proj_output_pretrain_representation.pth index b7734db3847adbcb027aee2fe86d34a995adccb9..0ec41bd3352bfaee0fc51deacea3ecd957242d66 100644 --- a/3b_layer7_mlp.down_proj_output_pretrain_representation.pth +++ b/3b_layer7_mlp.down_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:4472707c2dbccbe5d279344cf4c64e4ecffa7db89c480c31dfe41b2d19d6837b -size 123344455 +oid sha256:5ddcc9dba3e5f603768c7e37abe50cf7dfc4ab067335d6f6154a3eb04c8151d4 +size 92585543 diff --git a/3b_layer7_mlp.up_proj_output_pretrain_representation.pth b/3b_layer7_mlp.up_proj_output_pretrain_representation.pth index 361c8f566f92cd3b24a640ee68c4f92fd4084032..6f098e6ec6cfeec40dc9f6f88bf65413a22491e4 100644 --- a/3b_layer7_mlp.up_proj_output_pretrain_representation.pth +++ b/3b_layer7_mlp.up_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:91fa0195b5a6eaa145508b8bf0e2be5a54aa2c110c7aeb05cbe36e0d3e0d8c62 -size 493372925 +oid sha256:606922a2d86d108b1373f3d9d8468c726063bf8459cd2c737a888332259ba74d +size 246892029 diff --git a/3b_layer7_self_attn.q_proj_input_pretrain_representation.pth b/3b_layer7_self_attn.q_proj_input_pretrain_representation.pth index 1543bf70169f5fd5a42515ad7efcef61713de872..ff9a2f4f951f321af7125eb1edfac5c9da81c99e 100644 --- a/3b_layer7_self_attn.q_proj_input_pretrain_representation.pth +++ b/3b_layer7_self_attn.q_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:df76ec36f394f067e0d9a52a250872941176debf0f18d2dadb764335fcd19663 -size 123344465 +oid sha256:af0daf29b036608d273d96b801b66883f264364ee36dc7544a340609526660e3 +size 92585553 diff --git a/3b_layer7_self_attn.q_proj_output_pretrain_representation.pth b/3b_layer7_self_attn.q_proj_output_pretrain_representation.pth index ce5402a60f73185b6a6c564979216cf64b6c31c2..69a102d7265c166482427e0ac33d9365708c4b45 100644 --- a/3b_layer7_self_attn.q_proj_output_pretrain_representation.pth +++ b/3b_layer7_self_attn.q_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:468031684ab077995cc69f7bd3d0d2d52a1e6df4be0873455d2ad23c818388e1 -size 123344470 +oid sha256:f8a44068b294e5d6b4f06c64d132b2d940f984e67c9bf00952ba79e3f76103c7 +size 92585558 diff --git a/3b_layer8_mlp.gate_proj_input_pretrain_representation.pth b/3b_layer8_mlp.gate_proj_input_pretrain_representation.pth index d4af8ae01eded69dc11cbe93b110cefe1f4637a8..f19c4929539fff1e89d96d25358dd41bb0a5b803 100644 --- a/3b_layer8_mlp.gate_proj_input_pretrain_representation.pth +++ b/3b_layer8_mlp.gate_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:472d56e725160b0ff46915a23ceac8a7e6ea804285280f769158ccb46a0705b4 -size 123344386 +oid sha256:7e44554eae4ffbcd94cb08c4ec343ef2f39f1164ded023fc88e7282846ed7fcb +size 92585474 diff --git a/3b_layer8_self_attn.q_proj_input_pretrain_representation.pth b/3b_layer8_self_attn.q_proj_input_pretrain_representation.pth index bea8f4447cb30df2e02f2dfa1eac6c977bed97d5..bddddf33022ca4602596b21706a35d402b0339fe 100644 --- a/3b_layer8_self_attn.q_proj_input_pretrain_representation.pth +++ b/3b_layer8_self_attn.q_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e49c291e86e729c5e4360d0407132d13b51356f2b013b0b067f76bfe80abcaba -size 123344465 +oid sha256:4b17e9346ee09d4123e49b65062d164c62e47137258ee4a5291713977a7f43ee +size 92585553 diff --git a/3b_layer8_self_attn.q_proj_output_pretrain_representation.pth b/3b_layer8_self_attn.q_proj_output_pretrain_representation.pth index c8acd99e7e9da3170f188ed94b539b5dd6df0f45..0f66fc3399109f8b105263a5bf505d4cb54924dc 100644 --- a/3b_layer8_self_attn.q_proj_output_pretrain_representation.pth +++ b/3b_layer8_self_attn.q_proj_output_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:efccc204f73476c31955f6ced942784955b35d3a3a1a655fedd02faa6ee269ea -size 123344470 +oid sha256:82aff2ec6035e9307df9111efb23c83b31b5be0ec808f65403ac2644f8ac00dd +size 92585558 diff --git a/3b_layer8_self_attn.v_proj_input_pretrain_representation.pth b/3b_layer8_self_attn.v_proj_input_pretrain_representation.pth index c5a8bbc6d75b3f37f2848b2b7ffce9230931f776..6068af73295924710f0d5b7bc769326251c66afa 100644 --- a/3b_layer8_self_attn.v_proj_input_pretrain_representation.pth +++ b/3b_layer8_self_attn.v_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e0c1c1ffd995a3a79f111e996b4c19669bc0508e052f9f8e82e5ab60da0125f2 -size 123344465 +oid sha256:245c2a695ed7481a61bbf715212dc22aeb37f646af9964c40cc4665900c43968 +size 92585553 diff --git a/3b_layer9_mlp.up_proj_input_pretrain_representation.pth b/3b_layer9_mlp.up_proj_input_pretrain_representation.pth index a2a353781086a396339db7a44607e4afdf3e32ed..a368ca310ff6ba26c655cd37d2a27720c8ef5514 100644 --- a/3b_layer9_mlp.up_proj_input_pretrain_representation.pth +++ b/3b_layer9_mlp.up_proj_input_pretrain_representation.pth @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:29c4cf5bcaa72a773b8e98d676858d938c608808ab35499458355e148437cc46 -size 123344376 +oid sha256:351a8375c95cd08333a090b4aba15ca7a6f451471f27e00882535993a70f170e +size 92585464