Upload folder using huggingface_hub

config.json

@@ -27,8 +27,40 @@
       }
     },
     "format": "pack-quantized",
-    "global_compression_ratio": 2.1823781700242217,
+    "global_compression_ratio": 2.0968847636762966,
     "ignore": [
+      "model.layers.0.block_sparse_moe.gate",
+      "model.layers.1.block_sparse_moe.gate",
+      "model.layers.2.block_sparse_moe.gate",
+      "model.layers.3.block_sparse_moe.gate",
+      "model.layers.4.block_sparse_moe.gate",
+      "model.layers.5.block_sparse_moe.gate",
+      "model.layers.6.block_sparse_moe.gate",
+      "model.layers.7.block_sparse_moe.gate",
+      "model.layers.8.block_sparse_moe.gate",
+      "model.layers.9.block_sparse_moe.gate",
+      "model.layers.10.block_sparse_moe.gate",
+      "model.layers.11.block_sparse_moe.gate",
+      "model.layers.12.block_sparse_moe.gate",
+      "model.layers.13.block_sparse_moe.gate",
+      "model.layers.14.block_sparse_moe.gate",
+      "model.layers.15.block_sparse_moe.gate",
+      "model.layers.16.block_sparse_moe.gate",
+      "model.layers.17.block_sparse_moe.gate",
+      "model.layers.18.block_sparse_moe.gate",
+      "model.layers.19.block_sparse_moe.gate",
+      "model.layers.20.block_sparse_moe.gate",
+      "model.layers.21.block_sparse_moe.gate",
+      "model.layers.22.block_sparse_moe.gate",
+      "model.layers.23.block_sparse_moe.gate",
+      "model.layers.24.block_sparse_moe.gate",
+      "model.layers.25.block_sparse_moe.gate",
+      "model.layers.26.block_sparse_moe.gate",
+      "model.layers.27.block_sparse_moe.gate",
+      "model.layers.28.block_sparse_moe.gate",
+      "model.layers.29.block_sparse_moe.gate",
+      "model.layers.30.block_sparse_moe.gate",
+      "model.layers.31.block_sparse_moe.gate",
       "lm_head"
     ],
     "kv_cache_scheme": null,
@@ -36,7 +68,7 @@
     "quantization_status": "frozen",
     "sparsity_config": {
       "format": "dense",
-      "global_sparsity": 0.15880852433171147,
+      "global_sparsity": 0.1588017024690434,
       "registry_requires_subclass": false,
       "sparsity_structure": "unstructured"
     }

model-00001-of-00005.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:203a9eb11275ff428f84f80fde387fd9096b385a789061e46dc842f46033580d
-size 4985745808
+oid sha256:80356de2146cfe15890405bcbe8e2c69a26fea557a631ccb972d0c0d99862b76
+size 4986084536

model-00002-of-00005.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:84c2deb8163905cf3b65d737eab333a3fc97c4b32b587630471b0bb841f7b110
-size 4974458544
+oid sha256:7a5aa1d5ee0fdf7a8df95b9dad3c95d6721a63f913ebb9fcda79be31f01479c2
+size 4974748880

model-00003-of-00005.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d5ae426b1d41e617d05b489b9c152ee69631d1834bfce678620ad76f5a30caa7
-size 4996121080
+oid sha256:1836d8ae0a57a331a2516cb82a052054dfbcecdeef5ee343fb78f44800de74e7
+size 4996459792

model-00004-of-00005.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eda77f6ebce9f9b21a9367dc39f36be7a61e8daa2e9948350e66f3f3651a2062
-size 4996121080
+oid sha256:80b4bc548dcf6df9f48beb2388238cee432cb3c15789782843c2bd4acf02b5d4
+size 4996459792

model-00005-of-00005.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:48fccb390f8bbab17213c9680cfd1be9c72ad50aba4f44fbd960913b0564bd8a
-size 4518570688
+oid sha256:660a49aed1b85e276e4767868bfdd9ed5ab2a47890de7f7564bbcd53b59ff111
+size 4518812624

model.safetensors.index.json

@@ -1,6 +1,6 @@
 {
   "metadata": {
-    "total_size": 24470657536
+    "total_size": 24472213504
   },
   "weight_map": {
     "lm_head.weight": "model-00005-of-00005.safetensors",
@@ -77,9 +77,7 @@
     "model.layers.0.block_sparse_moe.experts.7.w3.weight_packed": "model-00001-of-00005.safetensors",
     "model.layers.0.block_sparse_moe.experts.7.w3.weight_scale": "model-00001-of-00005.safetensors",
     "model.layers.0.block_sparse_moe.experts.7.w3.weight_shape": "model-00001-of-00005.safetensors",
-    "model.layers.0.block_sparse_moe.gate.weight_packed": "model-00001-of-00005.safetensors",
-    "model.layers.0.block_sparse_moe.gate.weight_scale": "model-00001-of-00005.safetensors",
-    "model.layers.0.block_sparse_moe.gate.weight_shape": "model-00001-of-00005.safetensors",
+    "model.layers.0.block_sparse_moe.gate.weight": "model-00001-of-00005.safetensors",
     "model.layers.0.input_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.0.self_attn.k_proj.weight_packed": "model-00001-of-00005.safetensors",
@@ -166,9 +164,7 @@
     "model.layers.1.block_sparse_moe.experts.7.w3.weight_packed": "model-00001-of-00005.safetensors",
     "model.layers.1.block_sparse_moe.experts.7.w3.weight_scale": "model-00001-of-00005.safetensors",
     "model.layers.1.block_sparse_moe.experts.7.w3.weight_shape": "model-00001-of-00005.safetensors",
-    "model.layers.1.block_sparse_moe.gate.weight_packed": "model-00001-of-00005.safetensors",
-    "model.layers.1.block_sparse_moe.gate.weight_scale": "model-00001-of-00005.safetensors",
-    "model.layers.1.block_sparse_moe.gate.weight_shape": "model-00001-of-00005.safetensors",
+    "model.layers.1.block_sparse_moe.gate.weight": "model-00001-of-00005.safetensors",
     "model.layers.1.input_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.1.self_attn.k_proj.weight_packed": "model-00001-of-00005.safetensors",
@@ -255,9 +251,7 @@
     "model.layers.10.block_sparse_moe.experts.7.w3.weight_packed": "model-00002-of-00005.safetensors",
     "model.layers.10.block_sparse_moe.experts.7.w3.weight_scale": "model-00002-of-00005.safetensors",
     "model.layers.10.block_sparse_moe.experts.7.w3.weight_shape": "model-00002-of-00005.safetensors",
-    "model.layers.10.block_sparse_moe.gate.weight_packed": "model-00002-of-00005.safetensors",
-    "model.layers.10.block_sparse_moe.gate.weight_scale": "model-00002-of-00005.safetensors",
-    "model.layers.10.block_sparse_moe.gate.weight_shape": "model-00002-of-00005.safetensors",
+    "model.layers.10.block_sparse_moe.gate.weight": "model-00002-of-00005.safetensors",
     "model.layers.10.input_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.10.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.10.self_attn.k_proj.weight_packed": "model-00002-of-00005.safetensors",
@@ -344,9 +338,7 @@
     "model.layers.11.block_sparse_moe.experts.7.w3.weight_packed": "model-00002-of-00005.safetensors",
     "model.layers.11.block_sparse_moe.experts.7.w3.weight_scale": "model-00002-of-00005.safetensors",
     "model.layers.11.block_sparse_moe.experts.7.w3.weight_shape": "model-00002-of-00005.safetensors",
-    "model.layers.11.block_sparse_moe.gate.weight_packed": "model-00002-of-00005.safetensors",
-    "model.layers.11.block_sparse_moe.gate.weight_scale": "model-00002-of-00005.safetensors",
-    "model.layers.11.block_sparse_moe.gate.weight_shape": "model-00002-of-00005.safetensors",
+    "model.layers.11.block_sparse_moe.gate.weight": "model-00002-of-00005.safetensors",
     "model.layers.11.input_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.11.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.11.self_attn.k_proj.weight_packed": "model-00002-of-00005.safetensors",
@@ -433,9 +425,7 @@
     "model.layers.12.block_sparse_moe.experts.7.w3.weight_packed": "model-00003-of-00005.safetensors",
     "model.layers.12.block_sparse_moe.experts.7.w3.weight_scale": "model-00003-of-00005.safetensors",
     "model.layers.12.block_sparse_moe.experts.7.w3.weight_shape": "model-00002-of-00005.safetensors",
-    "model.layers.12.block_sparse_moe.gate.weight_packed": "model-00002-of-00005.safetensors",
-    "model.layers.12.block_sparse_moe.gate.weight_scale": "model-00002-of-00005.safetensors",
-    "model.layers.12.block_sparse_moe.gate.weight_shape": "model-00002-of-00005.safetensors",
+    "model.layers.12.block_sparse_moe.gate.weight": "model-00002-of-00005.safetensors",
     "model.layers.12.input_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.12.self_attn.k_proj.weight_packed": "model-00002-of-00005.safetensors",
@@ -522,9 +512,7 @@
     "model.layers.13.block_sparse_moe.experts.7.w3.weight_packed": "model-00003-of-00005.safetensors",
     "model.layers.13.block_sparse_moe.experts.7.w3.weight_scale": "model-00003-of-00005.safetensors",
     "model.layers.13.block_sparse_moe.experts.7.w3.weight_shape": "model-00003-of-00005.safetensors",
-    "model.layers.13.block_sparse_moe.gate.weight_packed": "model-00003-of-00005.safetensors",
-    "model.layers.13.block_sparse_moe.gate.weight_scale": "model-00003-of-00005.safetensors",
-    "model.layers.13.block_sparse_moe.gate.weight_shape": "model-00003-of-00005.safetensors",
+    "model.layers.13.block_sparse_moe.gate.weight": "model-00003-of-00005.safetensors",
     "model.layers.13.input_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.13.self_attn.k_proj.weight_packed": "model-00003-of-00005.safetensors",
@@ -611,9 +599,7 @@
     "model.layers.14.block_sparse_moe.experts.7.w3.weight_packed": "model-00003-of-00005.safetensors",
     "model.layers.14.block_sparse_moe.experts.7.w3.weight_scale": "model-00003-of-00005.safetensors",
     "model.layers.14.block_sparse_moe.experts.7.w3.weight_shape": "model-00003-of-00005.safetensors",
-    "model.layers.14.block_sparse_moe.gate.weight_packed": "model-00003-of-00005.safetensors",
-    "model.layers.14.block_sparse_moe.gate.weight_scale": "model-00003-of-00005.safetensors",
-    "model.layers.14.block_sparse_moe.gate.weight_shape": "model-00003-of-00005.safetensors",
+    "model.layers.14.block_sparse_moe.gate.weight": "model-00003-of-00005.safetensors",
     "model.layers.14.input_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.14.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.14.self_attn.k_proj.weight_packed": "model-00003-of-00005.safetensors",
@@ -700,9 +686,7 @@
     "model.layers.15.block_sparse_moe.experts.7.w3.weight_packed": "model-00003-of-00005.safetensors",
     "model.layers.15.block_sparse_moe.experts.7.w3.weight_scale": "model-00003-of-00005.safetensors",
     "model.layers.15.block_sparse_moe.experts.7.w3.weight_shape": "model-00003-of-00005.safetensors",
-    "model.layers.15.block_sparse_moe.gate.weight_packed": "model-00003-of-00005.safetensors",
-    "model.layers.15.block_sparse_moe.gate.weight_scale": "model-00003-of-00005.safetensors",
-    "model.layers.15.block_sparse_moe.gate.weight_shape": "model-00003-of-00005.safetensors",
+    "model.layers.15.block_sparse_moe.gate.weight": "model-00003-of-00005.safetensors",
     "model.layers.15.input_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.15.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.15.self_attn.k_proj.weight_packed": "model-00003-of-00005.safetensors",
@@ -789,9 +773,7 @@
     "model.layers.16.block_sparse_moe.experts.7.w3.weight_packed": "model-00003-of-00005.safetensors",
     "model.layers.16.block_sparse_moe.experts.7.w3.weight_scale": "model-00003-of-00005.safetensors",
     "model.layers.16.block_sparse_moe.experts.7.w3.weight_shape": "model-00003-of-00005.safetensors",
-    "model.layers.16.block_sparse_moe.gate.weight_packed": "model-00003-of-00005.safetensors",
-    "model.layers.16.block_sparse_moe.gate.weight_scale": "model-00003-of-00005.safetensors",
-    "model.layers.16.block_sparse_moe.gate.weight_shape": "model-00003-of-00005.safetensors",
+    "model.layers.16.block_sparse_moe.gate.weight": "model-00003-of-00005.safetensors",
     "model.layers.16.input_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.16.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.16.self_attn.k_proj.weight_packed": "model-00003-of-00005.safetensors",
@@ -878,9 +860,7 @@
     "model.layers.17.block_sparse_moe.experts.7.w3.weight_packed": "model-00003-of-00005.safetensors",
     "model.layers.17.block_sparse_moe.experts.7.w3.weight_scale": "model-00003-of-00005.safetensors",
     "model.layers.17.block_sparse_moe.experts.7.w3.weight_shape": "model-00003-of-00005.safetensors",
-    "model.layers.17.block_sparse_moe.gate.weight_packed": "model-00003-of-00005.safetensors",
-    "model.layers.17.block_sparse_moe.gate.weight_scale": "model-00003-of-00005.safetensors",
-    "model.layers.17.block_sparse_moe.gate.weight_shape": "model-00003-of-00005.safetensors",
+    "model.layers.17.block_sparse_moe.gate.weight": "model-00003-of-00005.safetensors",
     "model.layers.17.input_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.17.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.17.self_attn.k_proj.weight_packed": "model-00003-of-00005.safetensors",
@@ -967,9 +947,7 @@
     "model.layers.18.block_sparse_moe.experts.7.w3.weight_packed": "model-00003-of-00005.safetensors",
     "model.layers.18.block_sparse_moe.experts.7.w3.weight_scale": "model-00003-of-00005.safetensors",
     "model.layers.18.block_sparse_moe.experts.7.w3.weight_shape": "model-00003-of-00005.safetensors",
-    "model.layers.18.block_sparse_moe.gate.weight_packed": "model-00003-of-00005.safetensors",
-    "model.layers.18.block_sparse_moe.gate.weight_scale": "model-00003-of-00005.safetensors",
-    "model.layers.18.block_sparse_moe.gate.weight_shape": "model-00003-of-00005.safetensors",
+    "model.layers.18.block_sparse_moe.gate.weight": "model-00003-of-00005.safetensors",
     "model.layers.18.input_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.18.post_attention_layernorm.weight": "model-00003-of-00005.safetensors",
     "model.layers.18.self_attn.k_proj.weight_packed": "model-00003-of-00005.safetensors",
@@ -1056,9 +1034,7 @@
     "model.layers.19.block_sparse_moe.experts.7.w3.weight_packed": "model-00004-of-00005.safetensors",
     "model.layers.19.block_sparse_moe.experts.7.w3.weight_scale": "model-00004-of-00005.safetensors",
     "model.layers.19.block_sparse_moe.experts.7.w3.weight_shape": "model-00004-of-00005.safetensors",
-    "model.layers.19.block_sparse_moe.gate.weight_packed": "model-00003-of-00005.safetensors",
-    "model.layers.19.block_sparse_moe.gate.weight_scale": "model-00003-of-00005.safetensors",
-    "model.layers.19.block_sparse_moe.gate.weight_shape": "model-00003-of-00005.safetensors",
+    "model.layers.19.block_sparse_moe.gate.weight": "model-00003-of-00005.safetensors",
     "model.layers.19.input_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.19.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.19.self_attn.k_proj.weight_packed": "model-00003-of-00005.safetensors",
@@ -1145,9 +1121,7 @@
     "model.layers.2.block_sparse_moe.experts.7.w3.weight_packed": "model-00001-of-00005.safetensors",
     "model.layers.2.block_sparse_moe.experts.7.w3.weight_scale": "model-00001-of-00005.safetensors",
     "model.layers.2.block_sparse_moe.experts.7.w3.weight_shape": "model-00001-of-00005.safetensors",
-    "model.layers.2.block_sparse_moe.gate.weight_packed": "model-00001-of-00005.safetensors",
-    "model.layers.2.block_sparse_moe.gate.weight_scale": "model-00001-of-00005.safetensors",
-    "model.layers.2.block_sparse_moe.gate.weight_shape": "model-00001-of-00005.safetensors",
+    "model.layers.2.block_sparse_moe.gate.weight": "model-00001-of-00005.safetensors",
     "model.layers.2.input_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.2.self_attn.k_proj.weight_packed": "model-00001-of-00005.safetensors",
@@ -1234,9 +1208,7 @@
     "model.layers.20.block_sparse_moe.experts.7.w3.weight_packed": "model-00004-of-00005.safetensors",
     "model.layers.20.block_sparse_moe.experts.7.w3.weight_scale": "model-00004-of-00005.safetensors",
     "model.layers.20.block_sparse_moe.experts.7.w3.weight_shape": "model-00004-of-00005.safetensors",
-    "model.layers.20.block_sparse_moe.gate.weight_packed": "model-00004-of-00005.safetensors",
-    "model.layers.20.block_sparse_moe.gate.weight_scale": "model-00004-of-00005.safetensors",
-    "model.layers.20.block_sparse_moe.gate.weight_shape": "model-00004-of-00005.safetensors",
+    "model.layers.20.block_sparse_moe.gate.weight": "model-00004-of-00005.safetensors",
     "model.layers.20.input_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.20.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.20.self_attn.k_proj.weight_packed": "model-00004-of-00005.safetensors",
@@ -1323,9 +1295,7 @@
     "model.layers.21.block_sparse_moe.experts.7.w3.weight_packed": "model-00004-of-00005.safetensors",
     "model.layers.21.block_sparse_moe.experts.7.w3.weight_scale": "model-00004-of-00005.safetensors",
     "model.layers.21.block_sparse_moe.experts.7.w3.weight_shape": "model-00004-of-00005.safetensors",
-    "model.layers.21.block_sparse_moe.gate.weight_packed": "model-00004-of-00005.safetensors",
-    "model.layers.21.block_sparse_moe.gate.weight_scale": "model-00004-of-00005.safetensors",
-    "model.layers.21.block_sparse_moe.gate.weight_shape": "model-00004-of-00005.safetensors",
+    "model.layers.21.block_sparse_moe.gate.weight": "model-00004-of-00005.safetensors",
     "model.layers.21.input_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.21.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.21.self_attn.k_proj.weight_packed": "model-00004-of-00005.safetensors",
@@ -1412,9 +1382,7 @@
     "model.layers.22.block_sparse_moe.experts.7.w3.weight_packed": "model-00004-of-00005.safetensors",
     "model.layers.22.block_sparse_moe.experts.7.w3.weight_scale": "model-00004-of-00005.safetensors",
     "model.layers.22.block_sparse_moe.experts.7.w3.weight_shape": "model-00004-of-00005.safetensors",
-    "model.layers.22.block_sparse_moe.gate.weight_packed": "model-00004-of-00005.safetensors",
-    "model.layers.22.block_sparse_moe.gate.weight_scale": "model-00004-of-00005.safetensors",
-    "model.layers.22.block_sparse_moe.gate.weight_shape": "model-00004-of-00005.safetensors",
+    "model.layers.22.block_sparse_moe.gate.weight": "model-00004-of-00005.safetensors",
     "model.layers.22.input_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.22.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.22.self_attn.k_proj.weight_packed": "model-00004-of-00005.safetensors",
@@ -1501,9 +1469,7 @@
     "model.layers.23.block_sparse_moe.experts.7.w3.weight_packed": "model-00004-of-00005.safetensors",
     "model.layers.23.block_sparse_moe.experts.7.w3.weight_scale": "model-00004-of-00005.safetensors",
     "model.layers.23.block_sparse_moe.experts.7.w3.weight_shape": "model-00004-of-00005.safetensors",
-    "model.layers.23.block_sparse_moe.gate.weight_packed": "model-00004-of-00005.safetensors",
-    "model.layers.23.block_sparse_moe.gate.weight_scale": "model-00004-of-00005.safetensors",
-    "model.layers.23.block_sparse_moe.gate.weight_shape": "model-00004-of-00005.safetensors",
+    "model.layers.23.block_sparse_moe.gate.weight": "model-00004-of-00005.safetensors",
     "model.layers.23.input_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.23.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.23.self_attn.k_proj.weight_packed": "model-00004-of-00005.safetensors",
@@ -1590,9 +1556,7 @@
     "model.layers.24.block_sparse_moe.experts.7.w3.weight_packed": "model-00004-of-00005.safetensors",
     "model.layers.24.block_sparse_moe.experts.7.w3.weight_scale": "model-00004-of-00005.safetensors",
     "model.layers.24.block_sparse_moe.experts.7.w3.weight_shape": "model-00004-of-00005.safetensors",
-    "model.layers.24.block_sparse_moe.gate.weight_packed": "model-00004-of-00005.safetensors",
-    "model.layers.24.block_sparse_moe.gate.weight_scale": "model-00004-of-00005.safetensors",
-    "model.layers.24.block_sparse_moe.gate.weight_shape": "model-00004-of-00005.safetensors",
+    "model.layers.24.block_sparse_moe.gate.weight": "model-00004-of-00005.safetensors",
     "model.layers.24.input_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.24.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.24.self_attn.k_proj.weight_packed": "model-00004-of-00005.safetensors",
@@ -1679,9 +1643,7 @@
     "model.layers.25.block_sparse_moe.experts.7.w3.weight_packed": "model-00004-of-00005.safetensors",
     "model.layers.25.block_sparse_moe.experts.7.w3.weight_scale": "model-00004-of-00005.safetensors",
     "model.layers.25.block_sparse_moe.experts.7.w3.weight_shape": "model-00004-of-00005.safetensors",
-    "model.layers.25.block_sparse_moe.gate.weight_packed": "model-00004-of-00005.safetensors",
-    "model.layers.25.block_sparse_moe.gate.weight_scale": "model-00004-of-00005.safetensors",
-    "model.layers.25.block_sparse_moe.gate.weight_shape": "model-00004-of-00005.safetensors",
+    "model.layers.25.block_sparse_moe.gate.weight": "model-00004-of-00005.safetensors",
     "model.layers.25.input_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.25.post_attention_layernorm.weight": "model-00004-of-00005.safetensors",
     "model.layers.25.self_attn.k_proj.weight_packed": "model-00004-of-00005.safetensors",
@@ -1768,9 +1730,7 @@
     "model.layers.26.block_sparse_moe.experts.7.w3.weight_packed": "model-00005-of-00005.safetensors",
     "model.layers.26.block_sparse_moe.experts.7.w3.weight_scale": "model-00005-of-00005.safetensors",
     "model.layers.26.block_sparse_moe.experts.7.w3.weight_shape": "model-00005-of-00005.safetensors",
-    "model.layers.26.block_sparse_moe.gate.weight_packed": "model-00004-of-00005.safetensors",
-    "model.layers.26.block_sparse_moe.gate.weight_scale": "model-00004-of-00005.safetensors",
-    "model.layers.26.block_sparse_moe.gate.weight_shape": "model-00004-of-00005.safetensors",
+    "model.layers.26.block_sparse_moe.gate.weight": "model-00004-of-00005.safetensors",
     "model.layers.26.input_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.26.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.26.self_attn.k_proj.weight_packed": "model-00004-of-00005.safetensors",
@@ -1857,9 +1817,7 @@
     "model.layers.27.block_sparse_moe.experts.7.w3.weight_packed": "model-00005-of-00005.safetensors",
     "model.layers.27.block_sparse_moe.experts.7.w3.weight_scale": "model-00005-of-00005.safetensors",
     "model.layers.27.block_sparse_moe.experts.7.w3.weight_shape": "model-00005-of-00005.safetensors",
-    "model.layers.27.block_sparse_moe.gate.weight_packed": "model-00005-of-00005.safetensors",
-    "model.layers.27.block_sparse_moe.gate.weight_scale": "model-00005-of-00005.safetensors",
-    "model.layers.27.block_sparse_moe.gate.weight_shape": "model-00005-of-00005.safetensors",
+    "model.layers.27.block_sparse_moe.gate.weight": "model-00005-of-00005.safetensors",
     "model.layers.27.input_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.27.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.27.self_attn.k_proj.weight_packed": "model-00005-of-00005.safetensors",
@@ -1946,9 +1904,7 @@
     "model.layers.28.block_sparse_moe.experts.7.w3.weight_packed": "model-00005-of-00005.safetensors",
     "model.layers.28.block_sparse_moe.experts.7.w3.weight_scale": "model-00005-of-00005.safetensors",
     "model.layers.28.block_sparse_moe.experts.7.w3.weight_shape": "model-00005-of-00005.safetensors",
-    "model.layers.28.block_sparse_moe.gate.weight_packed": "model-00005-of-00005.safetensors",
-    "model.layers.28.block_sparse_moe.gate.weight_scale": "model-00005-of-00005.safetensors",
-    "model.layers.28.block_sparse_moe.gate.weight_shape": "model-00005-of-00005.safetensors",
+    "model.layers.28.block_sparse_moe.gate.weight": "model-00005-of-00005.safetensors",
     "model.layers.28.input_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.28.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.28.self_attn.k_proj.weight_packed": "model-00005-of-00005.safetensors",
@@ -2035,9 +1991,7 @@
     "model.layers.29.block_sparse_moe.experts.7.w3.weight_packed": "model-00005-of-00005.safetensors",
     "model.layers.29.block_sparse_moe.experts.7.w3.weight_scale": "model-00005-of-00005.safetensors",
     "model.layers.29.block_sparse_moe.experts.7.w3.weight_shape": "model-00005-of-00005.safetensors",
-    "model.layers.29.block_sparse_moe.gate.weight_packed": "model-00005-of-00005.safetensors",
-    "model.layers.29.block_sparse_moe.gate.weight_scale": "model-00005-of-00005.safetensors",
-    "model.layers.29.block_sparse_moe.gate.weight_shape": "model-00005-of-00005.safetensors",
+    "model.layers.29.block_sparse_moe.gate.weight": "model-00005-of-00005.safetensors",
     "model.layers.29.input_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.29.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.29.self_attn.k_proj.weight_packed": "model-00005-of-00005.safetensors",
@@ -2124,9 +2078,7 @@
     "model.layers.3.block_sparse_moe.experts.7.w3.weight_packed": "model-00001-of-00005.safetensors",
     "model.layers.3.block_sparse_moe.experts.7.w3.weight_scale": "model-00001-of-00005.safetensors",
     "model.layers.3.block_sparse_moe.experts.7.w3.weight_shape": "model-00001-of-00005.safetensors",
-    "model.layers.3.block_sparse_moe.gate.weight_packed": "model-00001-of-00005.safetensors",
-    "model.layers.3.block_sparse_moe.gate.weight_scale": "model-00001-of-00005.safetensors",
-    "model.layers.3.block_sparse_moe.gate.weight_shape": "model-00001-of-00005.safetensors",
+    "model.layers.3.block_sparse_moe.gate.weight": "model-00001-of-00005.safetensors",
     "model.layers.3.input_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.3.self_attn.k_proj.weight_packed": "model-00001-of-00005.safetensors",
@@ -2213,9 +2165,7 @@
     "model.layers.30.block_sparse_moe.experts.7.w3.weight_packed": "model-00005-of-00005.safetensors",
     "model.layers.30.block_sparse_moe.experts.7.w3.weight_scale": "model-00005-of-00005.safetensors",
     "model.layers.30.block_sparse_moe.experts.7.w3.weight_shape": "model-00005-of-00005.safetensors",
-    "model.layers.30.block_sparse_moe.gate.weight_packed": "model-00005-of-00005.safetensors",
-    "model.layers.30.block_sparse_moe.gate.weight_scale": "model-00005-of-00005.safetensors",
-    "model.layers.30.block_sparse_moe.gate.weight_shape": "model-00005-of-00005.safetensors",
+    "model.layers.30.block_sparse_moe.gate.weight": "model-00005-of-00005.safetensors",
     "model.layers.30.input_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.30.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.30.self_attn.k_proj.weight_packed": "model-00005-of-00005.safetensors",
@@ -2302,9 +2252,7 @@
     "model.layers.31.block_sparse_moe.experts.7.w3.weight_packed": "model-00005-of-00005.safetensors",
     "model.layers.31.block_sparse_moe.experts.7.w3.weight_scale": "model-00005-of-00005.safetensors",
     "model.layers.31.block_sparse_moe.experts.7.w3.weight_shape": "model-00005-of-00005.safetensors",
-    "model.layers.31.block_sparse_moe.gate.weight_packed": "model-00005-of-00005.safetensors",
-    "model.layers.31.block_sparse_moe.gate.weight_scale": "model-00005-of-00005.safetensors",
-    "model.layers.31.block_sparse_moe.gate.weight_shape": "model-00005-of-00005.safetensors",
+    "model.layers.31.block_sparse_moe.gate.weight": "model-00005-of-00005.safetensors",
     "model.layers.31.input_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.31.post_attention_layernorm.weight": "model-00005-of-00005.safetensors",
     "model.layers.31.self_attn.k_proj.weight_packed": "model-00005-of-00005.safetensors",
@@ -2391,9 +2339,7 @@
     "model.layers.4.block_sparse_moe.experts.7.w3.weight_packed": "model-00001-of-00005.safetensors",
     "model.layers.4.block_sparse_moe.experts.7.w3.weight_scale": "model-00001-of-00005.safetensors",
     "model.layers.4.block_sparse_moe.experts.7.w3.weight_shape": "model-00001-of-00005.safetensors",
-    "model.layers.4.block_sparse_moe.gate.weight_packed": "model-00001-of-00005.safetensors",
-    "model.layers.4.block_sparse_moe.gate.weight_scale": "model-00001-of-00005.safetensors",
-    "model.layers.4.block_sparse_moe.gate.weight_shape": "model-00001-of-00005.safetensors",
+    "model.layers.4.block_sparse_moe.gate.weight": "model-00001-of-00005.safetensors",
     "model.layers.4.input_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.4.self_attn.k_proj.weight_packed": "model-00001-of-00005.safetensors",
@@ -2480,9 +2426,7 @@
     "model.layers.5.block_sparse_moe.experts.7.w3.weight_packed": "model-00001-of-00005.safetensors",
     "model.layers.5.block_sparse_moe.experts.7.w3.weight_scale": "model-00001-of-00005.safetensors",
     "model.layers.5.block_sparse_moe.experts.7.w3.weight_shape": "model-00001-of-00005.safetensors",
-    "model.layers.5.block_sparse_moe.gate.weight_packed": "model-00001-of-00005.safetensors",
-    "model.layers.5.block_sparse_moe.gate.weight_scale": "model-00001-of-00005.safetensors",
-    "model.layers.5.block_sparse_moe.gate.weight_shape": "model-00001-of-00005.safetensors",
+    "model.layers.5.block_sparse_moe.gate.weight": "model-00001-of-00005.safetensors",
     "model.layers.5.input_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00005.safetensors",
     "model.layers.5.self_attn.k_proj.weight_packed": "model-00001-of-00005.safetensors",
@@ -2569,9 +2513,7 @@
     "model.layers.6.block_sparse_moe.experts.7.w3.weight_packed": "model-00002-of-00005.safetensors",
     "model.layers.6.block_sparse_moe.experts.7.w3.weight_scale": "model-00002-of-00005.safetensors",
     "model.layers.6.block_sparse_moe.experts.7.w3.weight_shape": "model-00002-of-00005.safetensors",
-    "model.layers.6.block_sparse_moe.gate.weight_packed": "model-00001-of-00005.safetensors",
-    "model.layers.6.block_sparse_moe.gate.weight_scale": "model-00001-of-00005.safetensors",
-    "model.layers.6.block_sparse_moe.gate.weight_shape": "model-00001-of-00005.safetensors",
+    "model.layers.6.block_sparse_moe.gate.weight": "model-00001-of-00005.safetensors",
     "model.layers.6.input_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.6.self_attn.k_proj.weight_packed": "model-00001-of-00005.safetensors",
@@ -2658,9 +2600,7 @@
     "model.layers.7.block_sparse_moe.experts.7.w3.weight_packed": "model-00002-of-00005.safetensors",
     "model.layers.7.block_sparse_moe.experts.7.w3.weight_scale": "model-00002-of-00005.safetensors",
     "model.layers.7.block_sparse_moe.experts.7.w3.weight_shape": "model-00002-of-00005.safetensors",
-    "model.layers.7.block_sparse_moe.gate.weight_packed": "model-00002-of-00005.safetensors",
-    "model.layers.7.block_sparse_moe.gate.weight_scale": "model-00002-of-00005.safetensors",
-    "model.layers.7.block_sparse_moe.gate.weight_shape": "model-00002-of-00005.safetensors",
+    "model.layers.7.block_sparse_moe.gate.weight": "model-00002-of-00005.safetensors",
     "model.layers.7.input_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.7.self_attn.k_proj.weight_packed": "model-00002-of-00005.safetensors",
@@ -2747,9 +2687,7 @@
     "model.layers.8.block_sparse_moe.experts.7.w3.weight_packed": "model-00002-of-00005.safetensors",
     "model.layers.8.block_sparse_moe.experts.7.w3.weight_scale": "model-00002-of-00005.safetensors",
     "model.layers.8.block_sparse_moe.experts.7.w3.weight_shape": "model-00002-of-00005.safetensors",
-    "model.layers.8.block_sparse_moe.gate.weight_packed": "model-00002-of-00005.safetensors",
-    "model.layers.8.block_sparse_moe.gate.weight_scale": "model-00002-of-00005.safetensors",
-    "model.layers.8.block_sparse_moe.gate.weight_shape": "model-00002-of-00005.safetensors",
+    "model.layers.8.block_sparse_moe.gate.weight": "model-00002-of-00005.safetensors",
     "model.layers.8.input_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.8.self_attn.k_proj.weight_packed": "model-00002-of-00005.safetensors",
@@ -2836,9 +2774,7 @@
     "model.layers.9.block_sparse_moe.experts.7.w3.weight_packed": "model-00002-of-00005.safetensors",
     "model.layers.9.block_sparse_moe.experts.7.w3.weight_scale": "model-00002-of-00005.safetensors",
     "model.layers.9.block_sparse_moe.experts.7.w3.weight_shape": "model-00002-of-00005.safetensors",
-    "model.layers.9.block_sparse_moe.gate.weight_packed": "model-00002-of-00005.safetensors",
-    "model.layers.9.block_sparse_moe.gate.weight_scale": "model-00002-of-00005.safetensors",
-    "model.layers.9.block_sparse_moe.gate.weight_shape": "model-00002-of-00005.safetensors",
+    "model.layers.9.block_sparse_moe.gate.weight": "model-00002-of-00005.safetensors",
     "model.layers.9.input_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00005.safetensors",
     "model.layers.9.self_attn.k_proj.weight_packed": "model-00002-of-00005.safetensors",

recipe.yaml

@@ -3,5 +3,5 @@ DEFAULT_stage:
     GPTQModifier:
       sequential_update: true
       targets: Linear
-      ignore: [lm_head]
+      ignore: [lm_head, 're:.*block_sparse_moe\.gate']
       scheme: W4A16