Browse Source

add smollm2 models

Alex Cheema 1 year ago
parent
commit
57c88f6bb7
1 changed files with 5 additions and 0 deletions
  1. 5 0
      exo/models.py

+ 5 - 0
exo/models.py

@@ -1,6 +1,11 @@
 from exo.inference.shard import Shard
 from exo.inference.shard import Shard
 
 
 model_base_shards = {
 model_base_shards = {
+  # smollm
+  "smollm2-135m": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/SmolLM2-135M-Instruct", start_layer=0, end_layer=0, n_layers=30),},
+  "smollm2-360m": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/SmolLM2-360M-Instruct", start_layer=0, end_layer=0, n_layers=32),},
+  "smollm2-1.7b": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/SmolLM2-1.7B-Instruct", start_layer=0, end_layer=0, n_layers=24),},
+
   ### llama
   ### llama
   "llama-3.2-1b": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/Llama-3.2-1B-Instruct-4bit", start_layer=0, end_layer=0, n_layers=16),},
   "llama-3.2-1b": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/Llama-3.2-1B-Instruct-4bit", start_layer=0, end_layer=0, n_layers=16),},
   "llama-3.2-3b": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/Llama-3.2-3B-Instruct-4bit", start_layer=0, end_layer=0, n_layers=28),},
   "llama-3.2-3b": {"MLXDynamicShardInferenceEngine": Shard(model_id="mlx-community/Llama-3.2-3B-Instruct-4bit", start_layer=0, end_layer=0, n_layers=28),},