Commit
•
30ab44e
1
Parent(s):
a244941
Upload LlamaForCausalLM
Browse files- model-00001-of-00003.safetensors +2 -2
- model-00002-of-00003.safetensors +2 -2
- model-00003-of-00003.safetensors +2 -2
- model.safetensors.index.json +452 -452
model-00001-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a39ce343bee15ce2b9ab9bdbb0e1131ba169f5e3cbacaf3f90011448bd73e717
|
| 3 |
+
size 6995733923
|
model-00002-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:76fa77cd165a1fbf2512f87622d211c853bf5c4531b52d2d147c3cbeed9fe67d
|
| 3 |
+
size 6934102407
|
model-00003-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f5991f2a8cb1f83db70ad7e2a11f8766caebf541c881cc144536e3cd74517f1a
|
| 3 |
+
size 5174173555
|
model.safetensors.index.json
CHANGED
|
@@ -621,38 +621,38 @@
|
|
| 621 |
"model.layers.20.self_attn.v_proj.weight.nested_quant_map": "model-00001-of-00003.safetensors",
|
| 622 |
"model.layers.20.self_attn.v_proj.weight.quant_map": "model-00001-of-00003.safetensors",
|
| 623 |
"model.layers.20.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00001-of-00003.safetensors",
|
| 624 |
-
"model.layers.21.input_layernorm.weight": "model-
|
| 625 |
-
"model.layers.21.mlp.down_proj.weight": "model-
|
| 626 |
-
"model.layers.21.mlp.down_proj.weight.absmax": "model-
|
| 627 |
-
"model.layers.21.mlp.down_proj.weight.nested_absmax": "model-
|
| 628 |
-
"model.layers.21.mlp.down_proj.weight.nested_quant_map": "model-
|
| 629 |
-
"model.layers.21.mlp.down_proj.weight.quant_map": "model-
|
| 630 |
-
"model.layers.21.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 631 |
-
"model.layers.21.mlp.gate_proj.weight": "model-
|
| 632 |
-
"model.layers.21.mlp.gate_proj.weight.absmax": "model-
|
| 633 |
-
"model.layers.21.mlp.gate_proj.weight.nested_absmax": "model-
|
| 634 |
-
"model.layers.21.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 635 |
-
"model.layers.21.mlp.gate_proj.weight.quant_map": "model-
|
| 636 |
-
"model.layers.21.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 637 |
-
"model.layers.21.mlp.up_proj.weight": "model-
|
| 638 |
-
"model.layers.21.mlp.up_proj.weight.absmax": "model-
|
| 639 |
-
"model.layers.21.mlp.up_proj.weight.nested_absmax": "model-
|
| 640 |
-
"model.layers.21.mlp.up_proj.weight.nested_quant_map": "model-
|
| 641 |
-
"model.layers.21.mlp.up_proj.weight.quant_map": "model-
|
| 642 |
-
"model.layers.21.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 643 |
-
"model.layers.21.post_attention_layernorm.weight": "model-
|
| 644 |
"model.layers.21.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
|
| 645 |
"model.layers.21.self_attn.k_proj.weight.absmax": "model-00001-of-00003.safetensors",
|
| 646 |
"model.layers.21.self_attn.k_proj.weight.nested_absmax": "model-00001-of-00003.safetensors",
|
| 647 |
"model.layers.21.self_attn.k_proj.weight.nested_quant_map": "model-00001-of-00003.safetensors",
|
| 648 |
"model.layers.21.self_attn.k_proj.weight.quant_map": "model-00001-of-00003.safetensors",
|
| 649 |
"model.layers.21.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00001-of-00003.safetensors",
|
| 650 |
-
"model.layers.21.self_attn.o_proj.weight": "model-
|
| 651 |
-
"model.layers.21.self_attn.o_proj.weight.absmax": "model-
|
| 652 |
-
"model.layers.21.self_attn.o_proj.weight.nested_absmax": "model-
|
| 653 |
-
"model.layers.21.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 654 |
-
"model.layers.21.self_attn.o_proj.weight.quant_map": "model-
|
| 655 |
-
"model.layers.21.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 656 |
"model.layers.21.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
|
| 657 |
"model.layers.21.self_attn.q_proj.weight.absmax": "model-00001-of-00003.safetensors",
|
| 658 |
"model.layers.21.self_attn.q_proj.weight.nested_absmax": "model-00001-of-00003.safetensors",
|
|
@@ -665,94 +665,94 @@
|
|
| 665 |
"model.layers.21.self_attn.v_proj.weight.nested_quant_map": "model-00001-of-00003.safetensors",
|
| 666 |
"model.layers.21.self_attn.v_proj.weight.quant_map": "model-00001-of-00003.safetensors",
|
| 667 |
"model.layers.21.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00001-of-00003.safetensors",
|
| 668 |
-
"model.layers.22.input_layernorm.weight": "model-
|
| 669 |
-
"model.layers.22.mlp.down_proj.weight": "model-
|
| 670 |
-
"model.layers.22.mlp.down_proj.weight.absmax": "model-
|
| 671 |
-
"model.layers.22.mlp.down_proj.weight.nested_absmax": "model-
|
| 672 |
-
"model.layers.22.mlp.down_proj.weight.nested_quant_map": "model-
|
| 673 |
-
"model.layers.22.mlp.down_proj.weight.quant_map": "model-
|
| 674 |
-
"model.layers.22.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 675 |
-
"model.layers.22.mlp.gate_proj.weight": "model-
|
| 676 |
-
"model.layers.22.mlp.gate_proj.weight.absmax": "model-
|
| 677 |
-
"model.layers.22.mlp.gate_proj.weight.nested_absmax": "model-
|
| 678 |
-
"model.layers.22.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 679 |
-
"model.layers.22.mlp.gate_proj.weight.quant_map": "model-
|
| 680 |
-
"model.layers.22.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 681 |
-
"model.layers.22.mlp.up_proj.weight": "model-
|
| 682 |
-
"model.layers.22.mlp.up_proj.weight.absmax": "model-
|
| 683 |
-
"model.layers.22.mlp.up_proj.weight.nested_absmax": "model-
|
| 684 |
-
"model.layers.22.mlp.up_proj.weight.nested_quant_map": "model-
|
| 685 |
-
"model.layers.22.mlp.up_proj.weight.quant_map": "model-
|
| 686 |
-
"model.layers.22.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 687 |
-
"model.layers.22.post_attention_layernorm.weight": "model-
|
| 688 |
-
"model.layers.22.self_attn.k_proj.weight": "model-
|
| 689 |
-
"model.layers.22.self_attn.k_proj.weight.absmax": "model-
|
| 690 |
-
"model.layers.22.self_attn.k_proj.weight.nested_absmax": "model-
|
| 691 |
-
"model.layers.22.self_attn.k_proj.weight.nested_quant_map": "model-
|
| 692 |
-
"model.layers.22.self_attn.k_proj.weight.quant_map": "model-
|
| 693 |
-
"model.layers.22.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 694 |
-
"model.layers.22.self_attn.o_proj.weight": "model-
|
| 695 |
-
"model.layers.22.self_attn.o_proj.weight.absmax": "model-
|
| 696 |
-
"model.layers.22.self_attn.o_proj.weight.nested_absmax": "model-
|
| 697 |
-
"model.layers.22.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 698 |
-
"model.layers.22.self_attn.o_proj.weight.quant_map": "model-
|
| 699 |
-
"model.layers.22.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 700 |
-
"model.layers.22.self_attn.q_proj.weight": "model-
|
| 701 |
-
"model.layers.22.self_attn.q_proj.weight.absmax": "model-
|
| 702 |
-
"model.layers.22.self_attn.q_proj.weight.nested_absmax": "model-
|
| 703 |
-
"model.layers.22.self_attn.q_proj.weight.nested_quant_map": "model-
|
| 704 |
-
"model.layers.22.self_attn.q_proj.weight.quant_map": "model-
|
| 705 |
-
"model.layers.22.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 706 |
-
"model.layers.22.self_attn.v_proj.weight": "model-
|
| 707 |
-
"model.layers.22.self_attn.v_proj.weight.absmax": "model-
|
| 708 |
-
"model.layers.22.self_attn.v_proj.weight.nested_absmax": "model-
|
| 709 |
-
"model.layers.22.self_attn.v_proj.weight.nested_quant_map": "model-
|
| 710 |
-
"model.layers.22.self_attn.v_proj.weight.quant_map": "model-
|
| 711 |
-
"model.layers.22.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 712 |
-
"model.layers.23.input_layernorm.weight": "model-
|
| 713 |
-
"model.layers.23.mlp.down_proj.weight": "model-
|
| 714 |
-
"model.layers.23.mlp.down_proj.weight.absmax": "model-
|
| 715 |
-
"model.layers.23.mlp.down_proj.weight.nested_absmax": "model-
|
| 716 |
-
"model.layers.23.mlp.down_proj.weight.nested_quant_map": "model-
|
| 717 |
-
"model.layers.23.mlp.down_proj.weight.quant_map": "model-
|
| 718 |
-
"model.layers.23.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 719 |
-
"model.layers.23.mlp.gate_proj.weight": "model-
|
| 720 |
-
"model.layers.23.mlp.gate_proj.weight.absmax": "model-
|
| 721 |
-
"model.layers.23.mlp.gate_proj.weight.nested_absmax": "model-
|
| 722 |
-
"model.layers.23.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 723 |
-
"model.layers.23.mlp.gate_proj.weight.quant_map": "model-
|
| 724 |
-
"model.layers.23.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 725 |
-
"model.layers.23.mlp.up_proj.weight": "model-
|
| 726 |
-
"model.layers.23.mlp.up_proj.weight.absmax": "model-
|
| 727 |
-
"model.layers.23.mlp.up_proj.weight.nested_absmax": "model-
|
| 728 |
-
"model.layers.23.mlp.up_proj.weight.nested_quant_map": "model-
|
| 729 |
-
"model.layers.23.mlp.up_proj.weight.quant_map": "model-
|
| 730 |
-
"model.layers.23.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 731 |
-
"model.layers.23.post_attention_layernorm.weight": "model-
|
| 732 |
-
"model.layers.23.self_attn.k_proj.weight": "model-
|
| 733 |
-
"model.layers.23.self_attn.k_proj.weight.absmax": "model-
|
| 734 |
-
"model.layers.23.self_attn.k_proj.weight.nested_absmax": "model-
|
| 735 |
-
"model.layers.23.self_attn.k_proj.weight.nested_quant_map": "model-
|
| 736 |
-
"model.layers.23.self_attn.k_proj.weight.quant_map": "model-
|
| 737 |
-
"model.layers.23.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 738 |
-
"model.layers.23.self_attn.o_proj.weight": "model-
|
| 739 |
-
"model.layers.23.self_attn.o_proj.weight.absmax": "model-
|
| 740 |
-
"model.layers.23.self_attn.o_proj.weight.nested_absmax": "model-
|
| 741 |
-
"model.layers.23.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 742 |
-
"model.layers.23.self_attn.o_proj.weight.quant_map": "model-
|
| 743 |
-
"model.layers.23.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 744 |
-
"model.layers.23.self_attn.q_proj.weight": "model-
|
| 745 |
-
"model.layers.23.self_attn.q_proj.weight.absmax": "model-
|
| 746 |
-
"model.layers.23.self_attn.q_proj.weight.nested_absmax": "model-
|
| 747 |
-
"model.layers.23.self_attn.q_proj.weight.nested_quant_map": "model-
|
| 748 |
-
"model.layers.23.self_attn.q_proj.weight.quant_map": "model-
|
| 749 |
-
"model.layers.23.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 750 |
-
"model.layers.23.self_attn.v_proj.weight": "model-
|
| 751 |
-
"model.layers.23.self_attn.v_proj.weight.absmax": "model-
|
| 752 |
-
"model.layers.23.self_attn.v_proj.weight.nested_absmax": "model-
|
| 753 |
-
"model.layers.23.self_attn.v_proj.weight.nested_quant_map": "model-
|
| 754 |
-
"model.layers.23.self_attn.v_proj.weight.quant_map": "model-
|
| 755 |
-
"model.layers.23.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 756 |
"model.layers.24.input_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 757 |
"model.layers.24.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
|
| 758 |
"model.layers.24.mlp.down_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
|
@@ -760,12 +760,12 @@
|
|
| 760 |
"model.layers.24.mlp.down_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 761 |
"model.layers.24.mlp.down_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 762 |
"model.layers.24.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 763 |
-
"model.layers.24.mlp.gate_proj.weight": "model-
|
| 764 |
-
"model.layers.24.mlp.gate_proj.weight.absmax": "model-
|
| 765 |
-
"model.layers.24.mlp.gate_proj.weight.nested_absmax": "model-
|
| 766 |
-
"model.layers.24.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 767 |
-
"model.layers.24.mlp.gate_proj.weight.quant_map": "model-
|
| 768 |
-
"model.layers.24.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 769 |
"model.layers.24.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
|
| 770 |
"model.layers.24.mlp.up_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 771 |
"model.layers.24.mlp.up_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
|
@@ -773,30 +773,30 @@
|
|
| 773 |
"model.layers.24.mlp.up_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 774 |
"model.layers.24.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 775 |
"model.layers.24.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 776 |
-
"model.layers.24.self_attn.k_proj.weight": "model-
|
| 777 |
-
"model.layers.24.self_attn.k_proj.weight.absmax": "model-
|
| 778 |
-
"model.layers.24.self_attn.k_proj.weight.nested_absmax": "model-
|
| 779 |
-
"model.layers.24.self_attn.k_proj.weight.nested_quant_map": "model-
|
| 780 |
-
"model.layers.24.self_attn.k_proj.weight.quant_map": "model-
|
| 781 |
-
"model.layers.24.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 782 |
-
"model.layers.24.self_attn.o_proj.weight": "model-
|
| 783 |
-
"model.layers.24.self_attn.o_proj.weight.absmax": "model-
|
| 784 |
-
"model.layers.24.self_attn.o_proj.weight.nested_absmax": "model-
|
| 785 |
-
"model.layers.24.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 786 |
-
"model.layers.24.self_attn.o_proj.weight.quant_map": "model-
|
| 787 |
-
"model.layers.24.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 788 |
-
"model.layers.24.self_attn.q_proj.weight": "model-
|
| 789 |
-
"model.layers.24.self_attn.q_proj.weight.absmax": "model-
|
| 790 |
-
"model.layers.24.self_attn.q_proj.weight.nested_absmax": "model-
|
| 791 |
-
"model.layers.24.self_attn.q_proj.weight.nested_quant_map": "model-
|
| 792 |
-
"model.layers.24.self_attn.q_proj.weight.quant_map": "model-
|
| 793 |
-
"model.layers.24.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 794 |
-
"model.layers.24.self_attn.v_proj.weight": "model-
|
| 795 |
-
"model.layers.24.self_attn.v_proj.weight.absmax": "model-
|
| 796 |
-
"model.layers.24.self_attn.v_proj.weight.nested_absmax": "model-
|
| 797 |
-
"model.layers.24.self_attn.v_proj.weight.nested_quant_map": "model-
|
| 798 |
-
"model.layers.24.self_attn.v_proj.weight.quant_map": "model-
|
| 799 |
-
"model.layers.24.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 800 |
"model.layers.25.input_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 801 |
"model.layers.25.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
|
| 802 |
"model.layers.25.mlp.down_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
|
@@ -1765,26 +1765,26 @@
|
|
| 1765 |
"model.layers.44.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 1766 |
"model.layers.44.self_attn.v_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 1767 |
"model.layers.44.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 1768 |
-
"model.layers.45.input_layernorm.weight": "model-
|
| 1769 |
-
"model.layers.45.mlp.down_proj.weight": "model-
|
| 1770 |
-
"model.layers.45.mlp.down_proj.weight.absmax": "model-
|
| 1771 |
-
"model.layers.45.mlp.down_proj.weight.nested_absmax": "model-
|
| 1772 |
-
"model.layers.45.mlp.down_proj.weight.nested_quant_map": "model-
|
| 1773 |
-
"model.layers.45.mlp.down_proj.weight.quant_map": "model-
|
| 1774 |
-
"model.layers.45.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1775 |
-
"model.layers.45.mlp.gate_proj.weight": "model-
|
| 1776 |
-
"model.layers.45.mlp.gate_proj.weight.absmax": "model-
|
| 1777 |
-
"model.layers.45.mlp.gate_proj.weight.nested_absmax": "model-
|
| 1778 |
-
"model.layers.45.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 1779 |
-
"model.layers.45.mlp.gate_proj.weight.quant_map": "model-
|
| 1780 |
-
"model.layers.45.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1781 |
-
"model.layers.45.mlp.up_proj.weight": "model-
|
| 1782 |
-
"model.layers.45.mlp.up_proj.weight.absmax": "model-
|
| 1783 |
-
"model.layers.45.mlp.up_proj.weight.nested_absmax": "model-
|
| 1784 |
-
"model.layers.45.mlp.up_proj.weight.nested_quant_map": "model-
|
| 1785 |
-
"model.layers.45.mlp.up_proj.weight.quant_map": "model-
|
| 1786 |
-
"model.layers.45.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1787 |
-
"model.layers.45.post_attention_layernorm.weight": "model-
|
| 1788 |
"model.layers.45.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
|
| 1789 |
"model.layers.45.self_attn.k_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 1790 |
"model.layers.45.self_attn.k_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
|
@@ -1809,182 +1809,182 @@
|
|
| 1809 |
"model.layers.45.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 1810 |
"model.layers.45.self_attn.v_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 1811 |
"model.layers.45.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 1812 |
-
"model.layers.46.input_layernorm.weight": "model-
|
| 1813 |
-
"model.layers.46.mlp.down_proj.weight": "model-
|
| 1814 |
-
"model.layers.46.mlp.down_proj.weight.absmax": "model-
|
| 1815 |
-
"model.layers.46.mlp.down_proj.weight.nested_absmax": "model-
|
| 1816 |
-
"model.layers.46.mlp.down_proj.weight.nested_quant_map": "model-
|
| 1817 |
-
"model.layers.46.mlp.down_proj.weight.quant_map": "model-
|
| 1818 |
-
"model.layers.46.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1819 |
-
"model.layers.46.mlp.gate_proj.weight": "model-
|
| 1820 |
-
"model.layers.46.mlp.gate_proj.weight.absmax": "model-
|
| 1821 |
-
"model.layers.46.mlp.gate_proj.weight.nested_absmax": "model-
|
| 1822 |
-
"model.layers.46.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 1823 |
-
"model.layers.46.mlp.gate_proj.weight.quant_map": "model-
|
| 1824 |
-
"model.layers.46.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1825 |
-
"model.layers.46.mlp.up_proj.weight": "model-
|
| 1826 |
-
"model.layers.46.mlp.up_proj.weight.absmax": "model-
|
| 1827 |
-
"model.layers.46.mlp.up_proj.weight.nested_absmax": "model-
|
| 1828 |
-
"model.layers.46.mlp.up_proj.weight.nested_quant_map": "model-
|
| 1829 |
-
"model.layers.46.mlp.up_proj.weight.quant_map": "model-
|
| 1830 |
-
"model.layers.46.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1831 |
-
"model.layers.46.post_attention_layernorm.weight": "model-
|
| 1832 |
-
"model.layers.46.self_attn.k_proj.weight": "model-
|
| 1833 |
-
"model.layers.46.self_attn.k_proj.weight.absmax": "model-
|
| 1834 |
-
"model.layers.46.self_attn.k_proj.weight.nested_absmax": "model-
|
| 1835 |
-
"model.layers.46.self_attn.k_proj.weight.nested_quant_map": "model-
|
| 1836 |
-
"model.layers.46.self_attn.k_proj.weight.quant_map": "model-
|
| 1837 |
-
"model.layers.46.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1838 |
-
"model.layers.46.self_attn.o_proj.weight": "model-
|
| 1839 |
-
"model.layers.46.self_attn.o_proj.weight.absmax": "model-
|
| 1840 |
-
"model.layers.46.self_attn.o_proj.weight.nested_absmax": "model-
|
| 1841 |
-
"model.layers.46.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 1842 |
-
"model.layers.46.self_attn.o_proj.weight.quant_map": "model-
|
| 1843 |
-
"model.layers.46.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1844 |
-
"model.layers.46.self_attn.q_proj.weight": "model-
|
| 1845 |
-
"model.layers.46.self_attn.q_proj.weight.absmax": "model-
|
| 1846 |
-
"model.layers.46.self_attn.q_proj.weight.nested_absmax": "model-
|
| 1847 |
-
"model.layers.46.self_attn.q_proj.weight.nested_quant_map": "model-
|
| 1848 |
-
"model.layers.46.self_attn.q_proj.weight.quant_map": "model-
|
| 1849 |
-
"model.layers.46.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1850 |
-
"model.layers.46.self_attn.v_proj.weight": "model-
|
| 1851 |
-
"model.layers.46.self_attn.v_proj.weight.absmax": "model-
|
| 1852 |
-
"model.layers.46.self_attn.v_proj.weight.nested_absmax": "model-
|
| 1853 |
-
"model.layers.46.self_attn.v_proj.weight.nested_quant_map": "model-
|
| 1854 |
-
"model.layers.46.self_attn.v_proj.weight.quant_map": "model-
|
| 1855 |
-
"model.layers.46.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1856 |
-
"model.layers.47.input_layernorm.weight": "model-
|
| 1857 |
-
"model.layers.47.mlp.down_proj.weight": "model-
|
| 1858 |
-
"model.layers.47.mlp.down_proj.weight.absmax": "model-
|
| 1859 |
-
"model.layers.47.mlp.down_proj.weight.nested_absmax": "model-
|
| 1860 |
-
"model.layers.47.mlp.down_proj.weight.nested_quant_map": "model-
|
| 1861 |
-
"model.layers.47.mlp.down_proj.weight.quant_map": "model-
|
| 1862 |
-
"model.layers.47.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1863 |
-
"model.layers.47.mlp.gate_proj.weight": "model-
|
| 1864 |
-
"model.layers.47.mlp.gate_proj.weight.absmax": "model-
|
| 1865 |
-
"model.layers.47.mlp.gate_proj.weight.nested_absmax": "model-
|
| 1866 |
-
"model.layers.47.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 1867 |
-
"model.layers.47.mlp.gate_proj.weight.quant_map": "model-
|
| 1868 |
-
"model.layers.47.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1869 |
-
"model.layers.47.mlp.up_proj.weight": "model-
|
| 1870 |
-
"model.layers.47.mlp.up_proj.weight.absmax": "model-
|
| 1871 |
-
"model.layers.47.mlp.up_proj.weight.nested_absmax": "model-
|
| 1872 |
-
"model.layers.47.mlp.up_proj.weight.nested_quant_map": "model-
|
| 1873 |
-
"model.layers.47.mlp.up_proj.weight.quant_map": "model-
|
| 1874 |
-
"model.layers.47.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1875 |
-
"model.layers.47.post_attention_layernorm.weight": "model-
|
| 1876 |
-
"model.layers.47.self_attn.k_proj.weight": "model-
|
| 1877 |
-
"model.layers.47.self_attn.k_proj.weight.absmax": "model-
|
| 1878 |
-
"model.layers.47.self_attn.k_proj.weight.nested_absmax": "model-
|
| 1879 |
-
"model.layers.47.self_attn.k_proj.weight.nested_quant_map": "model-
|
| 1880 |
-
"model.layers.47.self_attn.k_proj.weight.quant_map": "model-
|
| 1881 |
-
"model.layers.47.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1882 |
-
"model.layers.47.self_attn.o_proj.weight": "model-
|
| 1883 |
-
"model.layers.47.self_attn.o_proj.weight.absmax": "model-
|
| 1884 |
-
"model.layers.47.self_attn.o_proj.weight.nested_absmax": "model-
|
| 1885 |
-
"model.layers.47.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 1886 |
-
"model.layers.47.self_attn.o_proj.weight.quant_map": "model-
|
| 1887 |
-
"model.layers.47.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1888 |
-
"model.layers.47.self_attn.q_proj.weight": "model-
|
| 1889 |
-
"model.layers.47.self_attn.q_proj.weight.absmax": "model-
|
| 1890 |
-
"model.layers.47.self_attn.q_proj.weight.nested_absmax": "model-
|
| 1891 |
-
"model.layers.47.self_attn.q_proj.weight.nested_quant_map": "model-
|
| 1892 |
-
"model.layers.47.self_attn.q_proj.weight.quant_map": "model-
|
| 1893 |
-
"model.layers.47.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1894 |
-
"model.layers.47.self_attn.v_proj.weight": "model-
|
| 1895 |
-
"model.layers.47.self_attn.v_proj.weight.absmax": "model-
|
| 1896 |
-
"model.layers.47.self_attn.v_proj.weight.nested_absmax": "model-
|
| 1897 |
-
"model.layers.47.self_attn.v_proj.weight.nested_quant_map": "model-
|
| 1898 |
-
"model.layers.47.self_attn.v_proj.weight.quant_map": "model-
|
| 1899 |
-
"model.layers.47.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1900 |
-
"model.layers.48.input_layernorm.weight": "model-
|
| 1901 |
-
"model.layers.48.mlp.down_proj.weight": "model-
|
| 1902 |
-
"model.layers.48.mlp.down_proj.weight.absmax": "model-
|
| 1903 |
-
"model.layers.48.mlp.down_proj.weight.nested_absmax": "model-
|
| 1904 |
-
"model.layers.48.mlp.down_proj.weight.nested_quant_map": "model-
|
| 1905 |
-
"model.layers.48.mlp.down_proj.weight.quant_map": "model-
|
| 1906 |
-
"model.layers.48.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1907 |
-
"model.layers.48.mlp.gate_proj.weight": "model-
|
| 1908 |
-
"model.layers.48.mlp.gate_proj.weight.absmax": "model-
|
| 1909 |
-
"model.layers.48.mlp.gate_proj.weight.nested_absmax": "model-
|
| 1910 |
-
"model.layers.48.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 1911 |
-
"model.layers.48.mlp.gate_proj.weight.quant_map": "model-
|
| 1912 |
-
"model.layers.48.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1913 |
-
"model.layers.48.mlp.up_proj.weight": "model-
|
| 1914 |
-
"model.layers.48.mlp.up_proj.weight.absmax": "model-
|
| 1915 |
-
"model.layers.48.mlp.up_proj.weight.nested_absmax": "model-
|
| 1916 |
-
"model.layers.48.mlp.up_proj.weight.nested_quant_map": "model-
|
| 1917 |
-
"model.layers.48.mlp.up_proj.weight.quant_map": "model-
|
| 1918 |
-
"model.layers.48.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1919 |
-
"model.layers.48.post_attention_layernorm.weight": "model-
|
| 1920 |
-
"model.layers.48.self_attn.k_proj.weight": "model-
|
| 1921 |
-
"model.layers.48.self_attn.k_proj.weight.absmax": "model-
|
| 1922 |
-
"model.layers.48.self_attn.k_proj.weight.nested_absmax": "model-
|
| 1923 |
-
"model.layers.48.self_attn.k_proj.weight.nested_quant_map": "model-
|
| 1924 |
-
"model.layers.48.self_attn.k_proj.weight.quant_map": "model-
|
| 1925 |
-
"model.layers.48.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1926 |
-
"model.layers.48.self_attn.o_proj.weight": "model-
|
| 1927 |
-
"model.layers.48.self_attn.o_proj.weight.absmax": "model-
|
| 1928 |
-
"model.layers.48.self_attn.o_proj.weight.nested_absmax": "model-
|
| 1929 |
-
"model.layers.48.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 1930 |
-
"model.layers.48.self_attn.o_proj.weight.quant_map": "model-
|
| 1931 |
-
"model.layers.48.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1932 |
-
"model.layers.48.self_attn.q_proj.weight": "model-
|
| 1933 |
-
"model.layers.48.self_attn.q_proj.weight.absmax": "model-
|
| 1934 |
-
"model.layers.48.self_attn.q_proj.weight.nested_absmax": "model-
|
| 1935 |
-
"model.layers.48.self_attn.q_proj.weight.nested_quant_map": "model-
|
| 1936 |
-
"model.layers.48.self_attn.q_proj.weight.quant_map": "model-
|
| 1937 |
-
"model.layers.48.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1938 |
-
"model.layers.48.self_attn.v_proj.weight": "model-
|
| 1939 |
-
"model.layers.48.self_attn.v_proj.weight.absmax": "model-
|
| 1940 |
-
"model.layers.48.self_attn.v_proj.weight.nested_absmax": "model-
|
| 1941 |
-
"model.layers.48.self_attn.v_proj.weight.nested_quant_map": "model-
|
| 1942 |
-
"model.layers.48.self_attn.v_proj.weight.quant_map": "model-
|
| 1943 |
-
"model.layers.48.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1944 |
-
"model.layers.49.input_layernorm.weight": "model-
|
| 1945 |
-
"model.layers.49.mlp.down_proj.weight": "model-
|
| 1946 |
-
"model.layers.49.mlp.down_proj.weight.absmax": "model-
|
| 1947 |
-
"model.layers.49.mlp.down_proj.weight.nested_absmax": "model-
|
| 1948 |
-
"model.layers.49.mlp.down_proj.weight.nested_quant_map": "model-
|
| 1949 |
-
"model.layers.49.mlp.down_proj.weight.quant_map": "model-
|
| 1950 |
-
"model.layers.49.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1951 |
-
"model.layers.49.mlp.gate_proj.weight": "model-
|
| 1952 |
-
"model.layers.49.mlp.gate_proj.weight.absmax": "model-
|
| 1953 |
-
"model.layers.49.mlp.gate_proj.weight.nested_absmax": "model-
|
| 1954 |
-
"model.layers.49.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 1955 |
-
"model.layers.49.mlp.gate_proj.weight.quant_map": "model-
|
| 1956 |
-
"model.layers.49.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1957 |
-
"model.layers.49.mlp.up_proj.weight": "model-
|
| 1958 |
-
"model.layers.49.mlp.up_proj.weight.absmax": "model-
|
| 1959 |
-
"model.layers.49.mlp.up_proj.weight.nested_absmax": "model-
|
| 1960 |
-
"model.layers.49.mlp.up_proj.weight.nested_quant_map": "model-
|
| 1961 |
-
"model.layers.49.mlp.up_proj.weight.quant_map": "model-
|
| 1962 |
-
"model.layers.49.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1963 |
-
"model.layers.49.post_attention_layernorm.weight": "model-
|
| 1964 |
-
"model.layers.49.self_attn.k_proj.weight": "model-
|
| 1965 |
-
"model.layers.49.self_attn.k_proj.weight.absmax": "model-
|
| 1966 |
-
"model.layers.49.self_attn.k_proj.weight.nested_absmax": "model-
|
| 1967 |
-
"model.layers.49.self_attn.k_proj.weight.nested_quant_map": "model-
|
| 1968 |
-
"model.layers.49.self_attn.k_proj.weight.quant_map": "model-
|
| 1969 |
-
"model.layers.49.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1970 |
-
"model.layers.49.self_attn.o_proj.weight": "model-
|
| 1971 |
-
"model.layers.49.self_attn.o_proj.weight.absmax": "model-
|
| 1972 |
-
"model.layers.49.self_attn.o_proj.weight.nested_absmax": "model-
|
| 1973 |
-
"model.layers.49.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 1974 |
-
"model.layers.49.self_attn.o_proj.weight.quant_map": "model-
|
| 1975 |
-
"model.layers.49.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1976 |
-
"model.layers.49.self_attn.q_proj.weight": "model-
|
| 1977 |
-
"model.layers.49.self_attn.q_proj.weight.absmax": "model-
|
| 1978 |
-
"model.layers.49.self_attn.q_proj.weight.nested_absmax": "model-
|
| 1979 |
-
"model.layers.49.self_attn.q_proj.weight.nested_quant_map": "model-
|
| 1980 |
-
"model.layers.49.self_attn.q_proj.weight.quant_map": "model-
|
| 1981 |
-
"model.layers.49.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1982 |
-
"model.layers.49.self_attn.v_proj.weight": "model-
|
| 1983 |
-
"model.layers.49.self_attn.v_proj.weight.absmax": "model-
|
| 1984 |
-
"model.layers.49.self_attn.v_proj.weight.nested_absmax": "model-
|
| 1985 |
-
"model.layers.49.self_attn.v_proj.weight.nested_quant_map": "model-
|
| 1986 |
-
"model.layers.49.self_attn.v_proj.weight.quant_map": "model-
|
| 1987 |
-
"model.layers.49.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 1988 |
"model.layers.5.input_layernorm.weight": "model-00001-of-00003.safetensors",
|
| 1989 |
"model.layers.5.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
|
| 1990 |
"model.layers.5.mlp.down_proj.weight.absmax": "model-00001-of-00003.safetensors",
|
|
@@ -2029,94 +2029,94 @@
|
|
| 2029 |
"model.layers.5.self_attn.v_proj.weight.nested_quant_map": "model-00001-of-00003.safetensors",
|
| 2030 |
"model.layers.5.self_attn.v_proj.weight.quant_map": "model-00001-of-00003.safetensors",
|
| 2031 |
"model.layers.5.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00001-of-00003.safetensors",
|
| 2032 |
-
"model.layers.50.input_layernorm.weight": "model-
|
| 2033 |
-
"model.layers.50.mlp.down_proj.weight": "model-
|
| 2034 |
-
"model.layers.50.mlp.down_proj.weight.absmax": "model-
|
| 2035 |
-
"model.layers.50.mlp.down_proj.weight.nested_absmax": "model-
|
| 2036 |
-
"model.layers.50.mlp.down_proj.weight.nested_quant_map": "model-
|
| 2037 |
-
"model.layers.50.mlp.down_proj.weight.quant_map": "model-
|
| 2038 |
-
"model.layers.50.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2039 |
-
"model.layers.50.mlp.gate_proj.weight": "model-
|
| 2040 |
-
"model.layers.50.mlp.gate_proj.weight.absmax": "model-
|
| 2041 |
-
"model.layers.50.mlp.gate_proj.weight.nested_absmax": "model-
|
| 2042 |
-
"model.layers.50.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 2043 |
-
"model.layers.50.mlp.gate_proj.weight.quant_map": "model-
|
| 2044 |
-
"model.layers.50.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2045 |
-
"model.layers.50.mlp.up_proj.weight": "model-
|
| 2046 |
-
"model.layers.50.mlp.up_proj.weight.absmax": "model-
|
| 2047 |
-
"model.layers.50.mlp.up_proj.weight.nested_absmax": "model-
|
| 2048 |
-
"model.layers.50.mlp.up_proj.weight.nested_quant_map": "model-
|
| 2049 |
-
"model.layers.50.mlp.up_proj.weight.quant_map": "model-
|
| 2050 |
-
"model.layers.50.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2051 |
-
"model.layers.50.post_attention_layernorm.weight": "model-
|
| 2052 |
-
"model.layers.50.self_attn.k_proj.weight": "model-
|
| 2053 |
-
"model.layers.50.self_attn.k_proj.weight.absmax": "model-
|
| 2054 |
-
"model.layers.50.self_attn.k_proj.weight.nested_absmax": "model-
|
| 2055 |
-
"model.layers.50.self_attn.k_proj.weight.nested_quant_map": "model-
|
| 2056 |
-
"model.layers.50.self_attn.k_proj.weight.quant_map": "model-
|
| 2057 |
-
"model.layers.50.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2058 |
-
"model.layers.50.self_attn.o_proj.weight": "model-
|
| 2059 |
-
"model.layers.50.self_attn.o_proj.weight.absmax": "model-
|
| 2060 |
-
"model.layers.50.self_attn.o_proj.weight.nested_absmax": "model-
|
| 2061 |
-
"model.layers.50.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 2062 |
-
"model.layers.50.self_attn.o_proj.weight.quant_map": "model-
|
| 2063 |
-
"model.layers.50.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2064 |
-
"model.layers.50.self_attn.q_proj.weight": "model-
|
| 2065 |
-
"model.layers.50.self_attn.q_proj.weight.absmax": "model-
|
| 2066 |
-
"model.layers.50.self_attn.q_proj.weight.nested_absmax": "model-
|
| 2067 |
-
"model.layers.50.self_attn.q_proj.weight.nested_quant_map": "model-
|
| 2068 |
-
"model.layers.50.self_attn.q_proj.weight.quant_map": "model-
|
| 2069 |
-
"model.layers.50.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2070 |
-
"model.layers.50.self_attn.v_proj.weight": "model-
|
| 2071 |
-
"model.layers.50.self_attn.v_proj.weight.absmax": "model-
|
| 2072 |
-
"model.layers.50.self_attn.v_proj.weight.nested_absmax": "model-
|
| 2073 |
-
"model.layers.50.self_attn.v_proj.weight.nested_quant_map": "model-
|
| 2074 |
-
"model.layers.50.self_attn.v_proj.weight.quant_map": "model-
|
| 2075 |
-
"model.layers.50.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2076 |
-
"model.layers.51.input_layernorm.weight": "model-
|
| 2077 |
-
"model.layers.51.mlp.down_proj.weight": "model-
|
| 2078 |
-
"model.layers.51.mlp.down_proj.weight.absmax": "model-
|
| 2079 |
-
"model.layers.51.mlp.down_proj.weight.nested_absmax": "model-
|
| 2080 |
-
"model.layers.51.mlp.down_proj.weight.nested_quant_map": "model-
|
| 2081 |
-
"model.layers.51.mlp.down_proj.weight.quant_map": "model-
|
| 2082 |
-
"model.layers.51.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2083 |
-
"model.layers.51.mlp.gate_proj.weight": "model-
|
| 2084 |
-
"model.layers.51.mlp.gate_proj.weight.absmax": "model-
|
| 2085 |
-
"model.layers.51.mlp.gate_proj.weight.nested_absmax": "model-
|
| 2086 |
-
"model.layers.51.mlp.gate_proj.weight.nested_quant_map": "model-
|
| 2087 |
-
"model.layers.51.mlp.gate_proj.weight.quant_map": "model-
|
| 2088 |
-
"model.layers.51.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2089 |
-
"model.layers.51.mlp.up_proj.weight": "model-
|
| 2090 |
-
"model.layers.51.mlp.up_proj.weight.absmax": "model-
|
| 2091 |
-
"model.layers.51.mlp.up_proj.weight.nested_absmax": "model-
|
| 2092 |
-
"model.layers.51.mlp.up_proj.weight.nested_quant_map": "model-
|
| 2093 |
-
"model.layers.51.mlp.up_proj.weight.quant_map": "model-
|
| 2094 |
-
"model.layers.51.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2095 |
-
"model.layers.51.post_attention_layernorm.weight": "model-
|
| 2096 |
-
"model.layers.51.self_attn.k_proj.weight": "model-
|
| 2097 |
-
"model.layers.51.self_attn.k_proj.weight.absmax": "model-
|
| 2098 |
-
"model.layers.51.self_attn.k_proj.weight.nested_absmax": "model-
|
| 2099 |
-
"model.layers.51.self_attn.k_proj.weight.nested_quant_map": "model-
|
| 2100 |
-
"model.layers.51.self_attn.k_proj.weight.quant_map": "model-
|
| 2101 |
-
"model.layers.51.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2102 |
-
"model.layers.51.self_attn.o_proj.weight": "model-
|
| 2103 |
-
"model.layers.51.self_attn.o_proj.weight.absmax": "model-
|
| 2104 |
-
"model.layers.51.self_attn.o_proj.weight.nested_absmax": "model-
|
| 2105 |
-
"model.layers.51.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 2106 |
-
"model.layers.51.self_attn.o_proj.weight.quant_map": "model-
|
| 2107 |
-
"model.layers.51.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2108 |
-
"model.layers.51.self_attn.q_proj.weight": "model-
|
| 2109 |
-
"model.layers.51.self_attn.q_proj.weight.absmax": "model-
|
| 2110 |
-
"model.layers.51.self_attn.q_proj.weight.nested_absmax": "model-
|
| 2111 |
-
"model.layers.51.self_attn.q_proj.weight.nested_quant_map": "model-
|
| 2112 |
-
"model.layers.51.self_attn.q_proj.weight.quant_map": "model-
|
| 2113 |
-
"model.layers.51.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2114 |
-
"model.layers.51.self_attn.v_proj.weight": "model-
|
| 2115 |
-
"model.layers.51.self_attn.v_proj.weight.absmax": "model-
|
| 2116 |
-
"model.layers.51.self_attn.v_proj.weight.nested_absmax": "model-
|
| 2117 |
-
"model.layers.51.self_attn.v_proj.weight.nested_quant_map": "model-
|
| 2118 |
-
"model.layers.51.self_attn.v_proj.weight.quant_map": "model-
|
| 2119 |
-
"model.layers.51.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2120 |
"model.layers.52.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 2121 |
"model.layers.52.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 2122 |
"model.layers.52.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
|
@@ -2137,30 +2137,30 @@
|
|
| 2137 |
"model.layers.52.mlp.up_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2138 |
"model.layers.52.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2139 |
"model.layers.52.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 2140 |
-
"model.layers.52.self_attn.k_proj.weight": "model-
|
| 2141 |
-
"model.layers.52.self_attn.k_proj.weight.absmax": "model-
|
| 2142 |
-
"model.layers.52.self_attn.k_proj.weight.nested_absmax": "model-
|
| 2143 |
-
"model.layers.52.self_attn.k_proj.weight.nested_quant_map": "model-
|
| 2144 |
-
"model.layers.52.self_attn.k_proj.weight.quant_map": "model-
|
| 2145 |
-
"model.layers.52.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2146 |
-
"model.layers.52.self_attn.o_proj.weight": "model-
|
| 2147 |
-
"model.layers.52.self_attn.o_proj.weight.absmax": "model-
|
| 2148 |
-
"model.layers.52.self_attn.o_proj.weight.nested_absmax": "model-
|
| 2149 |
-
"model.layers.52.self_attn.o_proj.weight.nested_quant_map": "model-
|
| 2150 |
-
"model.layers.52.self_attn.o_proj.weight.quant_map": "model-
|
| 2151 |
-
"model.layers.52.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2152 |
-
"model.layers.52.self_attn.q_proj.weight": "model-
|
| 2153 |
-
"model.layers.52.self_attn.q_proj.weight.absmax": "model-
|
| 2154 |
-
"model.layers.52.self_attn.q_proj.weight.nested_absmax": "model-
|
| 2155 |
-
"model.layers.52.self_attn.q_proj.weight.nested_quant_map": "model-
|
| 2156 |
-
"model.layers.52.self_attn.q_proj.weight.quant_map": "model-
|
| 2157 |
-
"model.layers.52.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2158 |
-
"model.layers.52.self_attn.v_proj.weight": "model-
|
| 2159 |
-
"model.layers.52.self_attn.v_proj.weight.absmax": "model-
|
| 2160 |
-
"model.layers.52.self_attn.v_proj.weight.nested_absmax": "model-
|
| 2161 |
-
"model.layers.52.self_attn.v_proj.weight.nested_quant_map": "model-
|
| 2162 |
-
"model.layers.52.self_attn.v_proj.weight.quant_map": "model-
|
| 2163 |
-
"model.layers.52.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-
|
| 2164 |
"model.layers.53.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 2165 |
"model.layers.53.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 2166 |
"model.layers.53.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
|
|
|
| 621 |
"model.layers.20.self_attn.v_proj.weight.nested_quant_map": "model-00001-of-00003.safetensors",
|
| 622 |
"model.layers.20.self_attn.v_proj.weight.quant_map": "model-00001-of-00003.safetensors",
|
| 623 |
"model.layers.20.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00001-of-00003.safetensors",
|
| 624 |
+
"model.layers.21.input_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 625 |
+
"model.layers.21.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
|
| 626 |
+
"model.layers.21.mlp.down_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 627 |
+
"model.layers.21.mlp.down_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 628 |
+
"model.layers.21.mlp.down_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 629 |
+
"model.layers.21.mlp.down_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 630 |
+
"model.layers.21.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 631 |
+
"model.layers.21.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
|
| 632 |
+
"model.layers.21.mlp.gate_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 633 |
+
"model.layers.21.mlp.gate_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 634 |
+
"model.layers.21.mlp.gate_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 635 |
+
"model.layers.21.mlp.gate_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 636 |
+
"model.layers.21.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 637 |
+
"model.layers.21.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
|
| 638 |
+
"model.layers.21.mlp.up_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 639 |
+
"model.layers.21.mlp.up_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 640 |
+
"model.layers.21.mlp.up_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 641 |
+
"model.layers.21.mlp.up_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 642 |
+
"model.layers.21.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 643 |
+
"model.layers.21.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 644 |
"model.layers.21.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
|
| 645 |
"model.layers.21.self_attn.k_proj.weight.absmax": "model-00001-of-00003.safetensors",
|
| 646 |
"model.layers.21.self_attn.k_proj.weight.nested_absmax": "model-00001-of-00003.safetensors",
|
| 647 |
"model.layers.21.self_attn.k_proj.weight.nested_quant_map": "model-00001-of-00003.safetensors",
|
| 648 |
"model.layers.21.self_attn.k_proj.weight.quant_map": "model-00001-of-00003.safetensors",
|
| 649 |
"model.layers.21.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00001-of-00003.safetensors",
|
| 650 |
+
"model.layers.21.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
|
| 651 |
+
"model.layers.21.self_attn.o_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 652 |
+
"model.layers.21.self_attn.o_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 653 |
+
"model.layers.21.self_attn.o_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 654 |
+
"model.layers.21.self_attn.o_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 655 |
+
"model.layers.21.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 656 |
"model.layers.21.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
|
| 657 |
"model.layers.21.self_attn.q_proj.weight.absmax": "model-00001-of-00003.safetensors",
|
| 658 |
"model.layers.21.self_attn.q_proj.weight.nested_absmax": "model-00001-of-00003.safetensors",
|
|
|
|
| 665 |
"model.layers.21.self_attn.v_proj.weight.nested_quant_map": "model-00001-of-00003.safetensors",
|
| 666 |
"model.layers.21.self_attn.v_proj.weight.quant_map": "model-00001-of-00003.safetensors",
|
| 667 |
"model.layers.21.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00001-of-00003.safetensors",
|
| 668 |
+
"model.layers.22.input_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 669 |
+
"model.layers.22.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
|
| 670 |
+
"model.layers.22.mlp.down_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 671 |
+
"model.layers.22.mlp.down_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 672 |
+
"model.layers.22.mlp.down_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 673 |
+
"model.layers.22.mlp.down_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 674 |
+
"model.layers.22.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 675 |
+
"model.layers.22.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
|
| 676 |
+
"model.layers.22.mlp.gate_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 677 |
+
"model.layers.22.mlp.gate_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 678 |
+
"model.layers.22.mlp.gate_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 679 |
+
"model.layers.22.mlp.gate_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 680 |
+
"model.layers.22.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 681 |
+
"model.layers.22.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
|
| 682 |
+
"model.layers.22.mlp.up_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 683 |
+
"model.layers.22.mlp.up_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 684 |
+
"model.layers.22.mlp.up_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 685 |
+
"model.layers.22.mlp.up_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 686 |
+
"model.layers.22.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 687 |
+
"model.layers.22.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 688 |
+
"model.layers.22.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
|
| 689 |
+
"model.layers.22.self_attn.k_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 690 |
+
"model.layers.22.self_attn.k_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 691 |
+
"model.layers.22.self_attn.k_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 692 |
+
"model.layers.22.self_attn.k_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 693 |
+
"model.layers.22.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 694 |
+
"model.layers.22.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
|
| 695 |
+
"model.layers.22.self_attn.o_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 696 |
+
"model.layers.22.self_attn.o_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 697 |
+
"model.layers.22.self_attn.o_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 698 |
+
"model.layers.22.self_attn.o_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 699 |
+
"model.layers.22.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 700 |
+
"model.layers.22.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
|
| 701 |
+
"model.layers.22.self_attn.q_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 702 |
+
"model.layers.22.self_attn.q_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 703 |
+
"model.layers.22.self_attn.q_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 704 |
+
"model.layers.22.self_attn.q_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 705 |
+
"model.layers.22.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 706 |
+
"model.layers.22.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
|
| 707 |
+
"model.layers.22.self_attn.v_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 708 |
+
"model.layers.22.self_attn.v_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 709 |
+
"model.layers.22.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 710 |
+
"model.layers.22.self_attn.v_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 711 |
+
"model.layers.22.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 712 |
+
"model.layers.23.input_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 713 |
+
"model.layers.23.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
|
| 714 |
+
"model.layers.23.mlp.down_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 715 |
+
"model.layers.23.mlp.down_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 716 |
+
"model.layers.23.mlp.down_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 717 |
+
"model.layers.23.mlp.down_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 718 |
+
"model.layers.23.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 719 |
+
"model.layers.23.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
|
| 720 |
+
"model.layers.23.mlp.gate_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 721 |
+
"model.layers.23.mlp.gate_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 722 |
+
"model.layers.23.mlp.gate_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 723 |
+
"model.layers.23.mlp.gate_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 724 |
+
"model.layers.23.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 725 |
+
"model.layers.23.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
|
| 726 |
+
"model.layers.23.mlp.up_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 727 |
+
"model.layers.23.mlp.up_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 728 |
+
"model.layers.23.mlp.up_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 729 |
+
"model.layers.23.mlp.up_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 730 |
+
"model.layers.23.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 731 |
+
"model.layers.23.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 732 |
+
"model.layers.23.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
|
| 733 |
+
"model.layers.23.self_attn.k_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 734 |
+
"model.layers.23.self_attn.k_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 735 |
+
"model.layers.23.self_attn.k_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 736 |
+
"model.layers.23.self_attn.k_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 737 |
+
"model.layers.23.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 738 |
+
"model.layers.23.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
|
| 739 |
+
"model.layers.23.self_attn.o_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 740 |
+
"model.layers.23.self_attn.o_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 741 |
+
"model.layers.23.self_attn.o_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 742 |
+
"model.layers.23.self_attn.o_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 743 |
+
"model.layers.23.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 744 |
+
"model.layers.23.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
|
| 745 |
+
"model.layers.23.self_attn.q_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 746 |
+
"model.layers.23.self_attn.q_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 747 |
+
"model.layers.23.self_attn.q_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 748 |
+
"model.layers.23.self_attn.q_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 749 |
+
"model.layers.23.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 750 |
+
"model.layers.23.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
|
| 751 |
+
"model.layers.23.self_attn.v_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 752 |
+
"model.layers.23.self_attn.v_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 753 |
+
"model.layers.23.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 754 |
+
"model.layers.23.self_attn.v_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 755 |
+
"model.layers.23.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 756 |
"model.layers.24.input_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 757 |
"model.layers.24.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
|
| 758 |
"model.layers.24.mlp.down_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
|
|
|
| 760 |
"model.layers.24.mlp.down_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 761 |
"model.layers.24.mlp.down_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 762 |
"model.layers.24.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 763 |
+
"model.layers.24.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
|
| 764 |
+
"model.layers.24.mlp.gate_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 765 |
+
"model.layers.24.mlp.gate_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 766 |
+
"model.layers.24.mlp.gate_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 767 |
+
"model.layers.24.mlp.gate_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 768 |
+
"model.layers.24.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 769 |
"model.layers.24.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
|
| 770 |
"model.layers.24.mlp.up_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 771 |
"model.layers.24.mlp.up_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
|
|
|
| 773 |
"model.layers.24.mlp.up_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 774 |
"model.layers.24.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 775 |
"model.layers.24.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 776 |
+
"model.layers.24.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
|
| 777 |
+
"model.layers.24.self_attn.k_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 778 |
+
"model.layers.24.self_attn.k_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 779 |
+
"model.layers.24.self_attn.k_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 780 |
+
"model.layers.24.self_attn.k_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 781 |
+
"model.layers.24.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 782 |
+
"model.layers.24.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
|
| 783 |
+
"model.layers.24.self_attn.o_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 784 |
+
"model.layers.24.self_attn.o_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 785 |
+
"model.layers.24.self_attn.o_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 786 |
+
"model.layers.24.self_attn.o_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 787 |
+
"model.layers.24.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 788 |
+
"model.layers.24.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
|
| 789 |
+
"model.layers.24.self_attn.q_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 790 |
+
"model.layers.24.self_attn.q_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 791 |
+
"model.layers.24.self_attn.q_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 792 |
+
"model.layers.24.self_attn.q_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 793 |
+
"model.layers.24.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 794 |
+
"model.layers.24.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
|
| 795 |
+
"model.layers.24.self_attn.v_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 796 |
+
"model.layers.24.self_attn.v_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
| 797 |
+
"model.layers.24.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 798 |
+
"model.layers.24.self_attn.v_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 799 |
+
"model.layers.24.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 800 |
"model.layers.25.input_layernorm.weight": "model-00002-of-00003.safetensors",
|
| 801 |
"model.layers.25.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
|
| 802 |
"model.layers.25.mlp.down_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
|
|
|
| 1765 |
"model.layers.44.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 1766 |
"model.layers.44.self_attn.v_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 1767 |
"model.layers.44.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 1768 |
+
"model.layers.45.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 1769 |
+
"model.layers.45.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 1770 |
+
"model.layers.45.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1771 |
+
"model.layers.45.mlp.down_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1772 |
+
"model.layers.45.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1773 |
+
"model.layers.45.mlp.down_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1774 |
+
"model.layers.45.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1775 |
+
"model.layers.45.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
|
| 1776 |
+
"model.layers.45.mlp.gate_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1777 |
+
"model.layers.45.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1778 |
+
"model.layers.45.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1779 |
+
"model.layers.45.mlp.gate_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1780 |
+
"model.layers.45.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1781 |
+
"model.layers.45.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
|
| 1782 |
+
"model.layers.45.mlp.up_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1783 |
+
"model.layers.45.mlp.up_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1784 |
+
"model.layers.45.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1785 |
+
"model.layers.45.mlp.up_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1786 |
+
"model.layers.45.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1787 |
+
"model.layers.45.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 1788 |
"model.layers.45.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
|
| 1789 |
"model.layers.45.self_attn.k_proj.weight.absmax": "model-00002-of-00003.safetensors",
|
| 1790 |
"model.layers.45.self_attn.k_proj.weight.nested_absmax": "model-00002-of-00003.safetensors",
|
|
|
|
| 1809 |
"model.layers.45.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00003.safetensors",
|
| 1810 |
"model.layers.45.self_attn.v_proj.weight.quant_map": "model-00002-of-00003.safetensors",
|
| 1811 |
"model.layers.45.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00003.safetensors",
|
| 1812 |
+
"model.layers.46.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 1813 |
+
"model.layers.46.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 1814 |
+
"model.layers.46.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1815 |
+
"model.layers.46.mlp.down_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1816 |
+
"model.layers.46.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1817 |
+
"model.layers.46.mlp.down_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1818 |
+
"model.layers.46.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1819 |
+
"model.layers.46.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
|
| 1820 |
+
"model.layers.46.mlp.gate_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1821 |
+
"model.layers.46.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1822 |
+
"model.layers.46.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1823 |
+
"model.layers.46.mlp.gate_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1824 |
+
"model.layers.46.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1825 |
+
"model.layers.46.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
|
| 1826 |
+
"model.layers.46.mlp.up_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1827 |
+
"model.layers.46.mlp.up_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1828 |
+
"model.layers.46.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1829 |
+
"model.layers.46.mlp.up_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1830 |
+
"model.layers.46.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1831 |
+
"model.layers.46.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 1832 |
+
"model.layers.46.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
|
| 1833 |
+
"model.layers.46.self_attn.k_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1834 |
+
"model.layers.46.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1835 |
+
"model.layers.46.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1836 |
+
"model.layers.46.self_attn.k_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1837 |
+
"model.layers.46.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1838 |
+
"model.layers.46.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
|
| 1839 |
+
"model.layers.46.self_attn.o_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1840 |
+
"model.layers.46.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1841 |
+
"model.layers.46.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1842 |
+
"model.layers.46.self_attn.o_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1843 |
+
"model.layers.46.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1844 |
+
"model.layers.46.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
|
| 1845 |
+
"model.layers.46.self_attn.q_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1846 |
+
"model.layers.46.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1847 |
+
"model.layers.46.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1848 |
+
"model.layers.46.self_attn.q_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1849 |
+
"model.layers.46.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1850 |
+
"model.layers.46.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
|
| 1851 |
+
"model.layers.46.self_attn.v_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1852 |
+
"model.layers.46.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1853 |
+
"model.layers.46.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1854 |
+
"model.layers.46.self_attn.v_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1855 |
+
"model.layers.46.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1856 |
+
"model.layers.47.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 1857 |
+
"model.layers.47.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 1858 |
+
"model.layers.47.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1859 |
+
"model.layers.47.mlp.down_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1860 |
+
"model.layers.47.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1861 |
+
"model.layers.47.mlp.down_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1862 |
+
"model.layers.47.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1863 |
+
"model.layers.47.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
|
| 1864 |
+
"model.layers.47.mlp.gate_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1865 |
+
"model.layers.47.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1866 |
+
"model.layers.47.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1867 |
+
"model.layers.47.mlp.gate_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1868 |
+
"model.layers.47.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1869 |
+
"model.layers.47.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
|
| 1870 |
+
"model.layers.47.mlp.up_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1871 |
+
"model.layers.47.mlp.up_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1872 |
+
"model.layers.47.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1873 |
+
"model.layers.47.mlp.up_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1874 |
+
"model.layers.47.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1875 |
+
"model.layers.47.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 1876 |
+
"model.layers.47.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
|
| 1877 |
+
"model.layers.47.self_attn.k_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1878 |
+
"model.layers.47.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1879 |
+
"model.layers.47.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1880 |
+
"model.layers.47.self_attn.k_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1881 |
+
"model.layers.47.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1882 |
+
"model.layers.47.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
|
| 1883 |
+
"model.layers.47.self_attn.o_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1884 |
+
"model.layers.47.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1885 |
+
"model.layers.47.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1886 |
+
"model.layers.47.self_attn.o_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1887 |
+
"model.layers.47.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1888 |
+
"model.layers.47.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
|
| 1889 |
+
"model.layers.47.self_attn.q_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1890 |
+
"model.layers.47.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1891 |
+
"model.layers.47.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1892 |
+
"model.layers.47.self_attn.q_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1893 |
+
"model.layers.47.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1894 |
+
"model.layers.47.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
|
| 1895 |
+
"model.layers.47.self_attn.v_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1896 |
+
"model.layers.47.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1897 |
+
"model.layers.47.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1898 |
+
"model.layers.47.self_attn.v_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1899 |
+
"model.layers.47.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1900 |
+
"model.layers.48.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 1901 |
+
"model.layers.48.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 1902 |
+
"model.layers.48.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1903 |
+
"model.layers.48.mlp.down_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1904 |
+
"model.layers.48.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1905 |
+
"model.layers.48.mlp.down_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1906 |
+
"model.layers.48.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1907 |
+
"model.layers.48.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
|
| 1908 |
+
"model.layers.48.mlp.gate_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1909 |
+
"model.layers.48.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1910 |
+
"model.layers.48.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1911 |
+
"model.layers.48.mlp.gate_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1912 |
+
"model.layers.48.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1913 |
+
"model.layers.48.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
|
| 1914 |
+
"model.layers.48.mlp.up_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1915 |
+
"model.layers.48.mlp.up_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1916 |
+
"model.layers.48.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1917 |
+
"model.layers.48.mlp.up_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1918 |
+
"model.layers.48.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1919 |
+
"model.layers.48.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 1920 |
+
"model.layers.48.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
|
| 1921 |
+
"model.layers.48.self_attn.k_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1922 |
+
"model.layers.48.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1923 |
+
"model.layers.48.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1924 |
+
"model.layers.48.self_attn.k_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1925 |
+
"model.layers.48.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1926 |
+
"model.layers.48.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
|
| 1927 |
+
"model.layers.48.self_attn.o_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1928 |
+
"model.layers.48.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1929 |
+
"model.layers.48.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1930 |
+
"model.layers.48.self_attn.o_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1931 |
+
"model.layers.48.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1932 |
+
"model.layers.48.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
|
| 1933 |
+
"model.layers.48.self_attn.q_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1934 |
+
"model.layers.48.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1935 |
+
"model.layers.48.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1936 |
+
"model.layers.48.self_attn.q_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1937 |
+
"model.layers.48.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1938 |
+
"model.layers.48.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
|
| 1939 |
+
"model.layers.48.self_attn.v_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1940 |
+
"model.layers.48.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1941 |
+
"model.layers.48.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1942 |
+
"model.layers.48.self_attn.v_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1943 |
+
"model.layers.48.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1944 |
+
"model.layers.49.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 1945 |
+
"model.layers.49.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 1946 |
+
"model.layers.49.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1947 |
+
"model.layers.49.mlp.down_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1948 |
+
"model.layers.49.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1949 |
+
"model.layers.49.mlp.down_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1950 |
+
"model.layers.49.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1951 |
+
"model.layers.49.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
|
| 1952 |
+
"model.layers.49.mlp.gate_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1953 |
+
"model.layers.49.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1954 |
+
"model.layers.49.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1955 |
+
"model.layers.49.mlp.gate_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1956 |
+
"model.layers.49.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1957 |
+
"model.layers.49.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
|
| 1958 |
+
"model.layers.49.mlp.up_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1959 |
+
"model.layers.49.mlp.up_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1960 |
+
"model.layers.49.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1961 |
+
"model.layers.49.mlp.up_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1962 |
+
"model.layers.49.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1963 |
+
"model.layers.49.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 1964 |
+
"model.layers.49.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
|
| 1965 |
+
"model.layers.49.self_attn.k_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1966 |
+
"model.layers.49.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1967 |
+
"model.layers.49.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1968 |
+
"model.layers.49.self_attn.k_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1969 |
+
"model.layers.49.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1970 |
+
"model.layers.49.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
|
| 1971 |
+
"model.layers.49.self_attn.o_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1972 |
+
"model.layers.49.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1973 |
+
"model.layers.49.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1974 |
+
"model.layers.49.self_attn.o_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1975 |
+
"model.layers.49.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1976 |
+
"model.layers.49.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
|
| 1977 |
+
"model.layers.49.self_attn.q_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1978 |
+
"model.layers.49.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1979 |
+
"model.layers.49.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1980 |
+
"model.layers.49.self_attn.q_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1981 |
+
"model.layers.49.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1982 |
+
"model.layers.49.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
|
| 1983 |
+
"model.layers.49.self_attn.v_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 1984 |
+
"model.layers.49.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 1985 |
+
"model.layers.49.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 1986 |
+
"model.layers.49.self_attn.v_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 1987 |
+
"model.layers.49.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 1988 |
"model.layers.5.input_layernorm.weight": "model-00001-of-00003.safetensors",
|
| 1989 |
"model.layers.5.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
|
| 1990 |
"model.layers.5.mlp.down_proj.weight.absmax": "model-00001-of-00003.safetensors",
|
|
|
|
| 2029 |
"model.layers.5.self_attn.v_proj.weight.nested_quant_map": "model-00001-of-00003.safetensors",
|
| 2030 |
"model.layers.5.self_attn.v_proj.weight.quant_map": "model-00001-of-00003.safetensors",
|
| 2031 |
"model.layers.5.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00001-of-00003.safetensors",
|
| 2032 |
+
"model.layers.50.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 2033 |
+
"model.layers.50.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 2034 |
+
"model.layers.50.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2035 |
+
"model.layers.50.mlp.down_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2036 |
+
"model.layers.50.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2037 |
+
"model.layers.50.mlp.down_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2038 |
+
"model.layers.50.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2039 |
+
"model.layers.50.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
|
| 2040 |
+
"model.layers.50.mlp.gate_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2041 |
+
"model.layers.50.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2042 |
+
"model.layers.50.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2043 |
+
"model.layers.50.mlp.gate_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2044 |
+
"model.layers.50.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2045 |
+
"model.layers.50.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
|
| 2046 |
+
"model.layers.50.mlp.up_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2047 |
+
"model.layers.50.mlp.up_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2048 |
+
"model.layers.50.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2049 |
+
"model.layers.50.mlp.up_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2050 |
+
"model.layers.50.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2051 |
+
"model.layers.50.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 2052 |
+
"model.layers.50.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
|
| 2053 |
+
"model.layers.50.self_attn.k_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2054 |
+
"model.layers.50.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2055 |
+
"model.layers.50.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2056 |
+
"model.layers.50.self_attn.k_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2057 |
+
"model.layers.50.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2058 |
+
"model.layers.50.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
|
| 2059 |
+
"model.layers.50.self_attn.o_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2060 |
+
"model.layers.50.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2061 |
+
"model.layers.50.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2062 |
+
"model.layers.50.self_attn.o_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2063 |
+
"model.layers.50.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2064 |
+
"model.layers.50.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
|
| 2065 |
+
"model.layers.50.self_attn.q_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2066 |
+
"model.layers.50.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2067 |
+
"model.layers.50.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2068 |
+
"model.layers.50.self_attn.q_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2069 |
+
"model.layers.50.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2070 |
+
"model.layers.50.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
|
| 2071 |
+
"model.layers.50.self_attn.v_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2072 |
+
"model.layers.50.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2073 |
+
"model.layers.50.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2074 |
+
"model.layers.50.self_attn.v_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2075 |
+
"model.layers.50.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2076 |
+
"model.layers.51.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 2077 |
+
"model.layers.51.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 2078 |
+
"model.layers.51.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2079 |
+
"model.layers.51.mlp.down_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2080 |
+
"model.layers.51.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2081 |
+
"model.layers.51.mlp.down_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2082 |
+
"model.layers.51.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2083 |
+
"model.layers.51.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
|
| 2084 |
+
"model.layers.51.mlp.gate_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2085 |
+
"model.layers.51.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2086 |
+
"model.layers.51.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2087 |
+
"model.layers.51.mlp.gate_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2088 |
+
"model.layers.51.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2089 |
+
"model.layers.51.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
|
| 2090 |
+
"model.layers.51.mlp.up_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2091 |
+
"model.layers.51.mlp.up_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2092 |
+
"model.layers.51.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2093 |
+
"model.layers.51.mlp.up_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2094 |
+
"model.layers.51.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2095 |
+
"model.layers.51.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 2096 |
+
"model.layers.51.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
|
| 2097 |
+
"model.layers.51.self_attn.k_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2098 |
+
"model.layers.51.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2099 |
+
"model.layers.51.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2100 |
+
"model.layers.51.self_attn.k_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2101 |
+
"model.layers.51.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2102 |
+
"model.layers.51.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
|
| 2103 |
+
"model.layers.51.self_attn.o_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2104 |
+
"model.layers.51.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2105 |
+
"model.layers.51.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2106 |
+
"model.layers.51.self_attn.o_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2107 |
+
"model.layers.51.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2108 |
+
"model.layers.51.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
|
| 2109 |
+
"model.layers.51.self_attn.q_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2110 |
+
"model.layers.51.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2111 |
+
"model.layers.51.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2112 |
+
"model.layers.51.self_attn.q_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2113 |
+
"model.layers.51.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2114 |
+
"model.layers.51.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
|
| 2115 |
+
"model.layers.51.self_attn.v_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2116 |
+
"model.layers.51.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2117 |
+
"model.layers.51.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2118 |
+
"model.layers.51.self_attn.v_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2119 |
+
"model.layers.51.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2120 |
"model.layers.52.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 2121 |
"model.layers.52.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 2122 |
"model.layers.52.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
|
|
|
| 2137 |
"model.layers.52.mlp.up_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2138 |
"model.layers.52.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2139 |
"model.layers.52.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 2140 |
+
"model.layers.52.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
|
| 2141 |
+
"model.layers.52.self_attn.k_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2142 |
+
"model.layers.52.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2143 |
+
"model.layers.52.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2144 |
+
"model.layers.52.self_attn.k_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2145 |
+
"model.layers.52.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2146 |
+
"model.layers.52.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
|
| 2147 |
+
"model.layers.52.self_attn.o_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2148 |
+
"model.layers.52.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2149 |
+
"model.layers.52.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2150 |
+
"model.layers.52.self_attn.o_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2151 |
+
"model.layers.52.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2152 |
+
"model.layers.52.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
|
| 2153 |
+
"model.layers.52.self_attn.q_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2154 |
+
"model.layers.52.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2155 |
+
"model.layers.52.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2156 |
+
"model.layers.52.self_attn.q_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2157 |
+
"model.layers.52.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2158 |
+
"model.layers.52.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
|
| 2159 |
+
"model.layers.52.self_attn.v_proj.weight.absmax": "model-00003-of-00003.safetensors",
|
| 2160 |
+
"model.layers.52.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00003.safetensors",
|
| 2161 |
+
"model.layers.52.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00003.safetensors",
|
| 2162 |
+
"model.layers.52.self_attn.v_proj.weight.quant_map": "model-00003-of-00003.safetensors",
|
| 2163 |
+
"model.layers.52.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00003.safetensors",
|
| 2164 |
"model.layers.53.input_layernorm.weight": "model-00003-of-00003.safetensors",
|
| 2165 |
"model.layers.53.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
|
| 2166 |
"model.layers.53.mlp.down_proj.weight.absmax": "model-00003-of-00003.safetensors",
|