diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/lm_head/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/lm_head/_s.pt
index e5f25a1fee5137fb273dc043741436f87c32cf78..e4efe72b2db4074e91b895ca803a525cf7851e97 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/lm_head/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/lm_head/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.fc1/_s.pt
index f0721afd471f5677a086cb231fc4dfc8b040aba8..7baac375e3cf65541ab7bdc475354b90ba8a3f52 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.fc2/_s.pt
index 5bbe2e4e9217fb33315ce054bd7d732a087553cc..8c33971b8a2696aeea910bf511de82919e42031f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.k_proj/_s.pt
index 8db6948e5f846b6b6a98d628ee046626ea2cdea7..a85bd4c746c2d197f76048c927a70070c47195a8 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.out_proj/_s.pt
index 77f10c0fe5e3bbdf207f3b1ade856c897ac042fa..fe9c7363191eacf6ea1286d78f41fc8cabf676b1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.q_proj/_s.pt
index 4f5f50037381e2efb3b7078970e656daf88487d9..dd4e0b831b4496cc4ed13c55fbad21f7204357cd 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.v_proj/_s.pt
index 50dc00cdea257650d5da7a64b342a955276f6f47..7cabd32d50cb1c2df0dc0261d9fec7df7077ad6f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.0.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.fc1/_s.pt
index b0db1381c7e83b2875ff5a255cb2a70baf8344bf..82e663df96cc8a5d567814f30d93abbd77989b94 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.fc2/_s.pt
index e5512dabbdfad06448a71543bcedb3198a7ede9a..5b3e9bd67ef071bcaf5cdd27d102822d22d40395 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.k_proj/_s.pt
index 52b2c1fdaa7a6dac04cd52cd272bef99e1aeaccb..ec27feeb2e53cd88e0b8680aa42e87ade95d894d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.out_proj/_s.pt
index abe0ffffc262442556295ce4dc81413c56112af6..20a00d9c867f4dde78b43ce40d68822a19b17b7d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.q_proj/_s.pt
index c7a26036b668f7a3b400ec37a7cc55292744f2a3..8c954653605f5db1e54f2cd34da7384e3ad85356 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.v_proj/_s.pt
index 0a0cff2d05bb7cafd9de2f0c19565aeac717709f..4f9414ed29b87f0115d0c24ce3c98a113aa9d219 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.1.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.fc1/_s.pt
index 3b855fbb0f315149830301719aff295b5c69d986..4356adcdd37f30c256d391aedfc572b75da2d8e1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.fc2/_s.pt
index e4887fd8177598a4aadd7be9aee1545ebaff50f2..e57b0d00c47c0e1c2e27dc386945411f031d38a4 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.k_proj/_s.pt
index 5c202bac4a43e40030d60f035c574846f7f1ca77..8b41e335d3b5928a5d0cfd4b205b3d0bc6978b08 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.out_proj/_s.pt
index 3195c8998d1e1b0774abb27465878fee5a58e274..4678c743d85c39157a1c6fe5336c180ecc59b259 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.q_proj/_s.pt
index 1af61f78d6f4624282b493181e5796d338369ea6..a3ea2f0a917bc51783ed952235f83b6dff1fca9b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.v_proj/_s.pt
index 8942da41059534ccf09a693023d16ec51511a73d..9db8a55cc39627e5ba058dc3d6b5f25f64fc144f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.10.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.fc1/_s.pt
index 6c730f8830c9774933c764ba706b98ce117ecc23..b2dbddd0064ab020c07ae32c42c5aab6c4cc06d9 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.fc2/_s.pt
index f4b7f265ac02f20f58cd16918880195b600894aa..e8181bf7a9e68d79adebeb54f4ee14dd1d60b3b3 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.k_proj/_s.pt
index afc5d15a0532fa45c8eacb5eaa1d884e57d4d45a..813ae1466e42bf16a74ad7eaf211551bb601bede 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.out_proj/_s.pt
index 2292fc419b5a097c801b9fce3867ca9d3bcc6a47..96948db0334bd632cd62f74b9c528f28cd896dc7 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.q_proj/_s.pt
index a5abab67f934cfd48d5ec5e0f8b4386f0c54cd4a..d9bbbe2d778788be28008de6d3471263e79c0e51 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.v_proj/_s.pt
index 8ad5f575f28846ae2f3719379d1274e9d94f3a73..a139fe451094030851c9a6349aa9e70873d56d61 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.11.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.fc1/_s.pt
index 1a89c5b755e6667854bbc6e5589a376c24d8fcb0..0d8b845c475f86c2f7b21226c59aa1ad8302f370 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.fc2/_s.pt
index 319f9b6530e388d488ed3892fd39960c3203284e..3ebb078617c64515ecabc9b55e6103ac987e53ed 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.k_proj/_s.pt
index a5974c3a0950cce16e825217757b868710f5e624..118b050d05b3b0cc0601b723e8630ee3ff88c22d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.out_proj/_s.pt
index 999208167acb45d2881d4ae1bb5a570680510033..87c9b95f6a7ffac9355f446556815fd07e1df694 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.q_proj/_s.pt
index 149808e83617d102f18e3e01288fe4b72fd2ef91..6891a989b3b99f9f99d205dc974d238347527852 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.v_proj/_s.pt
index ce205df2fe5e50a55de2b824b36f8d9800edb59e..d09daaf822eb6a6b8b993dd12ffc07def4e3e94d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.12.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.fc1/_s.pt
index c91a20600b48124abb333f77ddfb76334a727daf..4d2187d6162c81f95ffc745699761466cd862163 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.fc2/_s.pt
index bec33cf00da06c7ec3e087543cc247467e2f0dcc..f93466ca539e0412c832105ec398a33f5bec5800 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.k_proj/_s.pt
index d119994e53f8015a30032e8e12c4256ee1c181dc..c506bd3312fc61546501e79a16f507f24f7af86e 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.out_proj/_s.pt
index 98336f63995c62f0f081a024efa649631ddde504..56da6bcfbd73dcd606e3eed3d6b4e2d58ba5edab 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.q_proj/_s.pt
index 0551dba942101c881043141b8ebcba654bd34764..720f682f12563cd270373eb53d1ffa52c980205c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.v_proj/_s.pt
index 4f34fe34e5b8cba0f3ba43dba7673757acbf0dc4..d3f506c7610da49ac6b8b2794f85e1b9c3dfa50c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.13.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.fc1/_s.pt
index d94896bd70d9cec9c8b27ee9a7f3e6af4a9f74c4..cd574dd656bff76f26e103becb19c1a1b2e89678 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.fc2/_s.pt
index 7fab8e7e4daa56d0e46b14fd91aaf77efe278780..b9ff3d33db94c076e71ff27d397909c08cfbbdca 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.k_proj/_s.pt
index 17c530365ba5809b4bf0b562d978384c29fac19d..af07f87166c53a642c9b104725fdecfbac19e3b0 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.out_proj/_s.pt
index 5ea31f50642f40797e8fcf0ea85f9188980973c6..aa494ce80f6b9f67a00488382c745009d114bc2c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.q_proj/_s.pt
index 9918887edb58b661c0f0932c162f69a21a808d58..769a8780d8f14c4956ce63e693d722642c0c2a81 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.v_proj/_s.pt
index 1b6343e61db9a49cdc97689aa3b0eeedb49941f3..eeb81a48f6a8d5ac201fd4a1fdaf9a60c03d3577 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.14.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.fc1/_s.pt
index 64c5d7f705f2fc72812efa76d3db39125c57e49a..267ec8c96609ceb1d205976a24e9d88bfb91b0b0 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.fc2/_s.pt
index 1ac50bbfca7e3e03bec9be0d0c42b89286d6040a..235e28b5aaadf240f2454da5706d5e9f04e71ef9 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.k_proj/_s.pt
index b44636b763769b2b036095315fc848c7ada22cfd..4268b24c1a52f40dcdcd75be586e954a250c0aba 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.out_proj/_s.pt
index d95921e800a1d64aeab9a0136837f333a8ef858f..a72a3543674df717e3167e761051cb385721c8c5 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.q_proj/_s.pt
index a251bab42d8a4c7d6306c527f397c8584f1b1bf7..d42cf65e9590e8541b91a8bf101f3fc78271a3f9 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.v_proj/_s.pt
index 94dfdadcf0e97d1e3858e464b1bfb5aa16bccc0e..44d9f6e15a8f8883bf0464d6c30e282784391b31 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.15.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.fc1/_s.pt
index 3fc854d0d8ebb69e59c76e7c1b5823ce7dd16c7a..ebe7a3e37cb97d6045253e5247745b0a64c33c2a 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.fc2/_s.pt
index fed7ea0842c01ee2e77cd25386db6f21d2330197..1ca9959af20547cc794c8ca5a3b506407a5537d7 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.k_proj/_s.pt
index 72a0d6cdbaec99c64e57b67d1c0978456dede048..1eb8461667eb6d530288eb783701499e1da989ba 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.out_proj/_s.pt
index 5afbf1b3ad4227457bb2c031b71fcc5e64781024..2e1d701be433062676e2cdffcab59b9a5c394fda 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.q_proj/_s.pt
index e3cd08974723488c933e3eb87cfb15730f44ba4c..727137024f2bba9075178ac6394dfb46e720d2d8 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.v_proj/_s.pt
index 31928a32661589f3a4595fe17aeb0a19bd63f4a5..f9f646cf28f529d36d7a0936b986e43f513e00a6 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.16.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.fc1/_s.pt
index e433001a576cb8ff37f91a96389c547ae93f66ee..f133c4cdeee38b7c21a06bdcc1bb2d1cd5af31f6 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.fc2/_s.pt
index b5d1fe11af2a7a4369dcf2f5226d18794057aa85..8143b80ceaeceb3b630942a9ed2d4521d43db0c4 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.k_proj/_s.pt
index 397925b39e6f6c710ee14fd75cb3c183d78668fc..22150c543db4c89ffb73bf97f2364d9bb3845906 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.out_proj/_s.pt
index 4fe29b91e5bd61eef212c76eefed2881e52b0d4f..0192269075015c76c99cb8b87e787182b2ef7992 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.q_proj/_s.pt
index f814f36e119243bd947e5c1203c69568e5c9bd3a..036395e163432384acc600b68b621c055222d175 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.v_proj/_s.pt
index cb599bd09825dbdc3431a68039eb8e76aa6078d9..6b36698f0cde49fa2b79272062211252023f8e7a 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.17.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.fc1/_s.pt
index 15ad41f2b78d508c7f00cc352a21350ecec51c81..f2c8be39cab8c84077a4f1509f085cd954fd2290 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.fc2/_s.pt
index decc6383cf1a9db407c467c0a674751efb7ea60d..aa579b924b6e03eef79bb524f28c8196668e376e 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.k_proj/_s.pt
index 014a5f4eae3701aa41e2f9bf98b1c38bd71c90dd..0dbefb2c45333ce147ae52dc9808059da98bc422 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.out_proj/_s.pt
index f93e5d9903d29e24d452ba664e8bf0a1d64662f9..1966a79d3e1eb1bf43db163646bea99e47eaa384 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.q_proj/_s.pt
index 7970d0380741ae1f6559a2e7e6ab87f49dd67861..6427f4699472d3949fd59f92530ca6fc0dadfcd1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.v_proj/_s.pt
index 665d8417b0fbea9c7a8162ee951c9ab89ccacdd3..4e652c60a6337526996fd2129c92f67be427557a 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.18.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.fc1/_s.pt
index e312da80a95177d61870f7f74cf1635b37d378b7..62bc01496b7eda92088e5063265f0846b1376753 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.fc2/_s.pt
index a282eb8137f15a4b8ffe3cd77f536b2ffb92eae5..f0b112253956a12783c7834d586c85bf7030feff 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.k_proj/_s.pt
index 694def176ed441669b5aaf9af909d3c858682184..d53caa4b803c6c50483c316b4aada66669f6fc85 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.out_proj/_s.pt
index 994573590ffcae4ea976cfc6dd1b103a0f14840b..c3f1b7949a31d0782c33c9446c44898638ad3fac 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.q_proj/_s.pt
index dc743310cb3b471bbfc24f35ca5c485827215467..a7c8399399e4ab245dc5784d3afb04ca32be3390 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.v_proj/_s.pt
index b1b6834921e6286b6febc0fc07db4f4cc0563f18..6f8fef3e9a5404dc50e047a6d9ec5b30b12d6a6a 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.19.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.fc1/_s.pt
index fa63be77d20c8dc35d8972d34fc24779a7e7d83d..827e3d04bb2a4955ab44c3ec9aec57094ebb6e75 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.fc2/_s.pt
index 3b0d7589c4a11336cc3fe343bfd6b8b6cbf0e156..359d79547512d3f064294d7a20e2ce5154bb33b1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.k_proj/_s.pt
index 1234abc00efd220332279a91528707fce95c9840..d2abb2ac4936c76df0f3d95fc5d0ef4b0a12924a 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.out_proj/_s.pt
index eda36af7c6b5d9151ab53ee7954e4e4c25f6929e..a0eab292d37e8344a406f397b35c9d49d0f62131 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.q_proj/_s.pt
index fe9a52d8a7384344326b4f0b14ae1b0b17775905..1183ec1143476faf6bbd67d12c1a9b25f07a48cc 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.v_proj/_s.pt
index f2d3aaffa74ab9784ef977739bef03b307950d67..667d7f3bab1cfc69742cf6bb8ae18ce132f5e600 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.2.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.fc1/_s.pt
index cbabdda709e3640cd831d276f95c96a8f5688ec1..2d8d668d7e9dd78daba0e9dd6e4b21bc1bc59355 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.fc2/_s.pt
index 3d9f9f022a5a29321683a72bf0e33aeefd5ffbc5..2fc7b952a7f1006e7a97b3f3d90ad180d2d08a57 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.k_proj/_s.pt
index 0f889ea508d5aa7b05d052d7f48db22f520a1bcb..5e281b7e3ce2e98bd402d8c20ce320dff667e444 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.out_proj/_s.pt
index 5e2288a1be8058148cf80ceade1fc7e2b1977170..db84c7b164f4c0dfda2dc467ca3cae20a81e43a3 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.q_proj/_s.pt
index e4f461a553f6b82fd95563cc5cc9077b37ffd71c..8bbf7c5a5bedd224b2dbd099cbfbfabc40f75bf6 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.v_proj/_s.pt
index f620c507cb60be6c027155fa2f2348b97a991446..80fcebaaf0c5042784aa47b68ecabccdc29edf88 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.20.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.fc1/_s.pt
index 0e4f7b2a3e89641cc10370cd5d2b059d2d3ac8b5..2691b159d1a52553491c7735ba498d18e05ae280 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.fc2/_s.pt
index 0fb3c5db4776d27a7e84352d1eda4cfa54261d44..c17199369c14fde963c29ee6e82e592a40660804 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.k_proj/_s.pt
index e59ebf79c03168da649cbbec73950b69cc9c3c7c..49a17c807e5ecb2e2dfcf09317cb4bc94586d452 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.out_proj/_s.pt
index 9f84d1c92bf0339edb7b4d01b4e0e8350d10a4b9..9220102ffbe24012f80a027667f5a2be355b03a1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.q_proj/_s.pt
index 5bb92cf651b8b6fb4d18326922d29c3128163347..3a615b3662f25d7a5dcc3e72048c999a66fa988c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.v_proj/_s.pt
index faabb07720da8d9b8a791ccb6a6a793d18049b6b..c4f9a1069321ea43ee80c7b48d6a6ae95d5c32b7 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.21.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.fc1/_s.pt
index 079eadda36fe2190dd8203f6fd05d51d6469f0a2..0f9fdd1195c0f70de771911389fd55fb78cc9128 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.fc2/_s.pt
index 66456027f723f1858651e900473642127297bab3..8b113477e24a750c74f5bc7e3f0c73fb63050bd6 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.k_proj/_s.pt
index 3dde9ef82aa334383b679565447327361ca60efa..6c2e59b0bdc38f06c05f414f9495e8d3def6a5d4 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.out_proj/_s.pt
index 425bbcc18f5f30d288e88e157078dce562343417..f93b7b084d2c0881e0b1056255039456f3e3ac7e 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.q_proj/_s.pt
index 87b32e08f55c064239d91647549b23633b6000a1..561e4f265508a415b6355b706c14c0926e94fabc 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.v_proj/_s.pt
index 6972b00107408027892791a268d9e237d0a563a5..07d20074c8fc79e4252e98eba6b041fd291a8825 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.22.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.fc1/_s.pt
index 3261a4f1c2e1afe98d5c450561b4b898d1f9a6b5..ffd470e57123caff317c488615422bf7ed42117d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.fc2/_s.pt
index 6bc801f8b5b3b406fee435c1c06cf12126f981ac..7b989398582910653d9f03e0f4b1bf530c7db2fe 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.k_proj/_s.pt
index 1d2b3d41c0cf3436f1a0060f5d1e3fa403952387..575afcf041597be85b68199a39add6b688278ebd 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.out_proj/_s.pt
index 0fadebff97a28ee78710befd50fc8564bcad55bc..e220f7764b8500331c533846c4be4876ad20fa10 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.q_proj/_s.pt
index 3ba8df08e09353402dfe0a1f5ebada048016d11f..516cf8a0f0eab7cc4000d993e860396b17fd4f94 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.v_proj/_s.pt
index eb68bcff741c48f66630158d8a50d7862675fb62..aa2f48dd457ffc2d6aff62e6a54e77f93107e539 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.23.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.fc1/_s.pt
index 77b71d7c487bc5bc2aa861f3e7837648b6435abb..1c8c5570e499681c7d2b355c98f64b4c2563c6d7 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.fc2/_s.pt
index 027175835919e9d0a7b22baa15996bbd1c746554..45a09fd8f2adee8684c526a4e2c7945598ca691f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.k_proj/_s.pt
index dbd0f1bd7366c31d71eaaaae70ad04cbd01a1ce0..ef2e013aa1f17d5db2e8b27273535b7386457d0b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.out_proj/_s.pt
index 274f2b7b77bd35943c559403d2e9aded3d797733..d6d3af2648fba2399899cbf4791b92330288dc16 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.q_proj/_s.pt
index 4265ceb277a5792a7bb1a32632585580705dd2ba..20acd0d2c24043667c0b1dd759ac385ce46a4179 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.v_proj/_s.pt
index fafce65ffbe77c4978e2efb5b3d7e4ea8572936b..4f1f5e1c9c48f73016fe5ec09c97f420ff4f06f3 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.3.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.fc1/_s.pt
index 77360a682aebc4dbb46f50b3e57410f1bafa6c19..7928d975ba9755a85703f0c88ad5cc932f764e66 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.fc2/_s.pt
index 91cb55e70303543790ec9f43cb5326a80366f368..f4ae2bad6680254a455c29a4f3a5a2d52b272816 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.k_proj/_s.pt
index b5445d774924feefe19fdc4c509ac54f2f10fbcc..ba4e29a3bb97d0e3a34ea19575aa9ebaee44e54c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.out_proj/_s.pt
index db84ddce430eab1fc52f08d3f857f482156ffaa2..6f411cf360b7f3fb5c2b280998c2f8509b75fc1d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.q_proj/_s.pt
index 216c25cf66b478546f59ffca1ee717e5ef65f5ab..d51ba8b1d467cbad427df2c100cbff0e58e2bea9 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.v_proj/_s.pt
index d5c2953ed1b8fc5983a1db5c88c1c408c8813cbf..cd648a52fff3125a80d3e6bdcc92ee5ddfcd4899 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.4.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.fc1/_s.pt
index aa705e4746e55c040dc3df77c066f35134c2d126..5410787be0f05e08d4daa05c50ce9e2a754da638 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.fc2/_s.pt
index e569f5f1049cbff51aa8cb93e0483cafe412d792..ec31ded5592f67cc9bd371c9076a8341e0b4c264 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.k_proj/_s.pt
index 9bc231d32b5a4e6877fffaf116dc2711372382a2..e0208f88ebc5a42b7efffaef18f9773acdf5187b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.out_proj/_s.pt
index 258767c8a7727d74a385c3262e07e6f67f580c2b..42e11558015a0dc6d747fde046a14f92333f3b43 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.q_proj/_s.pt
index 8f13947526c9fc7615937395d76e4e7a17911043..116472f6f516743cea07563fcc7ed9bbe184911b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.v_proj/_s.pt
index aaacac4aed877c7fa226d88275d1901f77db3106..2d987004682bbd3eb872211b87637fc41eb692ae 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.5.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.fc1/_s.pt
index c53382da3132809b79bfd91e7158949b5b64fd19..cf292d6e171dfbb29d636ab6bc7c477cddab100a 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.fc2/_s.pt
index 3a289384d22a9b8157b929358450b325c57b2137..81f4a7d8d8d49cc923c08ce8c5cc0839cf785aa6 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.k_proj/_s.pt
index 9069a023e4c365e3a1699452cc425e48aa0a6be7..7962518f9189d4e8c0e308823175f23e925e15d0 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.out_proj/_s.pt
index e4ad6066e74194d85ad040e05dace01efbcc9eab..cc726666d72538ad25ff0533912c339c0776b8ed 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.q_proj/_s.pt
index 6836c2d1c4c617e4b6cdd16ef555320ecc10fb1e..95bed5349ef18116aff9dc230aa21923a8745c85 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.v_proj/_s.pt
index 0b24d43e526597ba9dbc784ef35ade40125dd85c..b829110ae72da3d3da9d4f5a22057e82f76a81e1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.6.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.fc1/_s.pt
index e4755809173accb84d88586ba09003ee2db56e09..56fc827e7e94210d98d6e4b726ee396b5ac9cd88 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.fc2/_s.pt
index cfa8ae68367b660afe33cf03414827e8ca6bc5f0..9c2d690ef98178bf4f8dd88ff9ffaa86ff499bba 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.k_proj/_s.pt
index 8a3cc824e6004ad3456a692ea3696a0106703fa0..760e09e5385e4dc66660b8d3feb6ab185e6d2d86 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.out_proj/_s.pt
index 29d7f626fa251a7249a8468ca1a8e4e62eea7d5b..7e1e0d56a88786724c2180c0f702b3c1e42d70d6 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.q_proj/_s.pt
index 708a0bc51d6b9bdabca7c3dc40feceecbf0678ba..3dbc8c968eb61afec7f40f6dda1faa9be4923e4c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.v_proj/_s.pt
index 61abf19f01f591a1a0595fca5da2c00537f1b524..847441fd0978a45c0011d8cbd3a872e435430c5f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.7.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.fc1/_s.pt
index 4e464381e8ab19899689d2417b0cb17c0b9cadce..9f7fb2fcded25d9ca973aaf1aaa13d96c04fcb6f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.fc2/_s.pt
index 5739a309ab27dd166550b24426154b47e8a91420..bd9b2ace713ccdd4e8a2e91f666668aa5db83303 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.k_proj/_s.pt
index 3b09c1071c707d1d8912ec8978d296c0fff70ea8..862f7af2d16315abc003077e5f2ed98730ec1196 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.out_proj/_s.pt
index 3eddacb2c5cc99e8f341d4c83d481232c1c07aa5..7bb52c83a06824a459e47f2bc5c4f88f18d13972 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.q_proj/_s.pt
index 00f146fc932a8725abcf1f0ee7def396a0683701..1125bf15f8b7396aebb2454f83e56b1ded73581b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.v_proj/_s.pt
index 32f9b1766b5181fbb0f54555228c5ae6ecae8048..cf87f90858e2f5640bc37f78a62ac847041abe64 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.8.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.fc1/_s.pt
index 385798e4856ea66acaf091e820a121f34893f8d9..0db00263e11ab7bedac62ba224b0a891b3b5b6a1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.fc2/_s.pt
index 34258df98a3d9db159019ae518e3e04009c46f2b..c8900cb2c99f6cd62a7980a52c43fe9e7be3dc9b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.k_proj/_s.pt
index b18c5620e04f3004dde874a032d2f0fee03cf087..bef7c4a85b0d7be5fe546915bb97e18151fce6fa 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.out_proj/_s.pt
index f60021b68c6358cf01e49be02c3a2f3bcb4aaa4e..bebc8bb2f42db29bdb54629de7aba2191b9c8328 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.q_proj/_s.pt
index a7c571f8902e22d7a2d6e9fc0ba959a72775e1df..1c6184ec7e39773547dfe1c1388491e0f53d4df3 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.v_proj/_s.pt
index 552783683755202c3a62dcd896be25c9837750cc..c32c76924b17e71d34b6c33ee47a5cc885641cf7 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.layers.9.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.project_in/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.project_in/_s.pt
index 0891eaf3ea890f8b41bbd27dcfd849494224acc9..e8c9b55d2186ce2fb593614c9441f5797e448474 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.project_in/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.project_in/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.project_out/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.project_out/_s.pt
index 2c9338e593d7f174121d11f1814abe2b228b4321..2bc34ba3ddb110888b3d02717da5c7032b82b9fe 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.project_out/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/init/model.decoder.project_out/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/lm_head/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/lm_head/_s.pt
index 003424abc48df87a17ab4d2eadf522bc2080907f..cc08434da2efa62b6ceaf20a508035c8ccd45538 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/lm_head/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/lm_head/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.fc1/_s.pt
index 2b7748d75eb958050828cd0ad1765658c0f8895d..710619d374b69ef8b907d6ba7e7e37848d6233cc 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.fc2/_s.pt
index 04e79718d0e61f2cf51cc9c746c2e8f0ffb5be52..53e5d26cf664a55f006c3967ffc8ac90f19f18b1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt
index bfea36bf8e414d924fe799b4dac1c0fa63af4599..b488ed176250cb14a1250989149e3e7faff756da 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt
index 6ae8f6887e9ab9a07dfe1f7d512447d2645c830c..8a3a536218e239d069216aa4e98a5410dfaaf284 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt
index 7e131da08534f1a9963e717127cd829301c2689f..eec317400657cb90a2bbed82986dfb5b3285e045 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt
index 737f6e1a2cefdc3b95b42dbb963b9f154294fcfe..590e225c564af14bbeca9832a6bd3b28c47a5689 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.fc1/_s.pt
index 38ba522e8937db9d7f728949cb96a04d21a70c70..19132f5c74c11a012ebdbf0ca117026ecdb2a378 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.fc2/_s.pt
index 173175c964f1a52932bef0dc481b049f5cd38deb..9f3dc2bf8df1730675772df29cff49a88f781145 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt
index f9874e80701d3fe0c66fc77dfd4b0898682471a3..69a6c4dcc2933bd6068f7beedaff4da851e13dde 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt
index 097a618ebe450b2543c213103217bd7efa1b8485..b2b1b18d411114bd1827c9e5aa6b51202d681b5b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt
index f8626533ffe866f86388fe66b3760a1b682e7f52..4d1d4544809af29296697bbda3f94d2902da4b17 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt
index 52a0ec74036e01f9f34029b72d0a11d9e97573fa..b729c88bcf5817dac392924cb21c7b5ad3ce7e4f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.fc1/_s.pt
index 816fec0e4733ed986b611da64b0a439c6514fdae..f611a2d3139997c3c55d82fa4530a29613dd9438 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.fc2/_s.pt
index aeccc91c050ba1f802ed920ac52d569419bae6c0..2cf1a3b070750933643966864cd28ab1b23a103b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt
index 86c4d36dc654eaec8f3ae47ba323a6ab43138cff..b40bb7eda1d97e54d9b091bcbd8de7d36621f3f9 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt
index 90beefeb57c98fc63a0f7d02e9a2cf7b323ddc13..820fd329f467e2dde0afd8c47d9ae6bcd8e40609 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt
index 02d2a1a660cd468cb03d2b3130cebdb74832a49c..cf8d0cbadffbb5d204fd240b798662804e1f7ad6 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt
index 29db503b4a30708d429940868d9053ddb9216d99..6b5031759afff09009bb55f817c5cd09989e00b5 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.fc1/_s.pt
index 435751ecb2861ffbbb48cd72d1bd017725172911..8f637e17d403591b9cf658796d232cd8ec0b0a81 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.fc2/_s.pt
index 967e7805dc516c0fef4ceeb873c2a761264c5690..ae0b8afb01a435d5d0411988b3d1e01d1be89337 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt
index 0d5868f3f13fb164cb665f5fdbbbec8caa4e446e..350b3dabe5fe38384fa43f221daead41ec245480 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt
index a4174202aedd0ff4d99a0fa6e23ac6d1e73e4d46..ca0caad1b35763e4c0250db904b5e6879dc1a9fa 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt
index ff8009adc0c0d3e26295c4cc528abb9390b7dc42..019fe1303c7d6d9da3970293afcd3a551659424f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt
index ffe6c73d95fcfae2e297c4fd4f19aa347e083353..b3df88daf2513e9d20442b666c010ca1bd0edccd 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.fc1/_s.pt
index e9c586ba031cf6475ad9f059183c8bab885b4c70..d37ca948ef1991ff6d05d06a5591125b495c1456 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.fc2/_s.pt
index 59efce0ce3ab88d65736bc8679db2016a87f0776..9af401e96a65add25f3f1e06bff4e95e421bd768 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt
index c3ca698e71524b49edf48db783d903d2ad4e8399..06800f722f3e7b2a01fb530291536bce003ebd4f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt
index 87d620cf5a47243b07323250834427bbddba4870..4fbcd5f6753fed232ea116025099d614fa99be09 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt
index c7efebf54aacf99f211bf367ab035c6af99ed838..9d7f2d4a72bc25571292c7808a1ad5bbb9f23958 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt
index f5fdbdc6f60e36ee2297e012ee404a7fe236f86a..897b6cc42b816fbc4e367474eb6804917f395196 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.fc1/_s.pt
index a28f0c0659af6af0eae24fb6891faa6ccc4f3363..b5d9cb6af65c82ab59b65e0885a6acace8a33753 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.fc2/_s.pt
index f9511c70581911b95bb230229072e1339ad39c61..71d21b8ce1284d71976fad478393fbe133ae018d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt
index f83952ad2125b17a42fe38525bbaf3a2b04bc60e..8efe8fdf9913359d1e8edf3f9654e19635d282a5 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt
index 7d5700b9c7949a006c9e6c11320085884bdfbc39..08142117b5f976cd03aea7388da442e169805ec9 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt
index 586b78e17ab9714d5ac4f5666e322cc1595b67e0..1e88ad2c7447a2b9ce18f11c566f7c007a2431b8 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt
index 423ed359f734b39b78b4b5d54ceafcca342227a4..fa09e6cba8e7aebf438553d4c859970c33b1a546 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.fc1/_s.pt
index 17df62825e8b15781d2c7eba30a0f2c3012d498f..fc3126ba19d08d4bf16da1efbd3999c305557d3c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.fc2/_s.pt
index a2d75c0f6a80f107c4736b246fdd458a95bcc48d..dd85b2794105ecf43781d3d3abe1aae3cdc91a4c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt
index de18ced1a72c3fa99b22abe5d90ac63bff158f1a..8a32fe69837180b7be7fc5f8bdf4d39a20e8e495 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt
index 336e755be4759368afc8c42b089adb1b09743c95..3efda905455306ac344d8ae9d53ed09c54ce461d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt
index e4fea263aaae649442b78c1f89e3670c9b3f8db5..d92b5034ce9b591a24450be021bae6eeb81b3dc2 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt
index 5a479e293f96737847c54db8d12a3b1b51d6a913..47294a0f19afc4e1aee3c4d84ca28d453d7aa304 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.fc1/_s.pt
index c4f3a66268c1117ac60510b7e259e8ec4bed4713..2728a4b1e832d2997140c47d6868ab227cca527d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.fc2/_s.pt
index 97c0018820d9e1cc0f6f6eb0b1f05f5fc3884fc5..c99b1eee0b6e61d0a85800ef87f429a998933ba1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt
index b418357abe10a4021c346c9f31b4044191bcb6f8..5494b5d37ef1829bf76133dbfbc9672fb62fd2ea 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt
index ef51299fd95ddf0a63c04bf431427842168ec503..0b315ff73edcb6381b870a05f0f970b78b41d383 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt
index ac2e84c6ed4fdc05be903db0adcfd7d46dcedfb0..c44c11f7ce1b4d1c86d11b5a2fd7361631e0ac01 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt
index 6b5610a6a6236e983cdd44040bb6b3f677f20caf..61a94414140881ffc85649c7d2fde4704a72f0b3 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.fc1/_s.pt
index e073a813f3457ad99f5d540ef4553f90ba0a93c6..bc3985ab14b70de41af04858bacc8ee15533cc9b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.fc2/_s.pt
index 96510afb616d15ad16c1456d517de4070317b8fa..b0b424d43646b815e9acd873c935e053bf2daec4 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt
index 278cdc0f7bc7bfb5562008aa38b7ad74ea7fe3d2..4529d61cfd74a4be8573744d6dd0cfe5dcd6e9c7 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt
index ac2cd01031d9c99d80bae19395e4133ae3c23940..accd99f9e04fd0159ccc4412d03b6ad856ff5b1e 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt
index 8fe970a38896b0fd8e32e7f16907f74098387c43..1dfb83e1379ca3a0066d825f784b19b979d8241e 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt
index 93f0f8c84cc67b47a52be6aefc1145d88f51046d..9de6804cc448f9a99858df2a0eba45a9cbc22610 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.fc1/_s.pt
index b42dfa78b2e06b94f689b6e13098a360cbcc4dd3..97f74af4bbfa817e5245d8786bcebbfa13a32360 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.fc2/_s.pt
index a178050ddef19864b52b2920f54867e369c045ca..0840cdb15ce6d44fa8888ac241432c4c033ba21c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt
index f6864cf5da861447e80f733c7f69110dfbb11765..59cedfd0e085c3b38f8d68fe80a2e9bdcd7557b1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt
index 09004b78fd29935fbeb63593e14fe560288e93df..b45e132a956c7d60052461a9ed1f460ace48ac3e 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt
index 005d894bb3b315305ebbd53de0f62bcc82e72ba7..820e6358b9e4fada712c3cb28d15c60b920e232c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt
index cf6f3437e7ce3ac0717effbd75691c44c1b3f1e0..261d84e9326141f27fe23b40991660a75197ae67 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.fc1/_s.pt
index aed52259e1e581e96e4db217cced7e1b5aa5b8b2..ac70a9a6ae85708de5d2500eabfb793fa6405866 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.fc2/_s.pt
index a8f87cea4a886fe9787a1662f38ea51fae28c917..43a05afe5a98efff97d123329795234f224e3588 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt
index c7556045eea18cbfae49d32981753046cb2147b5..f971cba7ce6d4d18ce703833d11d8d8439511610 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt
index 9e1b1b5154769f20e9391635cf46acc87d3fe392..ca7723a72915e33fccd6a991f0b292f9c330b6cc 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt
index c5bae4b2d636042b1e4420a8f5b2a2179f905a1e..f506bfca34e907ac054c9cbe7848da74870cc8a7 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt
index a6bd9762d65fe6cc5b83b31c62d767d7a391c428..4d05b88c12590b5de27b6fe677d2e15f86b16a73 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.fc1/_s.pt
index 94b3e57ae4329aeb3b466bc115fb878225b10125..0bc77a43a254e0bc15826aeb4e22867dc621eddf 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.fc2/_s.pt
index 41bf2ec5b7c2629fe8e7ec44533fe1bdf1ba3f71..014c9248d66bf4f4a9832739ca5c8ab7fb4d95dd 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt
index eee070c130344db0e84b524720b7566a51e7367f..deced2a7ee30963d7a8931dedc6c155dd1244266 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt
index ce5360da371b66abe36b85905d3f53555a5ec752..071c62db46a22522fb0647299d6710e097efd71f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt
index 4fdd0a1d0051e491807a91fac4d6a69ce8062e60..2e5a32eaf7e90c436ed9761a83de27d5d3f699b1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt
index c0db46f0c815bbeef1e61dcd45fbf5d428a5f235..30e53823020dafc9646f9673cd5f6ca36608b71f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.fc1/_s.pt
index 3431d1e325fd30b67648810ea0daafad17454b6c..b04982c0c2b561523f8b6e471d7c00d55b0a433c 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.fc2/_s.pt
index 97da9873131958ee7eab1625feb0fecede2ceeba..788215b2fabc72511edc46c07d56dac23c2aebcf 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt
index 637c949e0d64cc758cdca28d860cabaea9907c47..307ca644cbd74fa0c1f337e82ae3c5c247eb9582 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt
index 7516ebaeff8b150fee74aa0f529ed9ed57facce7..e1b4fdec04ec7a2146036da7907e4007e19b032e 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt
index 3b10f0c4e7b51801baadf6613a59f0c09d4b2c99..37a122943646532629efadd8c53f0a14e6d7ba4a 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt
index d45351834111897892fb3041cd74d13c435c4fc7..a6950367e20549806d64160d4c039c4f2511a475 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.fc1/_s.pt
index 96ffc8d39164d67d2cb0a6930ae71fbc37ed3f24..03ada8bd4ac716295f63e7828faad26f2845a4f8 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.fc2/_s.pt
index ab7fbf2c095cae46630a22682eabcec2d43e63e6..cae641d49ef6ccd175ff7f38deab0cca9ca54fd0 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt
index d382160392b330ed31ae333bf50dab1cb78aca7c..1dd87ca983be2cac68b25690922dab786dc34146 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt
index 7e20a2c94c4c0945afd55bece5dddcc7e112e28b..01acb2873e4279a23cb3de4659fdaa5cf232a3fb 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt
index 4c07c4786a425b1fe34fbf43b4f600b2f4c17b3f..4908dcd08608cbdb7b2a98df8d6c1cdf0f4ea2c6 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt
index d7f533d6f3d54e92c6f6ef86de46df1a2b31d71e..7d272c95ceb185d62a8b3cdaaeb7bcff27f1add3 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.fc1/_s.pt
index d73789c17963d22a0b6e0d9ae445a8ae921ceb99..d1d18fdc1408538c7290c776fbbd2cd0c1a9b51b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.fc2/_s.pt
index 67f5e85274947ca5024a9d33c895808f173bf1f8..984411351995e9da49f23c68449a701f03926db0 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt
index 724ed106b3ec694d45115e8815eef840fe761c6d..20284c300c55de5070b0ce6387b096f173996c38 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt
index decd01cf8215130c6537fc4e522ea1d1dca9efd3..84a8ec71d80fc5d402bff4be5716c3daa08f8a59 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt
index 1e3be687519903605ffb49d6f71a0cbaa8d650ca..f75bf0aff5aef4332f9a76398f5ff17e546bbbd1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt
index ec2f4f5944c52dc2c1b65d1a54ff0e441f7864d6..49832783a3f15da3133fcb4095d34f06a5fc886e 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.fc1/_s.pt
index f0bc7fc300528241a9744a2bfbf1c3a7156b3408..311d33ff4e00899d8967793fd4b2267afe0448b3 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.fc2/_s.pt
index 94153243d0d3725d0a9c1d34e1a848c3778e4bfd..1a1b51c1622430a9feeb3359d38b98889be6542a 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt
index 791aa8b2b6e8fe3e11fc91c2266b44dab0431501..8056e38188c03cc1c885b217ca6bbf112cb6671b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt
index deaf594a06de0c162deee3b75024b19ada7a2772..7564d496afbe3225cc14d26dd92d1056896a60a8 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt
index e373de034eca3b7d68b20503ef2739172331d618..1eed81b531de70dfb42d8e3ad57276c146bd5ea0 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt
index 96a2082e67f079f5652261fd3a7f195c745d7a00..629b7bde644fa4d2f0eb505d6bc7ba42de10ca37 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.fc1/_s.pt
index 0867ae637c5f2c4001b9046f46f8cdf97624c186..ecf3338417d7891931c7d90b229903c6ead2d220 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.fc2/_s.pt
index c4c114451d361d91d9eee7634775e256d7428a1d..18953104ee779469724be125dbfdc241209e1339 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt
index 849bc2cafd71d051661d00fa6fda5b786a6a15fd..b83e7522c84ef699cf8be3d5c1579ff4ddecb4ab 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt
index 41620b780c7acc3efc2135734b147a6b7f4b3a68..b6c82c098582a10e2c99fe695e3c0146cd6459a3 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt
index 294636ff45f80a5b86e2cae75d838ddee2350061..18d34b268ea784ca3501e4847eff305d9f7d3c8d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt
index 469eb9a5aaf27ff12471ef7b5006f686879031ca..744368c9a404ae4a14235358cdbada1d6e7150a1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.fc1/_s.pt
index c883505b976886fcb42fb48b81edfe54ed228d9a..cf3dc9e8c96a8aedca260307862a8c1440eb29c0 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.fc2/_s.pt
index e22de76edd871580324280cb1c61374e73e498e0..184f1800690400f79d8824a44f4dbb6fbe7d28a4 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt
index 1868e51738ca19e6e9041f21d547922b16afc1e4..ebc56cda0236fa4c24f9f6dbdc0e0d2059fe2ee2 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt
index d7dac92810d1be9edd0c7bf93962f5f48a7baceb..c923178e3fdde5816363380a89458e6f35751533 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt
index ff8842ea5028a795bffc1ebcf7b77429b60cf37e..755a17a66bf1cc524115e399ab9637b2f981c875 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt
index 29942e76a046fcc09e1eec4e3caaef1189492510..4f6e3264d2eb84e72d195f191c70bc1be746b937 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.fc1/_s.pt
index 042a4ee45d5ddd4b4c69db1046d5fe2527201ba4..80c10a9bf29b5f08af148f706b53916bbd2ce4e4 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.fc2/_s.pt
index 40ac2835c1f797a815d4faf235c88ce16b64458f..518e4b91e9425842d1d835f5d307c46f726547b9 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt
index 0ea6dc48862e198246348270f27f8d98eab7e0f6..3c77148695c8fef6a89b13457ea09c0cacfbde83 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt
index d199873a953740cc1f35f399b4e99f1baf3961e0..18850d5041698e6011cfd87460e304c5ee5d287a 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt
index 49fe62855d0758fbcad5e9367e33bd9c899ad37e..b75cdf15ac7de3ef868daae7ed5142630fa3a764 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt
index af753ec59eff17c1ae97fdd3cf7e739c22b76902..41e285dff9ff93fc69990ce28f5f17713552a684 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.fc1/_s.pt
index db515ec47076e40062f13c463374690e8691ba6c..65a5ee37ff59af41bc5b292bdaf6addf4ab610c4 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.fc2/_s.pt
index aa4d9893df5a30ed0ade3b938df92f38b3b00e9f..5ef5a28ff3e66f50bda6038613fb8b7e09b94e22 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt
index ff264f6fe8f348aa35a06744a181c81397b7937b..7f7f3f6c8eb102cf493da5c2a974b08eea482daa 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt
index 5af9efa796ef455c94b33d58c1acfca5788d065d..9b78d6bcf391aa907eec3bba9e55d487b335ab19 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt
index 3c84b7b55536a8f947f9fb21ec225b85e5c53a27..bb7e9700083344c79a72a46354eb84a78da499c9 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt
index 221bd4329b7402d4573d87a10c365b7bce1f2806..3b25b8f84abbf04fa1ae94bd2a878bac9712021d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.fc1/_s.pt
index 7405d4ab81806f53b3c0542d19e2008bc52feae0..6928e9960c0a714dd3da341b3e2736a09587fd10 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.fc2/_s.pt
index fc67ec27756e2dba85158b4947fe685d7e175c4b..57187d7df65a6f0198f1f1f37ac5aee3f2e43d40 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt
index 9761f242c982044e6e29ad812c13561e1f12fdca..06fb199664e09e1a902638d817aecea8c0f8d25f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt
index 08149d0d22930145035407e0427e3fad1b746ef1..842a13a0f5695a659d0ad48dfc9e8a3c4dfd390b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt
index 970409c9abf54a2c2d8f585f86d567788b78e3c4..31dc474bf72e9b1409982058e896c43c59e4aab4 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt
index ade54f04ff83d20c3e936a2580fafe88f8af4f21..d3690632d795a7b0b2fbf3febe006851e803c989 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.fc1/_s.pt
index c840626a7615c73ab3f16658e2be82070408907b..af927733e176beef0c93903d46ba5ce907ae99f0 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.fc2/_s.pt
index 401336b91b9c9a0aaae4d55a9ff770b82055953a..2e7a58beac07345985cac5ea5c48b48491fb7dcf 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt
index abeb0dc443c7bea53a48bcab775299af0d107059..c86b6118c3acbd835adaef0096eefeabcf8d149d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt
index 31147783cfdf8efce4f424b8ce805bde5ab10f6f..a2773ddbf2d6a9915abcd5bb11e4385751c1d9f5 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt
index 396e29d3a6ee897d20b9955decb578c09f184cb8..fa13f9f6c616c49d317ce44e5361f60a679262f8 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt
index ed1a46f3db63757b33790cec2d36ae7a03a5395b..e863cb8fea45babe3f974453d2e18c70c76a62e5 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.fc1/_s.pt
index 962c52583f9b3136cf10f1b2c566b8129fbd701a..5b8f0c7cf9710109a6e913e5cabf6ece4cf81130 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.fc2/_s.pt
index 97ef07753936f3b836ce974607176edbeb98cad7..3b553ab34eeba53571f95d12c159137bfb6707ec 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt
index 8665e27ee5b276031d97c511898201f501e0859d..22cf32fd7ad965cd46361fdc33f779fce5bb8028 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt
index 862b05c3bb7048054e60b510bd6760ce4fa9afd1..9032e8737f33b9acf1d8fb7d243168b5c5a6785b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt
index 3f65dbb650bef93fbc5890c9b7ab06960a8f6959..e4da22200b8036a1293dbfdd06143092f27a60da 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt
index f3449949b27eea83b24008220cc168766f062268..70e8e72ac07edda30c0b3370b8243eb7754dd2f1 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.fc1/_s.pt
index e0ae47e4d63004b88df1452c3f4788fdac610efb..9356c375d2a03eccf26983e390f9a6b7cf49eb04 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.fc1/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.fc2/_s.pt
index 65ac92d4694762537f40f32e38efdfbc60742162..3533bb77dd6311da1d92bf484172d85b881aa29d 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.fc2/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt
index 15ce2f77b353fb5efc561e6d315aab650759687c..c3372ce830ccc6f80de8dcc114f662c4062774d8 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt
index ecd65a3d41baf5637f235328da6c6eda701725f2..c1b1a77a34f21954bcdeed6344472c094f03565b 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt
index 96544e1694ba290455d06c26457b195ac75c8fa9..d7a892edd96c5e3b1446ab5378519201a1fadce9 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt
index b0a8772dfd1a0d891a2021f2a4a639c3acf2a3af..534a231a00f44a8d9635fcc97e0074391d1420b4 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.project_in/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.project_in/_s.pt
index 2a586023fbbaa2df033574b4dbb78393d6671d95..931ae1ce0a18c87d54b3a417e19f62a6b1bafa9f 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.project_in/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.project_in/_s.pt differ
diff --git a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.project_out/_s.pt b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.project_out/_s.pt
index f90bb44cc6c7a154245b980271915eab54a412a9..a46c988ad667a3d22feb4d352e875b13a38a4cd5 100644
Binary files a/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.project_out/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_15/trained/model.decoder.project_out/_s.pt differ