fix GenerationMixin ImportError in transformers main branch (#12)

- fix GenerationMixin ImportError in transformers main branch (4ba2441d1ca881cf9cb3e687184d490d7687c1ca)
- revert (2109fd69a52f3ff0aaf75c2f13b932a566bd1190)

config.json

@@ -19,7 +19,8 @@
     "merge_kernel_size": [
       2,
       2
-    ]
+    ],
+    "torch_dtype": "bfloat16"
   },
   "text_config": {
     "vocab_size": 163840,

modeling_kimi_vl.py

@@ -55,10 +55,8 @@ import torch.distributed as dist
 from torch.nn import CrossEntropyLoss
 from transformers.activations import GELUActivation, ACT2FN, PytorchGELUTanh
 from transformers.cache_utils import Cache, DynamicCache
-from transformers.modeling_utils import (
-    PreTrainedModel,
-    GenerationMixin,
-)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.generation.utils import GenerationMixin
 from transformers.models.llava.modeling_llava import LlavaCausalLMOutputWithPast
 from transformers.modeling_attn_mask_utils import _prepare_4d_causal_attention_mask
 from transformers.modeling_outputs import (