Upload folder using huggingface_hub

config.json

@@ -90,7 +90,7 @@
     "quantization_status": "compressed",
     "sparsity_config": {},
     "transform_config": {},
-    "version": "0.17.2.a20260724"
+    "version": "0.17.2.a20260727"
   },
   "rms_norm_eps": 1e-05,
   "rope_parameters": {

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ab4225d762be9931307e54b7c4609ea4ceffaa3e5e97500ca10f443657776438
+oid sha256:705802da34391b8d69991d29199b472b0b815183cee13fe268894f6d00046819
 size 1131284624

recipe.yaml

@@ -1,6 +1,7 @@
 quant_stage:
   quant_modifiers:
     AWQModifier:
+      requires_calibration_data: true
       mappings:
       - smooth_layer: re:.*input_layernorm$
         balance_layers: ['re:.*q_proj$', 're:.*k_proj$', 're:.*v_proj$']
@@ -12,11 +13,13 @@ quant_stage:
       ignore: [lm_head]
       scheme: W4A16
       bypass_divisibility_checks: false
+      requires_calibration_data: false
     GPTQModifier:
       targets: ['re:.*mlp\.(down|gate|up)_proj$']
       ignore: [lm_head]
       scheme: W8A8
       bypass_divisibility_checks: false
+      requires_calibration_data: true
       block_size: 128
       dampening_frac: 0.001
       offload_hessians: false