processor load

2849128d6acad1c139d1c8963bf054ec3f16316a · 2024-07-28 01:03:34 +0530 · Varshith

Files touched

M exo/inference/mlx/sharded_utils.py

Diff

commit 2849128d6acad1c139d1c8963bf054ec3f16316a
Author: Varshith <varshith.bathini@sprinklr.com>
Date:   Sun Jul 28 01:03:34 2024 +0530

    processor load
---
 exo/inference/mlx/sharded_utils.py | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/exo/inference/mlx/sharded_utils.py b/exo/inference/mlx/sharded_utils.py
index 0bc8efe6..2964d544 100644
--- a/exo/inference/mlx/sharded_utils.py
+++ b/exo/inference/mlx/sharded_utils.py
@@ -235,10 +235,10 @@ async def load_shard(
         model = apply_lora_layers(model, adapter_path)
         model.eval()
 
-    # TODO: figure out a better way
-    if "llama" in str(model_path):
-        tokenizer = load_tokenizer(model_path, tokenizer_config)
-        return model, tokenizer
-    elif "llava" in str(model_path):
+    # TODO: figure out a generic solution
+    if model.model_type == "llava":
         processor = AutoProcessor.from_pretrained(model_path)
-        return model, processor
\ No newline at end of file
+        return model, processor
+    else:
+        tokenizer = load_tokenizer(model_path, tokenizer_config)
+        return model, tokenizer
\ No newline at end of file

← 54993995 conflicts · back to Exo · try/except for decode, #75 5eafd5a3 →