define qwen2.5-3b

exo/models.py

@@ -69,7 +69,19 @@ model_cards = {
   "qwen-2.5-coder-7b": { "layers": 28, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/Qwen2.5-Coder-7B-Instruct-4bit", }, },
   "qwen-2.5-coder-14b": { "layers": 48, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/Qwen2.5-Coder-14B-Instruct-4bit", }, },
   "qwen-2.5-coder-32b": { "layers": 64, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/Qwen2.5-Coder-32B-Instruct-4bit", }, },
-  "qwen-2.5-7b": { "layers": 28, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/Qwen2.5-7B-Instruct-4bit", }, },
+  "qwen-2.5-3b": {
+    "layers": 36,
+    "repo": {
+      "MLXDynamicShardInferenceEngine": "mlx-community/Qwen2.5-3B-Instruct-4bit",
+      "TinygradDynamicShardInferenceEngine": "unsloth/Qwen2.5-3B-Instruct",
+    },
+  },
+  "qwen-2.5-7b": {
+    "layers": 28,
+    "repo": {
+      "MLXDynamicShardInferenceEngine": "mlx-community/Qwen2.5-7B-Instruct-4bit",
+    },
+  },
   "qwen-2.5-math-7b": { "layers": 28, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/Qwen2.5-Math-7B-Instruct-4bit", }, },
   "qwen-2.5-14b": { "layers": 48, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/Qwen2.5-14B-Instruct-4bit", }, },
   "qwen-2.5-72b": { "layers": 80, "repo": { "MLXDynamicShardInferenceEngine": "mlx-community/Qwen2.5-72B-Instruct-4bit", }, },