Add MLX int4 variant

.gitattributes

@@ -38,3 +38,4 @@ original/phi-4-mini_bf16.pte filter=lfs diff=lfs merge=lfs -text
 quantized/phi-4-mini_8da4w.pte filter=lfs diff=lfs merge=lfs -text
 xnnpack/phi_4_mini_xnnpack_bf16.pte filter=lfs diff=lfs merge=lfs -text
 xnnpack/phi_4_mini_xnnpack_8da4w.pte filter=lfs diff=lfs merge=lfs -text
+mlx/phi_4_mini_mlx_int4.pte filter=lfs diff=lfs merge=lfs -text

mlx/config.json

@@ -0,0 +1,61 @@
+{
+  "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
+  "model": "phi_4_mini",
+  "family": "phi",
+  "capabilities": [
+    "text-generation"
+  ],
+  "backend": "mlx",
+  "license": "mit",
+  "variants": [
+    {
+      "file": "phi_4_mini_mlx_int4.pte",
+      "precision": "4w",
+      "quantized": true,
+      "default": true,
+      "methods": {
+        "enable_dynamic_shape": true,
+        "forward": {
+          "inputs": [
+            {
+              "shape": [
+                1,
+                {
+                  "min": 1,
+                  "max": 2047
+                }
+              ],
+              "dtype": "int64"
+            },
+            {
+              "shape": [
+                1
+              ],
+              "dtype": "int64"
+            }
+          ],
+          "outputs": [
+            {
+              "shape": [
+                1,
+                200064
+              ],
+              "dtype": "bfloat16"
+            }
+          ]
+        },
+        "get_bos_id": 199999,
+        "get_eos_ids": [
+          200020,
+          199999
+        ],
+        "get_max_context_len": 2048,
+        "get_max_seq_len": 2048,
+        "get_n_layers": 32,
+        "get_vocab_size": 200064,
+        "use_kv_cache": true,
+        "use_sdpa_with_kv_cache": true
+      }
+    }
+  ]
+}

mlx/phi_4_mini_mlx_int4.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:135d86947cab70e2e6f72a915afba1cd797baf5cf3d1babf591ad36c3bb0bd04
+size 3267644544