diff --git "a/model.safetensors.index.json" "b/model.safetensors.index.json"
--- "a/model.safetensors.index.json"
+++ "b/model.safetensors.index.json"
@@ -1,6 +1,6 @@
 {
   "metadata": {
-    "total_size": 22483575955
+    "total_size": 23744109829
   },
   "weight_map": {
     "lm_head.weight": "model-00005-of-00005.safetensors",
@@ -101,23 +101,8 @@
     "model.layers.1.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00001-of-00005.safetensors",
     "model.layers.10.input_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.10.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.down_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.down_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.down_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.down_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
     "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.gate_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.gate_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.gate_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.gate_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
     "model.layers.10.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.up_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.up_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.up_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.up_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.10.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
     "model.layers.10.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.10.self_attn.k_proj.bias": "model-00002-of-00005.safetensors",
     "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
@@ -413,26 +398,26 @@
     "model.layers.16.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
     "model.layers.16.self_attn.v_proj.weight.quant_map": "model-00002-of-00005.safetensors",
     "model.layers.16.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.17.input_layernorm.weight": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.down_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.down_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.down_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.down_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.gate_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.gate_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.gate_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.gate_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.up_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.up_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.up_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.up_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.17.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.17.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.down_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.down_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.down_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.gate_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.gate_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.up_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.up_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.up_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.17.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.17.self_attn.k_proj.bias": "model-00002-of-00005.safetensors",
     "model.layers.17.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
     "model.layers.17.self_attn.k_proj.weight.absmax": "model-00002-of-00005.safetensors",
@@ -440,12 +425,12 @@
     "model.layers.17.self_attn.k_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
     "model.layers.17.self_attn.k_proj.weight.quant_map": "model-00002-of-00005.safetensors",
     "model.layers.17.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.17.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.17.self_attn.o_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.17.self_attn.o_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.17.self_attn.o_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.17.self_attn.o_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.17.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.17.self_attn.o_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.17.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.17.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.17.self_attn.o_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.17.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
     "model.layers.17.self_attn.q_proj.bias": "model-00002-of-00005.safetensors",
     "model.layers.17.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
     "model.layers.17.self_attn.q_proj.weight.absmax": "model-00002-of-00005.safetensors",
@@ -460,53 +445,53 @@
     "model.layers.17.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
     "model.layers.17.self_attn.v_proj.weight.quant_map": "model-00002-of-00005.safetensors",
     "model.layers.17.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.18.input_layernorm.weight": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.down_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.down_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.down_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.down_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.gate_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.gate_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.gate_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.gate_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.up_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.up_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.up_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.up_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.18.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.k_proj.bias": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.k_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.k_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.k_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.k_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.o_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.o_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.o_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.o_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.q_proj.bias": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.q_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.q_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.q_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.q_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.v_proj.bias": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.v_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.v_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.v_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.18.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.down_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.down_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.down_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.gate_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.gate_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.up_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.up_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.up_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.k_proj.bias": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.k_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.k_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.o_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.o_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.q_proj.bias": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.q_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.q_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.v_proj.bias": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.v_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.v_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.18.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
     "model.layers.19.input_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.19.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
     "model.layers.19.mlp.down_proj.weight.absmax": "model-00003-of-00005.safetensors",
@@ -514,12 +499,12 @@
     "model.layers.19.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
     "model.layers.19.mlp.down_proj.weight.quant_map": "model-00003-of-00005.safetensors",
     "model.layers.19.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.19.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.19.mlp.gate_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.19.mlp.gate_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.19.mlp.gate_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.19.mlp.gate_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.19.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.19.mlp.gate_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.19.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.19.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.19.mlp.gate_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.19.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
     "model.layers.19.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
     "model.layers.19.mlp.up_proj.weight.absmax": "model-00003-of-00005.safetensors",
     "model.layers.19.mlp.up_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
@@ -527,33 +512,33 @@
     "model.layers.19.mlp.up_proj.weight.quant_map": "model-00003-of-00005.safetensors",
     "model.layers.19.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
     "model.layers.19.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
-    "model.layers.19.self_attn.k_proj.bias": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.k_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.k_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.k_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.k_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.o_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.o_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.o_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.o_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.o_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.q_proj.bias": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.q_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.q_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.q_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.q_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.q_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.v_proj.bias": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.v_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.v_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.v_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.v_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.v_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.19.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
+    "model.layers.19.self_attn.k_proj.bias": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.k_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.k_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.o_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.o_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.q_proj.bias": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.q_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.q_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.v_proj.bias": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.v_proj.weight.absmax": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.v_proj.weight.quant_map": "model-00003-of-00005.safetensors",
+    "model.layers.19.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
     "model.layers.2.input_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.2.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
     "model.layers.2.mlp.down_proj.weight.absmax": "model-00001-of-00005.safetensors",
@@ -697,23 +682,8 @@
     "model.layers.21.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
     "model.layers.22.input_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.22.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.down_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.down_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.down_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
     "model.layers.22.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.gate_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.gate_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
     "model.layers.22.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.up_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.up_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.up_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.22.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
     "model.layers.22.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.22.self_attn.k_proj.bias": "model-00003-of-00005.safetensors",
     "model.layers.22.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
@@ -1306,26 +1276,26 @@
     "model.layers.33.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
     "model.layers.33.self_attn.v_proj.weight.quant_map": "model-00003-of-00005.safetensors",
     "model.layers.33.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.34.input_layernorm.weight": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.down_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.down_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.down_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.34.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.down_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.down_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.down_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
     "model.layers.34.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
     "model.layers.34.mlp.gate_proj.weight.absmax": "model-00003-of-00005.safetensors",
     "model.layers.34.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
     "model.layers.34.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
     "model.layers.34.mlp.gate_proj.weight.quant_map": "model-00003-of-00005.safetensors",
     "model.layers.34.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.up_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.up_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.up_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.34.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.34.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
+    "model.layers.34.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.up_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.up_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.up_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.34.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.34.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.34.self_attn.k_proj.bias": "model-00003-of-00005.safetensors",
     "model.layers.34.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
     "model.layers.34.self_attn.k_proj.weight.absmax": "model-00003-of-00005.safetensors",
@@ -1353,194 +1323,194 @@
     "model.layers.34.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
     "model.layers.34.self_attn.v_proj.weight.quant_map": "model-00003-of-00005.safetensors",
     "model.layers.34.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.35.input_layernorm.weight": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.down_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.down_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.down_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.gate_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.gate_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.up_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.up_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.up_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.35.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.k_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.k_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.k_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.o_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.o_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.q_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.q_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.q_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.v_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.v_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.v_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.35.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.36.input_layernorm.weight": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.down_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.down_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.down_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.gate_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.gate_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.up_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.up_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.up_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.36.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.k_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.k_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.k_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.o_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.o_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.q_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.q_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.q_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.v_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.v_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.v_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.36.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.37.input_layernorm.weight": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.down_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.down_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.down_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.gate_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.gate_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.up_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.up_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.up_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.37.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.k_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.k_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.k_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.o_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.o_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.q_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.q_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.q_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.v_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.v_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.v_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.37.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.38.input_layernorm.weight": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.down_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.down_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.down_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.down_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.down_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.gate_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.gate_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.gate_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.gate_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.gate_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.up_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.up_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.up_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.up_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.up_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.38.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.k_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.k_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.k_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.o_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.o_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.q_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.q_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.q_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.v_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.v_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.v_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.38.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.35.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.down_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.down_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.down_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.gate_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.gate_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.gate_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.gate_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.up_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.up_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.up_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.35.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.k_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.k_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.k_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.k_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.o_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.o_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.o_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.o_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.q_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.q_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.q_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.q_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.v_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.v_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.35.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.36.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.down_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.down_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.down_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.gate_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.gate_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.gate_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.gate_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.up_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.up_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.up_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.36.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.k_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.k_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.k_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.k_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.o_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.o_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.o_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.o_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.q_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.q_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.q_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.q_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.v_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.v_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.36.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.37.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.down_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.down_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.down_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.gate_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.gate_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.gate_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.gate_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.up_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.up_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.up_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.37.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.k_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.k_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.k_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.k_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.o_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.o_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.o_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.o_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.q_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.q_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.q_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.q_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.v_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.v_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.37.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.38.input_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.down_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.down_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.down_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.gate_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.gate_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.gate_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.gate_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.up_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.up_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.up_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.38.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.k_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.k_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.k_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.k_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.o_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.o_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.o_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.o_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.q_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.q_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.q_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.q_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.v_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.v_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.38.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
     "model.layers.39.input_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.39.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
     "model.layers.39.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
@@ -1561,33 +1531,33 @@
     "model.layers.39.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
     "model.layers.39.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
     "model.layers.39.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
-    "model.layers.39.self_attn.k_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.k_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.k_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.k_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.k_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.k_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.o_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.o_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.o_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.o_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.o_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.q_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.q_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.q_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.q_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.q_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.q_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.v_proj.bias": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.v_proj.weight": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.v_proj.weight.absmax": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.v_proj.weight.nested_absmax": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.v_proj.weight.nested_quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.v_proj.weight.quant_map": "model-00003-of-00005.safetensors",
-    "model.layers.39.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00003-of-00005.safetensors",
+    "model.layers.39.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.k_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.k_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.k_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.k_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.o_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.o_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.o_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.o_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.q_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.q_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.q_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.q_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.v_proj.weight.absmax": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.v_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
+    "model.layers.39.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
     "model.layers.4.input_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.4.mlp.down_proj.weight": "model-00001-of-00005.safetensors",
     "model.layers.4.mlp.down_proj.weight.absmax": "model-00001-of-00005.safetensors",
@@ -2340,26 +2310,26 @@
     "model.layers.53.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
     "model.layers.53.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
     "model.layers.53.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.54.input_layernorm.weight": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.down_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.down_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.down_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.gate_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.gate_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.gate_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.gate_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.up_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.up_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.up_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.54.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.54.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
+    "model.layers.54.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.down_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.down_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.down_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.down_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.gate_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.gate_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.gate_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.gate_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.gate_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.up_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.up_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.up_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.up_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.54.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.54.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.54.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
     "model.layers.54.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
     "model.layers.54.self_attn.k_proj.weight.absmax": "model-00004-of-00005.safetensors",
@@ -2387,194 +2357,194 @@
     "model.layers.54.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
     "model.layers.54.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
     "model.layers.54.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.55.input_layernorm.weight": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.down_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.down_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.down_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.gate_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.gate_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.gate_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.gate_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.up_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.up_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.up_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.55.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.k_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.k_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.k_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.k_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.o_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.o_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.o_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.o_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.q_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.q_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.q_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.q_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.v_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.v_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.55.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.56.input_layernorm.weight": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.down_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.down_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.down_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.gate_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.gate_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.gate_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.gate_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.up_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.up_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.up_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.56.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.k_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.k_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.k_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.k_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.o_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.o_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.o_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.o_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.q_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.q_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.q_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.q_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.v_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.v_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.56.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.57.input_layernorm.weight": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.down_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.down_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.down_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.gate_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.gate_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.gate_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.gate_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.up_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.up_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.up_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.57.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.k_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.k_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.k_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.k_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.o_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.o_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.o_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.o_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.q_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.q_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.q_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.q_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.v_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.v_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.57.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.58.input_layernorm.weight": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.down_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.down_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.down_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.down_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.down_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.gate_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.gate_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.gate_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.gate_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.gate_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.up_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.up_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.up_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.up_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.up_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.58.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.k_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.k_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.k_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.k_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.k_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.k_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.o_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.o_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.o_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.o_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.o_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.q_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.q_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.q_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.q_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.q_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.q_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.v_proj.bias": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.v_proj.weight": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.v_proj.weight.absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.v_proj.weight.nested_absmax": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.v_proj.weight.nested_quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.v_proj.weight.quant_map": "model-00004-of-00005.safetensors",
-    "model.layers.58.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00004-of-00005.safetensors",
+    "model.layers.55.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.down_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.down_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.down_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.down_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.gate_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.gate_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.gate_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.gate_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.gate_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.up_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.up_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.up_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.up_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.55.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.k_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.k_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.k_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.k_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.k_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.k_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.o_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.o_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.o_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.o_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.o_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.q_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.q_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.q_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.q_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.q_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.q_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.v_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.v_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.v_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.v_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.v_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.v_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.55.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.56.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.down_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.down_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.down_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.down_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.gate_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.gate_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.gate_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.gate_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.gate_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.up_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.up_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.up_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.up_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.56.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.k_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.k_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.k_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.k_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.k_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.k_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.o_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.o_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.o_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.o_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.o_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.q_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.q_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.q_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.q_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.q_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.q_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.v_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.v_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.v_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.v_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.v_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.v_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.56.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.57.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.down_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.down_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.down_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.down_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.gate_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.gate_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.gate_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.gate_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.gate_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.up_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.up_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.up_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.up_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.57.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.k_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.k_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.k_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.k_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.k_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.k_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.o_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.o_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.o_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.o_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.o_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.q_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.q_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.q_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.q_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.q_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.q_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.v_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.v_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.v_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.v_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.v_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.v_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.57.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.58.input_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.down_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.down_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.down_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.down_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.gate_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.gate_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.gate_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.gate_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.gate_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.up_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.up_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.up_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.up_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.up_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.58.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.k_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.k_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.k_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.k_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.k_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.k_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.k_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.o_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.o_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.o_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.o_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.o_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.o_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.q_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.q_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.q_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.q_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.q_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.q_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.q_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.v_proj.bias": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.v_proj.weight": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.v_proj.weight.absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.v_proj.weight.nested_absmax": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.v_proj.weight.nested_quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.v_proj.weight.quant_map": "model-00005-of-00005.safetensors",
+    "model.layers.58.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
     "model.layers.59.input_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.59.mlp.down_proj.weight": "model-00005-of-00005.safetensors",
     "model.layers.59.mlp.down_proj.weight.absmax": "model-00005-of-00005.safetensors",
@@ -2844,23 +2814,8 @@
     "model.layers.63.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00005-of-00005.safetensors",
     "model.layers.7.input_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.7.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.down_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.down_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.down_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.down_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
     "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.gate_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.gate_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.gate_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.gate_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
     "model.layers.7.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.up_proj.weight.absmax": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.up_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.up_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.up_proj.weight.quant_map": "model-00002-of-00005.safetensors",
-    "model.layers.7.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
     "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.7.self_attn.k_proj.bias": "model-00002-of-00005.safetensors",
     "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",
@@ -2891,8 +2846,23 @@
     "model.layers.7.self_attn.v_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
     "model.layers.8.input_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.8.mlp.down_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.down_proj.weight.absmax": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.down_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.down_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.down_proj.weight.quant_map": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.down_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
     "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.gate_proj.weight.absmax": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.gate_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.gate_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.gate_proj.weight.quant_map": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.gate_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
     "model.layers.8.mlp.up_proj.weight": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.up_proj.weight.absmax": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.up_proj.weight.nested_absmax": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.up_proj.weight.nested_quant_map": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.up_proj.weight.quant_map": "model-00002-of-00005.safetensors",
+    "model.layers.8.mlp.up_proj.weight.quant_state.bitsandbytes__nf4": "model-00002-of-00005.safetensors",
     "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.8.self_attn.k_proj.bias": "model-00002-of-00005.safetensors",
     "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00005.safetensors",