diff --git a/.gitattributes b/.gitattributes index bfc0d36f94db218b914dee14f41b1ff98804ba9b..7e835791415579b7b239b1229b51e94ade03cfc3 100644 --- a/.gitattributes +++ b/.gitattributes @@ -103,3 +103,9 @@ gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json filter=lfs gsm8k__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json filter=lfs diff=lfs merge=lfs -text csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json filter=lfs diff=lfs merge=lfs -text csqa__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json filter=lfs diff=lfs merge=lfs -text +csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json filter=lfs diff=lfs merge=lfs -text +csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json filter=lfs diff=lfs merge=lfs -text +csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json filter=lfs diff=lfs merge=lfs -text +csqa__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json filter=lfs diff=lfs merge=lfs -text +csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json filter=lfs diff=lfs merge=lfs -text +csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json filter=lfs diff=lfs merge=lfs -text diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth new file mode 100644 index 0000000000000000000000000000000000000000..eed508682149de4c1131d3222faaf97d0f99af6c --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0d51a89f706019e82a1bece244691bbef31c0dc9544212f828d427b8052b077 +size 16389 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth new file mode 100644 index 0000000000000000000000000000000000000000..1582f870d588010e54a84e87cad2a82c64383531 --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9db2a9d890a3bdfb6477301adcccedc994df442a5cf1f4f617b4737086e632ef +size 16389 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth new file mode 100644 index 0000000000000000000000000000000000000000..6c5f68874a2cbb1ee13b8701913fee25a565465d --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74f99367832c29a5aff6042c0dc14d87a3ffc16ce9cc6466678227421d213c43 +size 16389 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth new file mode 100644 index 0000000000000000000000000000000000000000..7cc2af470bce75db7a93a2eff6e9cf7abc7f29c0 --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2984cd1dfb7f3fdddf5cc99bb1dab2759a14eef425d83154d76b310151f5d01c +size 16389 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth new file mode 100644 index 0000000000000000000000000000000000000000..fe9ff85525efda6057785fc346a0b95c5c320269 --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4cb1eac572046e66de7e9c52bba4d61d758728542e4fdc6d8b194a6c29e10f96 +size 16389 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth new file mode 100644 index 0000000000000000000000000000000000000000..aafad03dd3947f1d8bbd9c5f74914185c6db95cf --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41da76e42d56dcd9ee4821dcdf66fd5f2e82e821081ec3e07cb7e1e952647992 +size 16389 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_6.pth b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_6.pth new file mode 100644 index 0000000000000000000000000000000000000000..f60d3ad5cafb5f38dd5f7dd7d21e6d1e23434c32 --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_6.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1ef332606e64d74e7d5d444ae550c6edf2fe161722c5d652fd47845e2aa1894 +size 16389 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth new file mode 100644 index 0000000000000000000000000000000000000000..3483165240faba63dddcaf2eb4aab33e8f3423d3 --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4b125a8a8f1fcea8543fd480628a947fbdff2ec9452d6a7be7e77a426d76b0c +size 16389 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..2e1fbe4ffe357b565f5ae2924b2767942728bba2 --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70583c9a51d736f1353b66ad07c780d71300daef1dfeacad27c4a2efb4fd9d68 +size 1465 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..172311123ab62378f1f6d90f3068a676b7d939ed --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c1dcab308e7cf5970ea38815e0a62887d705c5b436f869ca27a5dcdd40c36a6 +size 17210148 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..f5dcc46a53ada8fc289cd8cf3abc6f5d7195ef1d --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdc1b3bc49e1e979a001baf2ca82a5f042bdc27bd295bc8237d9f5c39026b8fd +size 6225 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..172311123ab62378f1f6d90f3068a676b7d939ed --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c1dcab308e7cf5970ea38815e0a62887d705c5b436f869ca27a5dcdd40c36a6 +size 17210148 diff --git a/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/training_args.bin b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..f5dcc46a53ada8fc289cd8cf3abc6f5d7195ef1d --- /dev/null +++ b/csqa__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fdc1b3bc49e1e979a001baf2ca82a5f042bdc27bd295bc8237d9f5c39026b8fd +size 6225 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth new file mode 100644 index 0000000000000000000000000000000000000000..eed508682149de4c1131d3222faaf97d0f99af6c --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0d51a89f706019e82a1bece244691bbef31c0dc9544212f828d427b8052b077 +size 16389 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth new file mode 100644 index 0000000000000000000000000000000000000000..1582f870d588010e54a84e87cad2a82c64383531 --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9db2a9d890a3bdfb6477301adcccedc994df442a5cf1f4f617b4737086e632ef +size 16389 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth new file mode 100644 index 0000000000000000000000000000000000000000..6c5f68874a2cbb1ee13b8701913fee25a565465d --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74f99367832c29a5aff6042c0dc14d87a3ffc16ce9cc6466678227421d213c43 +size 16389 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth new file mode 100644 index 0000000000000000000000000000000000000000..7cc2af470bce75db7a93a2eff6e9cf7abc7f29c0 --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2984cd1dfb7f3fdddf5cc99bb1dab2759a14eef425d83154d76b310151f5d01c +size 16389 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth new file mode 100644 index 0000000000000000000000000000000000000000..fe9ff85525efda6057785fc346a0b95c5c320269 --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4cb1eac572046e66de7e9c52bba4d61d758728542e4fdc6d8b194a6c29e10f96 +size 16389 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth new file mode 100644 index 0000000000000000000000000000000000000000..aafad03dd3947f1d8bbd9c5f74914185c6db95cf --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41da76e42d56dcd9ee4821dcdf66fd5f2e82e821081ec3e07cb7e1e952647992 +size 16389 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_6.pth b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_6.pth new file mode 100644 index 0000000000000000000000000000000000000000..f60d3ad5cafb5f38dd5f7dd7d21e6d1e23434c32 --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_6.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1ef332606e64d74e7d5d444ae550c6edf2fe161722c5d652fd47845e2aa1894 +size 16389 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth new file mode 100644 index 0000000000000000000000000000000000000000..3483165240faba63dddcaf2eb4aab33e8f3423d3 --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4b125a8a8f1fcea8543fd480628a947fbdff2ec9452d6a7be7e77a426d76b0c +size 16389 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..2e1fbe4ffe357b565f5ae2924b2767942728bba2 --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70583c9a51d736f1353b66ad07c780d71300daef1dfeacad27c4a2efb4fd9d68 +size 1465 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..172311123ab62378f1f6d90f3068a676b7d939ed --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c1dcab308e7cf5970ea38815e0a62887d705c5b436f869ca27a5dcdd40c36a6 +size 17210148 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..979e07588be25051288cfea3f3b12c9f8b3d9686 --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c7765c5b78f375dcc02c499a3caae283f561d92654cc152d633b21dc10b0b9e1 +size 6225 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..172311123ab62378f1f6d90f3068a676b7d939ed --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8c1dcab308e7cf5970ea38815e0a62887d705c5b436f869ca27a5dcdd40c36a6 +size 17210148 diff --git a/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/training_args.bin b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..979e07588be25051288cfea3f3b12c9f8b3d9686 --- /dev/null +++ b/csqa__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c7765c5b78f375dcc02c499a3caae283f561d92654cc152d633b21dc10b0b9e1 +size 6225 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth new file mode 100644 index 0000000000000000000000000000000000000000..eed508682149de4c1131d3222faaf97d0f99af6c --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0d51a89f706019e82a1bece244691bbef31c0dc9544212f828d427b8052b077 +size 16389 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth new file mode 100644 index 0000000000000000000000000000000000000000..1582f870d588010e54a84e87cad2a82c64383531 --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9db2a9d890a3bdfb6477301adcccedc994df442a5cf1f4f617b4737086e632ef +size 16389 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth new file mode 100644 index 0000000000000000000000000000000000000000..6c5f68874a2cbb1ee13b8701913fee25a565465d --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74f99367832c29a5aff6042c0dc14d87a3ffc16ce9cc6466678227421d213c43 +size 16389 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth new file mode 100644 index 0000000000000000000000000000000000000000..7cc2af470bce75db7a93a2eff6e9cf7abc7f29c0 --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2984cd1dfb7f3fdddf5cc99bb1dab2759a14eef425d83154d76b310151f5d01c +size 16389 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth new file mode 100644 index 0000000000000000000000000000000000000000..fe9ff85525efda6057785fc346a0b95c5c320269 --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4cb1eac572046e66de7e9c52bba4d61d758728542e4fdc6d8b194a6c29e10f96 +size 16389 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth new file mode 100644 index 0000000000000000000000000000000000000000..aafad03dd3947f1d8bbd9c5f74914185c6db95cf --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41da76e42d56dcd9ee4821dcdf66fd5f2e82e821081ec3e07cb7e1e952647992 +size 16389 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_6.pth b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_6.pth new file mode 100644 index 0000000000000000000000000000000000000000..f60d3ad5cafb5f38dd5f7dd7d21e6d1e23434c32 --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_6.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1ef332606e64d74e7d5d444ae550c6edf2fe161722c5d652fd47845e2aa1894 +size 16389 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth new file mode 100644 index 0000000000000000000000000000000000000000..3483165240faba63dddcaf2eb4aab33e8f3423d3 --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4b125a8a8f1fcea8543fd480628a947fbdff2ec9452d6a7be7e77a426d76b0c +size 16389 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..2e1fbe4ffe357b565f5ae2924b2767942728bba2 --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70583c9a51d736f1353b66ad07c780d71300daef1dfeacad27c4a2efb4fd9d68 +size 1465 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..cd71f61a15a522601badb3dc960d800d9cb3766c --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4 +size 11422654 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..362fdbbedd56812ded7e934d8b7f3341f6021c41 --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:18faf0b870b38c912c75e1ce9afdbe19af1c3fadd92209c7a90c663c6da06db4 +size 6161 diff --git a/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/training_args.bin b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..362fdbbedd56812ded7e934d8b7f3341f6021c41 --- /dev/null +++ b/csqa__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:18faf0b870b38c912c75e1ce9afdbe19af1c3fadd92209c7a90c663c6da06db4 +size 6161 diff --git a/csqa__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json b/csqa__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json new file mode 100644 index 0000000000000000000000000000000000000000..cd71f61a15a522601badb3dc960d800d9cb3766c --- /dev/null +++ b/csqa__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/tokenizer.json @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4 +size 11422654 diff --git a/csqa__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/training_args.bin b/csqa__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..876b610ff40e0a52fdab2c316e07b7191fc22660 --- /dev/null +++ b/csqa__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8659b993a9025fced8eda027b3eb2c8012015107d38086de64143bfb5704ef31 +size 6161 diff --git a/gsm8k__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth b/gsm8k__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth new file mode 100644 index 0000000000000000000000000000000000000000..fe9ff85525efda6057785fc346a0b95c5c320269 --- /dev/null +++ b/gsm8k__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4cb1eac572046e66de7e9c52bba4d61d758728542e4fdc6d8b194a6c29e10f96 +size 16389 diff --git a/gsm8k__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin b/gsm8k__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..fa8cd069ab8ad6fb38cfd4ff7784d804a4255e40 --- /dev/null +++ b/gsm8k__llama3_8b_instruct__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d818ffd17d6503309f5c3b497030d0e3a41785b1e06d81ccae31f09c44a5ba6 +size 6225 diff --git a/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth new file mode 100644 index 0000000000000000000000000000000000000000..eed508682149de4c1131d3222faaf97d0f99af6c --- /dev/null +++ b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0d51a89f706019e82a1bece244691bbef31c0dc9544212f828d427b8052b077 +size 16389 diff --git a/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth new file mode 100644 index 0000000000000000000000000000000000000000..1582f870d588010e54a84e87cad2a82c64383531 --- /dev/null +++ b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9db2a9d890a3bdfb6477301adcccedc994df442a5cf1f4f617b4737086e632ef +size 16389 diff --git a/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth new file mode 100644 index 0000000000000000000000000000000000000000..6c5f68874a2cbb1ee13b8701913fee25a565465d --- /dev/null +++ b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74f99367832c29a5aff6042c0dc14d87a3ffc16ce9cc6466678227421d213c43 +size 16389 diff --git a/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth new file mode 100644 index 0000000000000000000000000000000000000000..7cc2af470bce75db7a93a2eff6e9cf7abc7f29c0 --- /dev/null +++ b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_3.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2984cd1dfb7f3fdddf5cc99bb1dab2759a14eef425d83154d76b310151f5d01c +size 16389 diff --git a/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth new file mode 100644 index 0000000000000000000000000000000000000000..fe9ff85525efda6057785fc346a0b95c5c320269 --- /dev/null +++ b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_4.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4cb1eac572046e66de7e9c52bba4d61d758728542e4fdc6d8b194a6c29e10f96 +size 16389 diff --git a/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth new file mode 100644 index 0000000000000000000000000000000000000000..aafad03dd3947f1d8bbd9c5f74914185c6db95cf --- /dev/null +++ b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:41da76e42d56dcd9ee4821dcdf66fd5f2e82e821081ec3e07cb7e1e952647992 +size 16389 diff --git a/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth new file mode 100644 index 0000000000000000000000000000000000000000..3483165240faba63dddcaf2eb4aab33e8f3423d3 --- /dev/null +++ b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4b125a8a8f1fcea8543fd480628a947fbdff2ec9452d6a7be7e77a426d76b0c +size 16389 diff --git a/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..2e1fbe4ffe357b565f5ae2924b2767942728bba2 --- /dev/null +++ b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70583c9a51d736f1353b66ad07c780d71300daef1dfeacad27c4a2efb4fd9d68 +size 1465 diff --git a/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..1e8fd05b0f448a5a0c1c8521c7e9de85bc3da7f7 --- /dev/null +++ b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f7a9c520160f72c51086fd83acfbbc629ce8de2e2586a1935813c9cf777e0dd +size 6225 diff --git a/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/training_args.bin b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..1e8fd05b0f448a5a0c1c8521c7e9de85bc3da7f7 --- /dev/null +++ b/gsm8k__llama3_8b_instruct__bare_100_r8_bs4_lr1e5_e3/lora/sft/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8f7a9c520160f72c51086fd83acfbbc629ce8de2e2586a1935813c9cf777e0dd +size 6225 diff --git a/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth b/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth new file mode 100644 index 0000000000000000000000000000000000000000..eed508682149de4c1131d3222faaf97d0f99af6c --- /dev/null +++ b/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_0.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0d51a89f706019e82a1bece244691bbef31c0dc9544212f828d427b8052b077 +size 16389 diff --git a/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth b/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth new file mode 100644 index 0000000000000000000000000000000000000000..1582f870d588010e54a84e87cad2a82c64383531 --- /dev/null +++ b/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_1.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9db2a9d890a3bdfb6477301adcccedc994df442a5cf1f4f617b4737086e632ef +size 16389 diff --git a/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth b/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth new file mode 100644 index 0000000000000000000000000000000000000000..6c5f68874a2cbb1ee13b8701913fee25a565465d --- /dev/null +++ b/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/rng_state_2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:74f99367832c29a5aff6042c0dc14d87a3ffc16ce9cc6466678227421d213c43 +size 16389 diff --git a/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/training_args.bin b/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..9a46576a4b228f736d7dccd8edf651fc2fc523fa --- /dev/null +++ b/gsm8k__qwen3_8b__all_variants_r8_bs4_lr1e5_e3/lora/sft/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ffcf5a84e896b05a24f66db8d60e9d2ccd852ff5b446838068d188c50b29612 +size 6161 diff --git a/gsm8k__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/optimizer.pt b/gsm8k__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..0b556035138d846beebdf8427e3f165a16e9beb8 --- /dev/null +++ b/gsm8k__qwen3_8b__bare_100_r8_bs4_lr1e5_e3/lora/sft/checkpoint-939/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5fd717ffb91dcc45503e07b80b274e30a0a2cf4845f3f6ec54a19e89944a6277 +size 175012683