diff --git a/adapter/adapter_config.json b/adapter/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/adapter_config.json +++ b/adapter/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/adapter_model.safetensors b/adapter/adapter_model.safetensors index b7733ff441ada9ccc544a5d5bf1c6a52935bd15e..ec9e58fe27d7c135c6a63ecc0be40cf7daddbcd5 100644 --- a/adapter/adapter_model.safetensors +++ b/adapter/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e3be2c8518433696a6aa5b088a43823facc099b20c7beab385e9352ec3023f2e +oid sha256:a595d437f4fcb0c6030840d9e70e7ff6818a5e490943aabe19f00c3dee6ba9c5 size 239536776 diff --git a/adapter/checkpoint-1/adapter_config.json b/adapter/checkpoint-1/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-1/adapter_config.json +++ b/adapter/checkpoint-1/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-10/adapter_config.json b/adapter/checkpoint-10/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-10/adapter_config.json +++ b/adapter/checkpoint-10/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-10/adapter_model.safetensors b/adapter/checkpoint-10/adapter_model.safetensors index ffb7514976a1c3b6b066f127037ee57121ebb387..29312e47dc7b6ea0ffa9dcedc5b0e29a51b3332c 100644 --- a/adapter/checkpoint-10/adapter_model.safetensors +++ b/adapter/checkpoint-10/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:05d17ae14cc83cb1bacf66ddbf8a731aa95e7ee88c27331d712b7f05316ab93b +oid sha256:2bf31fa16c7c88660017f477ad7d02c7780ab77c6958e8963d96b8136b997444 size 239536776 diff --git a/adapter/checkpoint-10/optimizer.pt b/adapter/checkpoint-10/optimizer.pt index 5bee592948320dba1f104883751d009efbba2193..c1d77c6241dc3aab9dce4942bc7a442691629573 100644 --- a/adapter/checkpoint-10/optimizer.pt +++ b/adapter/checkpoint-10/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:5f101ceedd36da51a74e912ba21ac9744d4bcf49d3acf3677c9a6a4f2f45e9b5 +oid sha256:3ba1ae4d12685af72d6883a2ab8c658110ab1baf47b7c5129877df055c8f519f size 243818437 diff --git a/adapter/checkpoint-10/trainer_state.json b/adapter/checkpoint-10/trainer_state.json index edbb0fdfa9e8d4554f74928d0aae3ef0dd229230..db478b164e434ae3e36414c12b288bd1b8f9b860 100644 --- a/adapter/checkpoint-10/trainer_state.json +++ b/adapter/checkpoint-10/trainer_state.json @@ -10,12 +10,12 @@ "is_world_process_zero": true, "log_history": [ { - "entropy": 0.8513693287968636, + "entropy": 3.8962098330259325, "epoch": 10.0, - "grad_norm": 1.34375, + "grad_norm": 368.0, "learning_rate": 8e-05, - "loss": 1.310697078704834, - "mean_token_accuracy": 0.8319162212312221, + "loss": 5.190721893310547, + "mean_token_accuracy": 0.37155905868858097, "num_tokens": 6910.0, "step": 10 } diff --git a/adapter/checkpoint-11/adapter_config.json b/adapter/checkpoint-11/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-11/adapter_config.json +++ b/adapter/checkpoint-11/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-11/adapter_model.safetensors b/adapter/checkpoint-11/adapter_model.safetensors index 5db8a313ccfab87261d2d0bea6c5bbc6d411bc92..2c542fd310356464674ac427462e1b09cb46142d 100644 --- a/adapter/checkpoint-11/adapter_model.safetensors +++ b/adapter/checkpoint-11/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:9d9681efd88617c5414ad8a2f16a9fe33ff44ea2cff0324cee219e52ebabff7c +oid sha256:1349bcf09b314d4e7c6011d14578a8ed965f03e7221a72299920aad5ff281fe3 size 239536776 diff --git a/adapter/checkpoint-11/optimizer.pt b/adapter/checkpoint-11/optimizer.pt index 5eefaafba2fcd9dd087d2e950fe46ba389c196cb..98f481ee16e2263de41998e4e60ab169f6e4eca3 100644 --- a/adapter/checkpoint-11/optimizer.pt +++ b/adapter/checkpoint-11/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:2fa7a3a67810ffd51f4413bb310c5e0b07749a8aef580fb8d5694adec4318e1a +oid sha256:d81b388a1cf32781d57740a88bc414884439ca3d4c067b097af482971c8c036e size 243818437 diff --git a/adapter/checkpoint-11/trainer_state.json b/adapter/checkpoint-11/trainer_state.json index 36982e121c080883b30cf9898bffa33ffd33651e..5d9165b2c247f44effac76a75460d8f6077064c0 100644 --- a/adapter/checkpoint-11/trainer_state.json +++ b/adapter/checkpoint-11/trainer_state.json @@ -10,12 +10,12 @@ "is_world_process_zero": true, "log_history": [ { - "entropy": 0.8513693287968636, + "entropy": 3.8962098330259325, "epoch": 10.0, - "grad_norm": 1.34375, + "grad_norm": 368.0, "learning_rate": 8e-05, - "loss": 1.310697078704834, - "mean_token_accuracy": 0.8319162212312221, + "loss": 5.190721893310547, + "mean_token_accuracy": 0.37155905868858097, "num_tokens": 6910.0, "step": 10 } diff --git a/adapter/checkpoint-12/adapter_config.json b/adapter/checkpoint-12/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-12/adapter_config.json +++ b/adapter/checkpoint-12/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-12/adapter_model.safetensors b/adapter/checkpoint-12/adapter_model.safetensors index 86119b0cd91512a2de49d44455878a185e5c3d7c..8cae5821f667e897a8c29816da4d93e9307087e6 100644 --- a/adapter/checkpoint-12/adapter_model.safetensors +++ b/adapter/checkpoint-12/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:6e2df4368fe55f5a79123103fc8d9321193f5a1c9b4771867f8c8a0e1d214108 +oid sha256:595eef2f37979c4a21aa3468807a944d4ef916f631a812a27c6d807dd1e44d7b size 239536776 diff --git a/adapter/checkpoint-12/optimizer.pt b/adapter/checkpoint-12/optimizer.pt index 42fdf661ed6a37c6dea4dccdec553095eab890c7..7406d2fa3c082dc210f9267ee4cb42b9f213fee6 100644 --- a/adapter/checkpoint-12/optimizer.pt +++ b/adapter/checkpoint-12/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:effe09760ff16f3e0e212c42ab9a3bf3df4644871b0cd48be8e746b423d3045a +oid sha256:a77a4ed360343bc275a374d0f7647552aa5145e9ae1d8312a8af34f8a4fd404e size 243818437 diff --git a/adapter/checkpoint-12/trainer_state.json b/adapter/checkpoint-12/trainer_state.json index 0a146af766721317b8649f537f0d22ffb7495671..fcdcc6d2caf567a2144626fb5a897838a1407ef2 100644 --- a/adapter/checkpoint-12/trainer_state.json +++ b/adapter/checkpoint-12/trainer_state.json @@ -10,12 +10,12 @@ "is_world_process_zero": true, "log_history": [ { - "entropy": 0.8513693287968636, + "entropy": 3.8962098330259325, "epoch": 10.0, - "grad_norm": 1.34375, + "grad_norm": 368.0, "learning_rate": 8e-05, - "loss": 1.310697078704834, - "mean_token_accuracy": 0.8319162212312221, + "loss": 5.190721893310547, + "mean_token_accuracy": 0.37155905868858097, "num_tokens": 6910.0, "step": 10 } diff --git a/adapter/checkpoint-13/adapter_config.json b/adapter/checkpoint-13/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-13/adapter_config.json +++ b/adapter/checkpoint-13/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-13/adapter_model.safetensors b/adapter/checkpoint-13/adapter_model.safetensors index 03ae81dfd6e2a3239cc66bfee6ff78ac4718bc3a..a38a25477f8a3126b6fe8f5badd2acddf277dfa3 100644 --- a/adapter/checkpoint-13/adapter_model.safetensors +++ b/adapter/checkpoint-13/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:90c8bf5cc5e3a9080536acd4e22d9b395f3b96ac285a926ec195764e9a3ac920 +oid sha256:0f4f5dcf4854154b3748d6241b60e69d6cf4d81671433917d80649ae3e7ebbec size 239536776 diff --git a/adapter/checkpoint-13/optimizer.pt b/adapter/checkpoint-13/optimizer.pt index 7cc75c0cfdf64170762c60f3e4563b2f267b352b..bd772eaeeb9f4fa99a7f71927f2f125435f47a36 100644 --- a/adapter/checkpoint-13/optimizer.pt +++ b/adapter/checkpoint-13/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:8de3a29fd07f2ee3595d0ac98721df096041694f03ff5aa11649d155f5ecb5ca +oid sha256:e88af799967be423526730277d7cde331772ddc7f14bfca26f3bd4bbb34ca4e8 size 243818437 diff --git a/adapter/checkpoint-13/trainer_state.json b/adapter/checkpoint-13/trainer_state.json index 647e953f449c6cdc0ab1df8e98363255e8ce43c7..cd0d9a966a38b4e78613a9fa4a343e521a5acc07 100644 --- a/adapter/checkpoint-13/trainer_state.json +++ b/adapter/checkpoint-13/trainer_state.json @@ -10,12 +10,12 @@ "is_world_process_zero": true, "log_history": [ { - "entropy": 0.8513693287968636, + "entropy": 3.8962098330259325, "epoch": 10.0, - "grad_norm": 1.34375, + "grad_norm": 368.0, "learning_rate": 8e-05, - "loss": 1.310697078704834, - "mean_token_accuracy": 0.8319162212312221, + "loss": 5.190721893310547, + "mean_token_accuracy": 0.37155905868858097, "num_tokens": 6910.0, "step": 10 } diff --git a/adapter/checkpoint-14/adapter_config.json b/adapter/checkpoint-14/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-14/adapter_config.json +++ b/adapter/checkpoint-14/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-14/adapter_model.safetensors b/adapter/checkpoint-14/adapter_model.safetensors index cdac2c0ab380df7093481859bcc49553623437ed..cae98ecea767cbf6e83f0039bb150f66738a1b33 100644 --- a/adapter/checkpoint-14/adapter_model.safetensors +++ b/adapter/checkpoint-14/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:d50aa8b388ec8733ac44bd472b9465ce8dd25750a07b47bb921b8abe78e93449 +oid sha256:61f5c3b0328c37da03c486ade9c2f7c128cefbbb64a64d30331445806c06066c size 239536776 diff --git a/adapter/checkpoint-14/optimizer.pt b/adapter/checkpoint-14/optimizer.pt index 25b100b2e210c879ed81b0800681353cf5e73aff..475c9e73c57406324244a97cc2e2b42c9b506509 100644 --- a/adapter/checkpoint-14/optimizer.pt +++ b/adapter/checkpoint-14/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b435012f854d460c7c66d1e22a7d951f7b7fc22aa9f957bc2f757111325d4c16 +oid sha256:d40565a8090ccbd941560ea637ff66f1a1d5f370441a578ceede47830bff632d size 243818437 diff --git a/adapter/checkpoint-14/trainer_state.json b/adapter/checkpoint-14/trainer_state.json index b2edd6ca4191a487abd2fb2eccf5976fc6101de8..19e47dc1bfd00481deb795ece74146ae4edc7078 100644 --- a/adapter/checkpoint-14/trainer_state.json +++ b/adapter/checkpoint-14/trainer_state.json @@ -10,12 +10,12 @@ "is_world_process_zero": true, "log_history": [ { - "entropy": 0.8513693287968636, + "entropy": 3.8962098330259325, "epoch": 10.0, - "grad_norm": 1.34375, + "grad_norm": 368.0, "learning_rate": 8e-05, - "loss": 1.310697078704834, - "mean_token_accuracy": 0.8319162212312221, + "loss": 5.190721893310547, + "mean_token_accuracy": 0.37155905868858097, "num_tokens": 6910.0, "step": 10 } diff --git a/adapter/checkpoint-15/adapter_config.json b/adapter/checkpoint-15/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-15/adapter_config.json +++ b/adapter/checkpoint-15/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-15/adapter_model.safetensors b/adapter/checkpoint-15/adapter_model.safetensors index b7733ff441ada9ccc544a5d5bf1c6a52935bd15e..ec9e58fe27d7c135c6a63ecc0be40cf7daddbcd5 100644 --- a/adapter/checkpoint-15/adapter_model.safetensors +++ b/adapter/checkpoint-15/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:e3be2c8518433696a6aa5b088a43823facc099b20c7beab385e9352ec3023f2e +oid sha256:a595d437f4fcb0c6030840d9e70e7ff6818a5e490943aabe19f00c3dee6ba9c5 size 239536776 diff --git a/adapter/checkpoint-15/optimizer.pt b/adapter/checkpoint-15/optimizer.pt index df78b83007f32859a84d45b7a31485ec1148bcb5..327314bc6c74ded18a443b53f1ce429e3f7b9a79 100644 --- a/adapter/checkpoint-15/optimizer.pt +++ b/adapter/checkpoint-15/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b4935ec169e25391ef800c9705cd37a3a5b35a3c6cde2dd9395cae3ff03c3ee2 +oid sha256:cfea4b6d476419e34dcd29293fede2338729433dce76a040899df509fcb32b5a size 243818437 diff --git a/adapter/checkpoint-15/trainer_state.json b/adapter/checkpoint-15/trainer_state.json index f84b0825d292ddace1a5c055a2db3dcaa203d2b0..011ca19410bf83ee59b853cf7d3689ae82488f5a 100644 --- a/adapter/checkpoint-15/trainer_state.json +++ b/adapter/checkpoint-15/trainer_state.json @@ -10,12 +10,12 @@ "is_world_process_zero": true, "log_history": [ { - "entropy": 0.8513693287968636, + "entropy": 3.8962098330259325, "epoch": 10.0, - "grad_norm": 1.34375, + "grad_norm": 368.0, "learning_rate": 8e-05, - "loss": 1.310697078704834, - "mean_token_accuracy": 0.8319162212312221, + "loss": 5.190721893310547, + "mean_token_accuracy": 0.37155905868858097, "num_tokens": 6910.0, "step": 10 } diff --git a/adapter/checkpoint-2/adapter_config.json b/adapter/checkpoint-2/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-2/adapter_config.json +++ b/adapter/checkpoint-2/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-2/adapter_model.safetensors b/adapter/checkpoint-2/adapter_model.safetensors index 105f732bb64e19ddf04aafad5561a8bf6195e6ba..e3b82fb4209d77c41f920833a547dc455db5910d 100644 --- a/adapter/checkpoint-2/adapter_model.safetensors +++ b/adapter/checkpoint-2/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:43e7a048bea5c9b88eae829f5a6ea192d63a9a9ee3be127d8931054344774a0b +oid sha256:b5f28440ff3f267bc7a6e3529f9ba86a1da9431cea7748c7014fd80838bae088 size 239536776 diff --git a/adapter/checkpoint-2/optimizer.pt b/adapter/checkpoint-2/optimizer.pt index 914df82d45ded461ae784a61f1e9fd2c9f7f83c2..b449be2e1312173ee77c526b1857f4bd4bc01d57 100644 --- a/adapter/checkpoint-2/optimizer.pt +++ b/adapter/checkpoint-2/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b7b530d3378e9cbb1e3f4ffbcd10e7d058234260cffaeab63775acb44a4da560 +oid sha256:73cbb9ecaa22640c0fffda414474bb65a07691b0d054549c3c56b7d87a77c80b size 243818437 diff --git a/adapter/checkpoint-3/adapter_config.json b/adapter/checkpoint-3/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-3/adapter_config.json +++ b/adapter/checkpoint-3/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-3/adapter_model.safetensors b/adapter/checkpoint-3/adapter_model.safetensors index e855914feea142cee6308e63e925ac666d96f870..c5190358b4e7c224e4cd0f81a98fb6eb812134ef 100644 --- a/adapter/checkpoint-3/adapter_model.safetensors +++ b/adapter/checkpoint-3/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:63a2df1a70456844c689f40525f9c3ac54f76c0275316566a7b60d88668ba86d +oid sha256:dd6f8f22893645472442d2ac34a88966a12206dc4b8fe844e92a46c2d7841f14 size 239536776 diff --git a/adapter/checkpoint-3/optimizer.pt b/adapter/checkpoint-3/optimizer.pt index 6e8c94f8b03681f5b7f9962e28152e02a4b504d0..3ab187966e22e353a40acb3a40eba7f691a12345 100644 --- a/adapter/checkpoint-3/optimizer.pt +++ b/adapter/checkpoint-3/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:ea813c605e804352ba8ff66dd96e9f8683da294ca935813cb507520dff866fd7 +oid sha256:35edc07011d9a778a6f92cddb94e16b5cf91466abd22943cf379ef288ac6c4a5 size 243818437 diff --git a/adapter/checkpoint-4/adapter_config.json b/adapter/checkpoint-4/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-4/adapter_config.json +++ b/adapter/checkpoint-4/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-4/adapter_model.safetensors b/adapter/checkpoint-4/adapter_model.safetensors index c80c673bb4f6e042d5430db2350652a37252090e..8e7e692f1a8f16e2c105c51d54ef058ad5f5cffb 100644 --- a/adapter/checkpoint-4/adapter_model.safetensors +++ b/adapter/checkpoint-4/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:0b759651321e274d2fac4ede5eae09ec051c644ddc3da7cd711b4fee072ff7ab +oid sha256:07c9721da7ae65cbc73b276b059a18f46de2579d51182898a55b6a81b25a3b70 size 239536776 diff --git a/adapter/checkpoint-4/optimizer.pt b/adapter/checkpoint-4/optimizer.pt index 49d5f9ed7ebb2d2af90cf782c9ba97b429e28ab9..7bcf1079f594ef5b72e43511e34e7df7b9d5cc6f 100644 --- a/adapter/checkpoint-4/optimizer.pt +++ b/adapter/checkpoint-4/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:d70fbb46583d0982619b54aa0647e44b07513ed9d5bbe96c792d690c4cd82385 +oid sha256:7c087c0ee39c0a3dbf77220bb7a45e3e49d2f7ad1c07b92960d3903c21220d29 size 243818437 diff --git a/adapter/checkpoint-5/adapter_config.json b/adapter/checkpoint-5/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-5/adapter_config.json +++ b/adapter/checkpoint-5/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-5/adapter_model.safetensors b/adapter/checkpoint-5/adapter_model.safetensors index d41707fb248b149938043e3ab88bbc556ebe9ba6..60c1056a65de1204589afb7e3ef05363d8e7c6c2 100644 --- a/adapter/checkpoint-5/adapter_model.safetensors +++ b/adapter/checkpoint-5/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:29e56d014bad376378d3b34998bcb3b3b9fb51b51b3f5db99fc3198c9852018d +oid sha256:0cc0ebb8aeefb322a48df1be8b5363be8a5a8f982e41070c3061e67b753c6ddd size 239536776 diff --git a/adapter/checkpoint-5/optimizer.pt b/adapter/checkpoint-5/optimizer.pt index 75b68afe0460ed295b133d5316d84c5eca14dc8c..faf6589cb83d923f5aac545f07f8c501117b5d36 100644 --- a/adapter/checkpoint-5/optimizer.pt +++ b/adapter/checkpoint-5/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:c01e814587fc44eb23dc47e7d9794565fa2c4a0982fc0e8ba8e92e6933f5bf3e +oid sha256:dd38cc2cb9cc376bd47c7b1eeabc584bd125abb715b1ad3b67cf702869dfe5ad size 243818437 diff --git a/adapter/checkpoint-6/adapter_config.json b/adapter/checkpoint-6/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-6/adapter_config.json +++ b/adapter/checkpoint-6/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-6/adapter_model.safetensors b/adapter/checkpoint-6/adapter_model.safetensors index d2f65c2b6f544b00910b8e7358a5aa3407e58d40..d9e5e38eefa1a5504f5d3a700b08f00682b42efd 100644 --- a/adapter/checkpoint-6/adapter_model.safetensors +++ b/adapter/checkpoint-6/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:2d520e27667af63bf65784a8245076234ec27b98f7e084552481b2c0ac118a13 +oid sha256:ac2b4923da7aa08825a2ef8e36bd4e2eb438de607846aaa3edcf5eeddbce4031 size 239536776 diff --git a/adapter/checkpoint-6/optimizer.pt b/adapter/checkpoint-6/optimizer.pt index e1b910fe4cd7b2897404ccd9e70d1dfe106fe89d..255c8c0761ced779c0e61e20bf631c46a7bd120c 100644 --- a/adapter/checkpoint-6/optimizer.pt +++ b/adapter/checkpoint-6/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:2fa67e6c425f93c8b24ac0d40e17dd9d2e7902edc418a3ff7636da87ffc1da8f +oid sha256:06ad60be8ae0e218d0edace7e76e29afe97d9474c0108b82bd962d94e08ce99f size 243818437 diff --git a/adapter/checkpoint-7/adapter_config.json b/adapter/checkpoint-7/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-7/adapter_config.json +++ b/adapter/checkpoint-7/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-7/adapter_model.safetensors b/adapter/checkpoint-7/adapter_model.safetensors index 2657da0dee3a433805a41b486c2626d9539cfede..53294440b86ddd35b669c71e3338625abc31e54e 100644 --- a/adapter/checkpoint-7/adapter_model.safetensors +++ b/adapter/checkpoint-7/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:6c8b39faf3a2c769ccbcd0f4bf97596484347460e0b96cdc227c636c8f7e523d +oid sha256:ffc9e1a2fb5f52dba2d02e58e21b841caf69d5cc3db65765036878a332b29617 size 239536776 diff --git a/adapter/checkpoint-7/optimizer.pt b/adapter/checkpoint-7/optimizer.pt index 189422e47e67b55620fbb03ae95db5926ef1cceb..c77b4a85d2ed7c9c85cb5994f352c7b9b8dd7a64 100644 --- a/adapter/checkpoint-7/optimizer.pt +++ b/adapter/checkpoint-7/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:b94dbe7064b764155bb9e2b3d791bb4f4a5fa064cd4d2ffbbd4fbb5fdcd4b70b +oid sha256:618eb287cd639948b42b7172333b50379df12661b16c7ebd1f944d841f833f81 size 243818437 diff --git a/adapter/checkpoint-8/adapter_config.json b/adapter/checkpoint-8/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-8/adapter_config.json +++ b/adapter/checkpoint-8/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-8/adapter_model.safetensors b/adapter/checkpoint-8/adapter_model.safetensors index b7aab915274c1413ff4c2f16d13986057f3a438a..a9be6afb0109865486e1c8436b58e8094de1089e 100644 --- a/adapter/checkpoint-8/adapter_model.safetensors +++ b/adapter/checkpoint-8/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:00f650b9434b1e9710dbda8ecdf0913aabff56e76246b504a909c8bf1cfd5820 +oid sha256:20fe7d37644a3c5e3936894db23065925af1124fbea2f3deae0ac82efdb88916 size 239536776 diff --git a/adapter/checkpoint-8/optimizer.pt b/adapter/checkpoint-8/optimizer.pt index 2fdaab061a6b40bbbe1fab0d8c4d4b6fac5e01ab..6c041adfd34146fb0c310516bf3ae71a03d07b8e 100644 --- a/adapter/checkpoint-8/optimizer.pt +++ b/adapter/checkpoint-8/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:16011f2abf7e734e798897529fdb53e299bc3ae0e164decc6417556b83bc625f +oid sha256:befc8812f05ed59f776319cb845c37271e189ef00b30bccb0a62754bba7fa75e size 243818437 diff --git a/adapter/checkpoint-9/adapter_config.json b/adapter/checkpoint-9/adapter_config.json index e934c4f8a7385575225a5b73e1281e2a8fbdb3e3..5fd4fcbf494e56470debb41e03496c1c739a98d3 100644 --- a/adapter/checkpoint-9/adapter_config.json +++ b/adapter/checkpoint-9/adapter_config.json @@ -31,13 +31,13 @@ "rank_pattern": {}, "revision": null, "target_modules": [ - "o_proj", "up_proj", - "k_proj", "gate_proj", - "down_proj", + "o_proj", "v_proj", - "q_proj" + "down_proj", + "q_proj", + "k_proj" ], "target_parameters": null, "task_type": "CAUSAL_LM", @@ -45,6 +45,6 @@ "use_bdlora": null, "use_dora": false, "use_qalora": false, - "use_rslora": false, + "use_rslora": true, "velora_config": null } \ No newline at end of file diff --git a/adapter/checkpoint-9/adapter_model.safetensors b/adapter/checkpoint-9/adapter_model.safetensors index c00dbef482dc9b2a443074747d745a416c0c1b78..57f8612a37397658c85944e5880dc39a0e5835df 100644 --- a/adapter/checkpoint-9/adapter_model.safetensors +++ b/adapter/checkpoint-9/adapter_model.safetensors @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:034d5337cafb6b34232eb6a40723e4d87caba75cdc73b10f25fa97eefb8a99a5 +oid sha256:ba07341a810f242ae89c79e431175a1eb5ca4db3a3a5da5d13cc4c33410d1f76 size 239536776 diff --git a/adapter/checkpoint-9/optimizer.pt b/adapter/checkpoint-9/optimizer.pt index 386f8c313d409fbd535b165a747332a2488c67c7..eaa9790a2f3a0c8aa2d485c35d7d5e119f69ae01 100644 --- a/adapter/checkpoint-9/optimizer.pt +++ b/adapter/checkpoint-9/optimizer.pt @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:7567a4fd6835febeb5724bd7394d79ac797cefd8b7d72ab41dc830ecf9b2e5e5 +oid sha256:6613889d1a4684553587028e623b6f9563264218a3e97dfa8aba72f391ba3040 size 243818437