Update Qwen 3 0.6B MLX model to int8 (v0.11.0)

.gitattributes

@@ -49,3 +49,4 @@ qwen-3-4B/quantized/qwen3_4b_8da4w.pte filter=lfs diff=lfs merge=lfs -text
 0_6b/mlx/qwen_3_0_6b_mlx_int4.pte filter=lfs diff=lfs merge=lfs -text
 1_7b/mlx/qwen_3_1_7b_mlx_int4.pte filter=lfs diff=lfs merge=lfs -text
 4b/mlx/qwen_3_4b_mlx_int4.pte filter=lfs diff=lfs merge=lfs -text
+0_6b/mlx/qwen_3_0_6b_mlx_int8.pte filter=lfs diff=lfs merge=lfs -text

0_6b/mlx/config.json

@@ -9,8 +9,8 @@
   "license": "apache-2.0",
   "variants": [
     {
-      "file": "qwen_3_0_6b_mlx_int4.pte",
-      "precision": "4w",
+      "file": "qwen_3_0_6b_mlx_int8.pte",
+      "precision": "8w",
       "quantized": true,
       "default": true,
       "methods": {
@@ -40,7 +40,7 @@
                 1,
                 151936
               ],
-              "dtype": "bfloat16"
+              "dtype": "float32"
             }
           ]
         },
@@ -48,7 +48,7 @@
         "get_eos_ids": [
           151645
         ],
-        "get_max_context_len": 2048,
+        "get_max_context_len": 8192,
         "get_max_seq_len": 2048,
         "get_n_layers": 28,
         "get_vocab_size": 151936,

0_6b/mlx/{qwen_3_0_6b_mlx_int4.pte → qwen_3_0_6b_mlx_int8.pte}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b22d3792afbd183ddb18aa4acc795dc2977d384dfc6d64ddc0360bc996dc12af
-size 628165248
+oid sha256:b7105828b2744c830770929f924c505161ff859a96481ff4f2b8f33b8cf56525
+size 926157568