Upload folder using huggingface_hub

README.md

@@ -17,8 +17,6 @@ from llmcompressor.transformers.compression.helpers import calculate_offload_dev
 model_id  = "Qwen/Qwen3-30B-A3B"
 model_out = model_id.split("/")[1] + ".w4a16"
 
-recipe = QuantizationModifier(targets="Linear", scheme="W4A16", ignore=["lm_head"], dampening_frac=0.1)
-
 device_map = calculate_offload_device_map(
     model_id, reserve_for_hessians=False, num_gpus=1, torch_dtype="bfloat16"
 )
@@ -33,5 +31,11 @@ model = AutoModelForCausalLM.from_pretrained(
   torch_dtype="bfloat16",
 )
 
+recipe = QuantizationModifier(
+  targets="Linear",
+  scheme="W4A16",
+  ignore=["lm_head", "re:.*mlp.gate$", "re:.*mlp.shared_expert_gate$"],
+)
+
 oneshot(model=model, recipe=recipe, output_dir=model_out)
 ```

config.json

@@ -49,6 +49,54 @@
     "format": "pack-quantized",
     "global_compression_ratio": null,
     "ignore": [
+      "model.layers.0.mlp.gate",
+      "model.layers.1.mlp.gate",
+      "model.layers.2.mlp.gate",
+      "model.layers.3.mlp.gate",
+      "model.layers.4.mlp.gate",
+      "model.layers.5.mlp.gate",
+      "model.layers.6.mlp.gate",
+      "model.layers.7.mlp.gate",
+      "model.layers.8.mlp.gate",
+      "model.layers.9.mlp.gate",
+      "model.layers.10.mlp.gate",
+      "model.layers.11.mlp.gate",
+      "model.layers.12.mlp.gate",
+      "model.layers.13.mlp.gate",
+      "model.layers.14.mlp.gate",
+      "model.layers.15.mlp.gate",
+      "model.layers.16.mlp.gate",
+      "model.layers.17.mlp.gate",
+      "model.layers.18.mlp.gate",
+      "model.layers.19.mlp.gate",
+      "model.layers.20.mlp.gate",
+      "model.layers.21.mlp.gate",
+      "model.layers.22.mlp.gate",
+      "model.layers.23.mlp.gate",
+      "model.layers.24.mlp.gate",
+      "model.layers.25.mlp.gate",
+      "model.layers.26.mlp.gate",
+      "model.layers.27.mlp.gate",
+      "model.layers.28.mlp.gate",
+      "model.layers.29.mlp.gate",
+      "model.layers.30.mlp.gate",
+      "model.layers.31.mlp.gate",
+      "model.layers.32.mlp.gate",
+      "model.layers.33.mlp.gate",
+      "model.layers.34.mlp.gate",
+      "model.layers.35.mlp.gate",
+      "model.layers.36.mlp.gate",
+      "model.layers.37.mlp.gate",
+      "model.layers.38.mlp.gate",
+      "model.layers.39.mlp.gate",
+      "model.layers.40.mlp.gate",
+      "model.layers.41.mlp.gate",
+      "model.layers.42.mlp.gate",
+      "model.layers.43.mlp.gate",
+      "model.layers.44.mlp.gate",
+      "model.layers.45.mlp.gate",
+      "model.layers.46.mlp.gate",
+      "model.layers.47.mlp.gate",
       "lm_head"
     ],
     "kv_cache_scheme": null,

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7057a9d4836c1931f7b026d075710e8171a30edd3312dce25a528540558a3ff7
-size 5001759432
+oid sha256:8367c13e9891801c60ef7f26060d972d1ea79e43fc4e87705106c1fd660e403a
+size 5001524144

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f79e6c9077f477aa405e6fd39e022e55c7915f57c2eae1c57cac5810e26abcf2
-size 5002072232
+oid sha256:c899896142c15978505e045a1935314fefd82781b162075487b181e61cf11e2b
+size 5001803304

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7198692a85de129d879adb808ee2b0dcc22463439c34ff4109d9fc4faf9e1520
-size 5001930568
+oid sha256:85fdd3a5b3b209608b1818cbb45de6e6a9db7fa57c3db3f73641af34a830e3b9
+size 5002084152

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dcd554978c38ba92b4e0d95d3241891667aa1a219dfa7645b6e2dd3a5e6b2c7f
-size 1668650384
+oid sha256:7c9a3ff0bd6804117efaa57d01be938f5c27ba865639c27010d18cb579df36ab
+size 1687667728

model.safetensors.index.json

@@ -1,6 +1,6 @@
 {
   "metadata": {
-    "total_size": 16667508480
+    "total_size": 16686185472
   },
   "weight_map": {
     "lm_head.weight": "model-00004-of-00004.safetensors",
@@ -1158,9 +1158,7 @@
     "model.layers.0.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.0.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.0.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.0.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.0.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.0.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.0.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.0.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -2329,9 +2327,7 @@
     "model.layers.1.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.1.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.1.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.1.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.1.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.1.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.1.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.1.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -3500,9 +3496,7 @@
     "model.layers.10.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.10.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.10.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.10.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.10.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.10.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.10.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.10.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.10.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.10.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -4671,9 +4665,7 @@
     "model.layers.11.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.11.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.11.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.11.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.11.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.11.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.11.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.11.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.11.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.11.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -5842,9 +5834,7 @@
     "model.layers.12.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.12.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.12.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.12.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.12.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.12.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.12.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.12.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.12.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.12.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -6779,33 +6769,33 @@
     "model.layers.13.mlp.experts.75.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.13.mlp.experts.75.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.13.mlp.experts.75.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.76.down_proj.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.76.down_proj.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.76.down_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.76.gate_proj.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.76.gate_proj.weight_scale": "model-00001-of-00004.safetensors",
+    "model.layers.13.mlp.experts.76.down_proj.weight_packed": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.76.down_proj.weight_scale": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.76.down_proj.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.76.gate_proj.weight_packed": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.76.gate_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.76.gate_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.76.up_proj.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.76.up_proj.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.76.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.77.down_proj.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.77.down_proj.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.77.down_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.77.gate_proj.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.77.gate_proj.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.77.gate_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.77.up_proj.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.77.up_proj.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.77.up_proj.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.13.mlp.experts.76.up_proj.weight_packed": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.76.up_proj.weight_scale": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.76.up_proj.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.77.down_proj.weight_packed": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.77.down_proj.weight_scale": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.77.down_proj.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.77.gate_proj.weight_packed": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.77.gate_proj.weight_scale": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.77.gate_proj.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.77.up_proj.weight_packed": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.77.up_proj.weight_scale": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.77.up_proj.weight_shape": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.78.down_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.78.down_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.78.down_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.13.mlp.experts.78.gate_proj.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.78.gate_proj.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.experts.78.gate_proj.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.13.mlp.experts.78.gate_proj.weight_packed": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.78.gate_proj.weight_scale": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.experts.78.gate_proj.weight_shape": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.78.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.78.up_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.13.mlp.experts.78.up_proj.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.13.mlp.experts.78.up_proj.weight_shape": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.79.down_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.79.down_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.79.down_proj.weight_shape": "model-00002-of-00004.safetensors",
@@ -7013,9 +7003,7 @@
     "model.layers.13.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.13.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.13.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.13.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.13.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.13.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.13.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.13.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -8184,9 +8172,7 @@
     "model.layers.14.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.14.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.14.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.14.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.14.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.14.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.14.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.14.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.14.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -9355,9 +9341,7 @@
     "model.layers.15.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.15.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.15.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.15.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.15.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.15.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.15.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.15.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.15.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -10526,9 +10510,7 @@
     "model.layers.16.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.16.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.16.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.16.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.16.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.16.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.16.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.16.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.16.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -11697,9 +11679,7 @@
     "model.layers.17.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.17.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.17.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.17.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.17.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.17.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.17.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.17.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.17.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -12868,9 +12848,7 @@
     "model.layers.18.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.18.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.18.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.18.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.18.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.18.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.18.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.18.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.18.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.18.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -14039,9 +14017,7 @@
     "model.layers.19.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.19.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.19.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.19.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.19.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.19.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.19.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.19.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.19.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.19.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -15210,9 +15186,7 @@
     "model.layers.2.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.2.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.2.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.2.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.2.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.2.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.2.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.2.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -16381,9 +16355,7 @@
     "model.layers.20.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.20.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.20.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.20.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.20.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.20.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.20.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.20.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.20.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.20.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -17552,9 +17524,7 @@
     "model.layers.21.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.21.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.21.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.21.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.21.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.21.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.21.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.21.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.21.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.21.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -18723,9 +18693,7 @@
     "model.layers.22.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.22.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.22.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.22.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.22.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.22.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.22.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.22.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.22.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.22.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -19894,9 +19862,7 @@
     "model.layers.23.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.23.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.23.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.23.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.23.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.23.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.23.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.23.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.23.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.23.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -21065,9 +21031,7 @@
     "model.layers.24.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.24.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.24.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.24.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.24.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.24.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.24.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.24.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.24.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.24.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -22236,9 +22200,7 @@
     "model.layers.25.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.25.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.25.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.25.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.25.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.25.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.25.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.25.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.25.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.25.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -23407,9 +23369,7 @@
     "model.layers.26.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.26.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.26.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.26.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.26.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.26.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.26.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.26.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.26.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.26.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -24578,9 +24538,7 @@
     "model.layers.27.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.27.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.27.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.27.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.27.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.27.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.27.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.27.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.27.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.27.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -25749,9 +25707,7 @@
     "model.layers.28.mlp.experts.99.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.28.mlp.experts.99.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.28.mlp.experts.99.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.28.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.28.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.28.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.28.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.28.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
     "model.layers.28.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.28.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -26083,51 +26039,51 @@
     "model.layers.29.mlp.experts.14.up_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.29.mlp.experts.14.up_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.29.mlp.experts.14.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.15.down_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.15.down_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.15.down_proj.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.29.mlp.experts.15.down_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.15.down_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.15.down_proj.weight_shape": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.15.gate_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.29.mlp.experts.15.gate_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.29.mlp.experts.15.gate_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.15.up_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.15.up_proj.weight_scale": "model-00002-of-00004.safetensors",
+    "model.layers.29.mlp.experts.15.up_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.15.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.15.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.16.down_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.16.down_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.16.down_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.16.gate_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.16.gate_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.16.gate_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.16.up_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.16.up_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.16.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.17.down_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.17.down_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.17.down_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.17.gate_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.17.gate_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.17.gate_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.17.up_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.17.up_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.17.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.18.down_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.18.down_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.18.down_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.18.gate_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.18.gate_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.18.gate_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.18.up_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.18.up_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.18.up_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.19.down_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.19.down_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.19.down_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.19.gate_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.19.gate_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.19.gate_proj.weight_shape": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.19.up_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.19.up_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.19.up_proj.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.29.mlp.experts.16.down_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.16.down_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.16.down_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.16.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.16.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.16.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.16.up_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.16.up_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.16.up_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.17.down_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.17.down_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.17.down_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.17.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.17.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.17.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.17.up_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.17.up_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.17.up_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.18.down_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.18.down_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.18.down_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.18.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.18.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.18.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.18.up_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.18.up_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.18.up_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.19.down_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.19.down_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.19.down_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.19.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.19.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.19.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.19.up_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.19.up_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.19.up_proj.weight_shape": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.2.down_proj.weight_packed": "model-00002-of-00004.safetensors",
     "model.layers.29.mlp.experts.2.down_proj.weight_scale": "model-00002-of-00004.safetensors",
     "model.layers.29.mlp.experts.2.down_proj.weight_shape": "model-00002-of-00004.safetensors",
@@ -26140,12 +26096,12 @@
     "model.layers.29.mlp.experts.20.down_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.20.down_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.20.down_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.29.mlp.experts.20.gate_proj.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.20.gate_proj.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.experts.20.gate_proj.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.29.mlp.experts.20.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.20.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.29.mlp.experts.20.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.20.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.20.up_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.29.mlp.experts.20.up_proj.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.29.mlp.experts.20.up_proj.weight_shape": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.21.down_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.21.down_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.21.down_proj.weight_shape": "model-00003-of-00004.safetensors",
@@ -26920,9 +26876,7 @@
     "model.layers.29.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.29.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.29.mlp.gate.weight_packed": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.gate.weight_scale": "model-00002-of-00004.safetensors",
-    "model.layers.29.mlp.gate.weight_shape": "model-00002-of-00004.safetensors",
+    "model.layers.29.mlp.gate.weight": "model-00002-of-00004.safetensors",
     "model.layers.29.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.29.self_attn.k_norm.weight": "model-00002-of-00004.safetensors",
     "model.layers.29.self_attn.k_proj.weight_packed": "model-00002-of-00004.safetensors",
@@ -28091,9 +28045,7 @@
     "model.layers.3.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.3.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.3.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.3.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.3.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.3.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.3.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.3.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -29262,9 +29214,7 @@
     "model.layers.30.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.30.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.30.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.30.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.30.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.30.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.30.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.30.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.30.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.30.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -30433,9 +30383,7 @@
     "model.layers.31.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.31.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.31.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.31.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.31.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.31.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.31.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.31.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.31.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.31.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -31604,9 +31552,7 @@
     "model.layers.32.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.32.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.32.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.32.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.32.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.32.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.32.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.32.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.32.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.32.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -32775,9 +32721,7 @@
     "model.layers.33.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.33.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.33.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.33.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.33.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.33.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.33.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.33.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.33.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.33.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -33946,9 +33890,7 @@
     "model.layers.34.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.34.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.34.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.34.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.34.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.34.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.34.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.34.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.34.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.34.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -35117,9 +35059,7 @@
     "model.layers.35.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.35.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.35.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.35.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.35.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.35.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.35.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.35.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.35.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.35.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -36288,9 +36228,7 @@
     "model.layers.36.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.36.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.36.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.36.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.36.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.36.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.36.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.36.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.36.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.36.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -37459,9 +37397,7 @@
     "model.layers.37.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.37.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.37.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.37.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.37.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.37.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.37.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.37.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.37.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.37.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -38630,9 +38566,7 @@
     "model.layers.38.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.38.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.38.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.38.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.38.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.38.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.38.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.38.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.38.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.38.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -39801,9 +39735,7 @@
     "model.layers.39.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.39.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.39.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.39.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.39.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.39.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.39.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.39.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.39.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.39.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -40972,9 +40904,7 @@
     "model.layers.4.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.4.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.4.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.4.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.4.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.4.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.4.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.4.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -42143,9 +42073,7 @@
     "model.layers.40.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.40.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.40.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.40.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.40.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.40.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.40.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.40.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.40.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.40.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -43314,9 +43242,7 @@
     "model.layers.41.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.41.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.41.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.41.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.41.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.41.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.41.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.41.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.41.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.41.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -44485,9 +44411,7 @@
     "model.layers.42.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.42.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.42.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.42.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.42.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.42.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.42.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.42.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.42.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.42.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -45656,9 +45580,7 @@
     "model.layers.43.mlp.experts.99.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.43.mlp.experts.99.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.43.mlp.experts.99.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.43.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.43.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.43.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.43.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.43.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
     "model.layers.43.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.43.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -46701,33 +46623,33 @@
     "model.layers.44.mlp.experts.86.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.44.mlp.experts.86.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.44.mlp.experts.86.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.87.down_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.87.down_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.87.down_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.87.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.87.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
+    "model.layers.44.mlp.experts.87.down_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.87.down_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.87.down_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.87.gate_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.87.gate_proj.weight_scale": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.87.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.87.up_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.87.up_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.87.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.88.down_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.88.down_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.88.down_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.88.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.88.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.88.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.88.up_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.88.up_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.88.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.89.down_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.89.down_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.89.down_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.89.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.89.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.89.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.89.up_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.89.up_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.89.up_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.44.mlp.experts.87.up_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.87.up_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.87.up_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.88.down_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.88.down_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.88.down_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.88.gate_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.88.gate_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.88.gate_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.88.up_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.88.up_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.88.up_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.89.down_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.89.down_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.89.down_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.89.gate_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.89.gate_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.89.gate_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.89.up_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.89.up_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.89.up_proj.weight_shape": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.9.down_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.44.mlp.experts.9.down_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.44.mlp.experts.9.down_proj.weight_shape": "model-00003-of-00004.safetensors",
@@ -46737,51 +46659,51 @@
     "model.layers.44.mlp.experts.9.up_proj.weight_packed": "model-00003-of-00004.safetensors",
     "model.layers.44.mlp.experts.9.up_proj.weight_scale": "model-00003-of-00004.safetensors",
     "model.layers.44.mlp.experts.9.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.90.down_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.90.down_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.90.down_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.90.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.90.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.90.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.90.up_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.90.up_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.90.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.91.down_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.91.down_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.91.down_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.91.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.91.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.91.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.91.up_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.91.up_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.91.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.92.down_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.92.down_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.92.down_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.92.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.92.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.92.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.92.up_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.92.up_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.92.up_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.93.down_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.93.down_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.93.down_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.93.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.93.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.93.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.93.up_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.93.up_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.93.up_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.44.mlp.experts.90.down_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.90.down_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.90.down_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.90.gate_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.90.gate_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.90.gate_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.90.up_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.90.up_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.90.up_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.91.down_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.91.down_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.91.down_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.91.gate_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.91.gate_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.91.gate_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.91.up_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.91.up_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.91.up_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.92.down_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.92.down_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.92.down_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.92.gate_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.92.gate_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.92.gate_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.92.up_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.92.up_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.92.up_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.93.down_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.93.down_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.93.down_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.93.gate_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.93.gate_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.93.gate_proj.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.93.up_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.93.up_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.93.up_proj.weight_shape": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.94.down_proj.weight_packed": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.94.down_proj.weight_scale": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.94.down_proj.weight_shape": "model-00004-of-00004.safetensors",
-    "model.layers.44.mlp.experts.94.gate_proj.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.94.gate_proj.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.experts.94.gate_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.44.mlp.experts.94.gate_proj.weight_packed": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.94.gate_proj.weight_scale": "model-00004-of-00004.safetensors",
+    "model.layers.44.mlp.experts.94.gate_proj.weight_shape": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.94.up_proj.weight_packed": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.94.up_proj.weight_scale": "model-00004-of-00004.safetensors",
-    "model.layers.44.mlp.experts.94.up_proj.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.44.mlp.experts.94.up_proj.weight_shape": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.95.down_proj.weight_packed": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.95.down_proj.weight_scale": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.95.down_proj.weight_shape": "model-00004-of-00004.safetensors",
@@ -46827,9 +46749,7 @@
     "model.layers.44.mlp.experts.99.up_proj.weight_packed": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.99.up_proj.weight_scale": "model-00004-of-00004.safetensors",
     "model.layers.44.mlp.experts.99.up_proj.weight_shape": "model-00004-of-00004.safetensors",
-    "model.layers.44.mlp.gate.weight_packed": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.gate.weight_scale": "model-00003-of-00004.safetensors",
-    "model.layers.44.mlp.gate.weight_shape": "model-00003-of-00004.safetensors",
+    "model.layers.44.mlp.gate.weight": "model-00003-of-00004.safetensors",
     "model.layers.44.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
     "model.layers.44.self_attn.k_norm.weight": "model-00003-of-00004.safetensors",
     "model.layers.44.self_attn.k_proj.weight_packed": "model-00003-of-00004.safetensors",
@@ -47998,9 +47918,7 @@
     "model.layers.45.mlp.experts.99.up_proj.weight_packed": "model-00004-of-00004.safetensors",
     "model.layers.45.mlp.experts.99.up_proj.weight_scale": "model-00004-of-00004.safetensors",
     "model.layers.45.mlp.experts.99.up_proj.weight_shape": "model-00004-of-00004.safetensors",
-    "model.layers.45.mlp.gate.weight_packed": "model-00004-of-00004.safetensors",
-    "model.layers.45.mlp.gate.weight_scale": "model-00004-of-00004.safetensors",
-    "model.layers.45.mlp.gate.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.45.mlp.gate.weight": "model-00004-of-00004.safetensors",
     "model.layers.45.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
     "model.layers.45.self_attn.k_norm.weight": "model-00004-of-00004.safetensors",
     "model.layers.45.self_attn.k_proj.weight_packed": "model-00004-of-00004.safetensors",
@@ -49169,9 +49087,7 @@
     "model.layers.46.mlp.experts.99.up_proj.weight_packed": "model-00004-of-00004.safetensors",
     "model.layers.46.mlp.experts.99.up_proj.weight_scale": "model-00004-of-00004.safetensors",
     "model.layers.46.mlp.experts.99.up_proj.weight_shape": "model-00004-of-00004.safetensors",
-    "model.layers.46.mlp.gate.weight_packed": "model-00004-of-00004.safetensors",
-    "model.layers.46.mlp.gate.weight_scale": "model-00004-of-00004.safetensors",
-    "model.layers.46.mlp.gate.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.46.mlp.gate.weight": "model-00004-of-00004.safetensors",
     "model.layers.46.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
     "model.layers.46.self_attn.k_norm.weight": "model-00004-of-00004.safetensors",
     "model.layers.46.self_attn.k_proj.weight_packed": "model-00004-of-00004.safetensors",
@@ -50340,9 +50256,7 @@
     "model.layers.47.mlp.experts.99.up_proj.weight_packed": "model-00004-of-00004.safetensors",
     "model.layers.47.mlp.experts.99.up_proj.weight_scale": "model-00004-of-00004.safetensors",
     "model.layers.47.mlp.experts.99.up_proj.weight_shape": "model-00004-of-00004.safetensors",
-    "model.layers.47.mlp.gate.weight_packed": "model-00004-of-00004.safetensors",
-    "model.layers.47.mlp.gate.weight_scale": "model-00004-of-00004.safetensors",
-    "model.layers.47.mlp.gate.weight_shape": "model-00004-of-00004.safetensors",
+    "model.layers.47.mlp.gate.weight": "model-00004-of-00004.safetensors",
     "model.layers.47.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
     "model.layers.47.self_attn.k_norm.weight": "model-00004-of-00004.safetensors",
     "model.layers.47.self_attn.k_proj.weight_packed": "model-00004-of-00004.safetensors",
@@ -51511,9 +51425,7 @@
     "model.layers.5.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.5.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.5.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.5.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.5.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.5.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.5.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.5.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -52682,9 +52594,7 @@
     "model.layers.6.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.6.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.6.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.6.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.6.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.6.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.6.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.6.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -53853,9 +53763,7 @@
     "model.layers.7.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.7.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.7.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.7.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.7.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.7.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.7.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.7.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -55024,9 +54932,7 @@
     "model.layers.8.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.8.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.8.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.8.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.8.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.8.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.8.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.8.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.8.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.8.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",
@@ -56195,9 +56101,7 @@
     "model.layers.9.mlp.experts.99.up_proj.weight_packed": "model-00001-of-00004.safetensors",
     "model.layers.9.mlp.experts.99.up_proj.weight_scale": "model-00001-of-00004.safetensors",
     "model.layers.9.mlp.experts.99.up_proj.weight_shape": "model-00001-of-00004.safetensors",
-    "model.layers.9.mlp.gate.weight_packed": "model-00001-of-00004.safetensors",
-    "model.layers.9.mlp.gate.weight_scale": "model-00001-of-00004.safetensors",
-    "model.layers.9.mlp.gate.weight_shape": "model-00001-of-00004.safetensors",
+    "model.layers.9.mlp.gate.weight": "model-00001-of-00004.safetensors",
     "model.layers.9.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
     "model.layers.9.self_attn.k_norm.weight": "model-00001-of-00004.safetensors",
     "model.layers.9.self_attn.k_proj.weight_packed": "model-00001-of-00004.safetensors",

recipe.yaml

@@ -1,6 +1,6 @@
 default_stage:
   default_modifiers:
     QuantizationModifier:
-      ignore: [lm_head]
+      ignore: [lm_head, 're:.*mlp.gate$', 're:.*mlp.shared_expert_gate$']
       targets: [Linear]
       scheme: W4A16

tokenizer.json

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:574de68a0f63f2004784a421c7d42c2b2786c05cb38542d2ed3525757a1f7fde
-size 11422932
+oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
+size 11422654