Add MLX int8 variants for 135M, 360M and 1.7B

.gitattributes

@@ -49,3 +49,6 @@ smolLm-2-1.7B/quantized/smolLm2_1_7B_8da4w.pte filter=lfs diff=lfs merge=lfs -te
 135m/xnnpack/smollm2_135m_xnnpack_8da8w.pte filter=lfs diff=lfs merge=lfs -text
 360m/xnnpack/smollm2_360m_xnnpack_8da8w.pte filter=lfs diff=lfs merge=lfs -text
 1_7b/xnnpack/smollm2_1_7b_xnnpack_8da8w.pte filter=lfs diff=lfs merge=lfs -text
+135m/mlx/smollm2_135m_mlx_int8.pte filter=lfs diff=lfs merge=lfs -text
+360m/mlx/smollm2_360m_mlx_int8.pte filter=lfs diff=lfs merge=lfs -text
+1_7b/mlx/smollm2_1_7b_mlx_int8.pte filter=lfs diff=lfs merge=lfs -text

135m/mlx/config.json

@@ -0,0 +1,61 @@
+{
+  "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
+  "model": "smollm2",
+  "family": "smollm",
+  "capabilities": [
+    "text-generation"
+  ],
+  "backend": "mlx",
+  "license": "apache-2.0",
+  "variants": [
+    {
+      "file": "smollm2_135m_mlx_int8.pte",
+      "precision": "8w",
+      "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,
+                49152
+              ],
+              "dtype": "bfloat16"
+            }
+          ]
+        },
+        "get_bos_id": 1,
+        "get_eos_ids": [
+          2
+        ],
+        "get_max_context_len": 2048,
+        "get_max_seq_len": 2048,
+        "get_n_layers": 30,
+        "get_vocab_size": 49152,
+        "use_kv_cache": true,
+        "use_sdpa_with_kv_cache": true
+      }
+    }
+  ],
+  "size": "135m"
+}

135m/mlx/smollm2_135m_mlx_int8.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b523149edadbf5c8511fc5b7d48e5edd00f3980520414d847b5c95a80491f66
+size 195630848

1_7b/mlx/config.json

@@ -0,0 +1,61 @@
+{
+  "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
+  "model": "smollm2",
+  "family": "smollm",
+  "capabilities": [
+    "text-generation"
+  ],
+  "backend": "mlx",
+  "license": "apache-2.0",
+  "variants": [
+    {
+      "file": "smollm2_1_7b_mlx_int8.pte",
+      "precision": "8w",
+      "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,
+                49152
+              ],
+              "dtype": "bfloat16"
+            }
+          ]
+        },
+        "get_bos_id": 1,
+        "get_eos_ids": [
+          2
+        ],
+        "get_max_context_len": 2048,
+        "get_max_seq_len": 2048,
+        "get_n_layers": 24,
+        "get_vocab_size": 49152,
+        "use_kv_cache": true,
+        "use_sdpa_with_kv_cache": true
+      }
+    }
+  ],
+  "size": "1_7b"
+}

1_7b/mlx/smollm2_1_7b_mlx_int8.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e591dbcd44b2c42ddd0a63746e771c0072a235d8da3b21e68ff7e3a5188d1687
+size 1966486016

360m/mlx/config.json

@@ -0,0 +1,61 @@
+{
+  "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
+  "model": "smollm2",
+  "family": "smollm",
+  "capabilities": [
+    "text-generation"
+  ],
+  "backend": "mlx",
+  "license": "apache-2.0",
+  "variants": [
+    {
+      "file": "smollm2_360m_mlx_int8.pte",
+      "precision": "8w",
+      "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,
+                49152
+              ],
+              "dtype": "bfloat16"
+            }
+          ]
+        },
+        "get_bos_id": 1,
+        "get_eos_ids": [
+          2
+        ],
+        "get_max_context_len": 2048,
+        "get_max_seq_len": 2048,
+        "get_n_layers": 32,
+        "get_vocab_size": 49152,
+        "use_kv_cache": true,
+        "use_sdpa_with_kv_cache": true
+      }
+    }
+  ],
+  "size": "360m"
+}

360m/mlx/smollm2_360m_mlx_int8.pte

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3574e1c68f7d1becf6dff573e638e0909e884e9ea74f2d05db02ab49e67ce42d
+size 467832064