cast to fp16 and change file path (#1955)

Author: qhzhuang

llm/inference/janus_pro/understanding.py

@@ -28,7 +28,7 @@
 )
 print('loaded processor and ckpt ')
 question = 'describe this image'
-image = "/home/HwHiAiUser/janus-pro-mindspore/inpain_model_cat.png"
+image = "./inpain_model_cat.png"
 conversation = [
     {
         "role": "<|User|>",

mindnlp/core/ops/array.py

@@ -21,8 +21,8 @@ def argwhere(input):
 # cat
 has_cat = hasattr(mindspore.mint, 'cat')
 def cat(tensors, dim=0):
-    if use_pyboost() and has_cat:
-        return mindspore.mint.cat(tensors, dim)
+    # if use_pyboost() and has_cat:
+    #     return mindspore.mint.cat(tensors, dim)
     return ops.cat(tensors, dim)
 
 # concat

mindnlp/transformers/cache_utils.py

@@ -364,7 +364,10 @@ def update(
         # Update the number of seen tokens
         if layer_idx == 0:
             self._seen_tokens += key_states.shape[-2]
-
+        if key_states.dtype!=mindspore.float16:
+            key_states = key_states.astype(mindspore.float16)
+        if key_states.dtype!=mindspore.float16:
+            value_states = value_states.astype(mindspore.float16)
         # Update the cache
         if len(self.key_cache) <= layer_idx:
             self.key_cache.append(key_states)
@@ -375,7 +378,8 @@ def update(
             self.key_cache[layer_idx] = key_states
             self.value_cache[layer_idx] = value_states
         else:
-            self.key_cache[layer_idx] = ops.cat([self.key_cache[layer_idx], key_states], dim=-2)
+            self.key_cache[layer_idx] = ops.cat(
+                [self.key_cache[layer_idx].astype(mindspore.float16), key_states.astype(mindspore.float16)], dim=-2)
             self.value_cache[layer_idx] = ops.cat([self.value_cache[layer_idx], value_states], dim=-2)
 
         return self.key_cache[layer_idx], self.value_cache[layer_idx]