Upload config.json with huggingface_hub

config.json

@@ -9,7 +9,12 @@
   "compression_config": {
     "config_groups": {
       "group_0": {
-        "input_activations": {
+        "input_activations": null,
+        "output_activations": null,
+        "targets": [
+          "Linear"
+        ],
+        "weights": {
           "block_structure": null,
           "dynamic": false,
           "group_size": null,
@@ -19,22 +24,26 @@
           "strategy": "channel",
           "symmetric": true,
           "type": "int"
-        },
-        "output_activations": null,
-        "targets": [
-          "Linear"
-        ],
-        "weights": {
+        }
+      },
+      "kv_cache": {
+        "input_activations": null,
+        "output_activations": {
           "block_structure": null,
           "dynamic": false,
           "group_size": null,
           "num_bits": 8,
           "observer": "minmax",
           "observer_kwargs": {},
-          "strategy": "channel",
+          "strategy": "tensor",
           "symmetric": true,
-          "type": "int"
-        }
+          "type": "float"
+        },
+        "targets": [
+          "re:.*k_proj",
+          "re:.*v_proj"
+        ],
+        "weights": null
       }
     },
     "format": "pack-quantized",