add deepseek-coder-v2.5

exo/models.py

@@ -26,8 +26,9 @@ model_base_shards = {
   ### mistral
   "mistral-nemo": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/Mistral-Nemo-Instruct-2407-4bit", start_layer=0, end_layer=0, n_layers=40),},
   "mistral-large": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/Mistral-Large-Instruct-2407-4bit", start_layer=0, end_layer=0, n_layers=88),},
-  ### deepseek v2
+  ### deepseek
   "deepseek-coder-v2-lite": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/DeepSeek-Coder-V2-Lite-Instruct-4bit-mlx", start_layer=0, end_layer=0, n_layers=27),},
+  "deepseek-coder-v2.5": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/DeepSeek-V2.5-MLX-AQ4_1_64", start_layer=0, end_layer=0, n_layers=60),},
   ### llava
   "llava-1.5-7b-hf": {"MLXDynamicShardInferenceEngine": Shard(model_id="llava-hf/llava-1.5-7b-hf", start_layer=0, end_layer=0, n_layers=32),},
   ### qwen

tinychat/examples/tinychat/index.html

@@ -36,6 +36,7 @@
 <option value="mistral-nemo">Mistral Nemo</option>
 <option value="mistral-large">Mistral Large</option>
 <option value="deepseek-coder-v2-lite">Deepseek Coder V2 Lite</option>
+<option value="deepseek-coder-v2.5">Deepseek Coder V2.5</option>
 <option value="llava-1.5-7b-hf">LLaVa 1.5 7B (Vision Model)</option>
 <option value="qwen-2.5-7b">Qwen 2.5 7B</option>
 <option value="qwen-2.5-math-7b">Qwen 2.5 7B (Math)</option>