777

requirements.txt

@@ -3,12 +3,11 @@
 torch==2.2.1
 torchvision==0.17.1
 torchaudio==2.2.1
-transformers==4.34.1
+transformers==4.51.3
 sentence-transformers==2.7.0
 diffusers==0.26.3
 decord==0.6.0
-huggingface_hub==0.22.1
-tokenizers==0.21.0
+huggingface_hub==0.30.2
 peft==0.11.1
 Pillow
 einops

robohusky/model/modeling_husky_embody2.py

@@ -39,13 +39,13 @@ from transformers.utils import (
     add_start_docstrings_to_model_forward,
     logging,
     replace_return_docstrings,
-    is_flash_attn_available
+    is_flash_attn_2_available
 )
 from transformers import AutoModelForCausalLM, GenerationConfig
 
 from .configuration_husky import HuskyConfig, HuskyQFormerConfig, HuskyVisionConfig
 
-if is_flash_attn_available():
+if is_flash_attn_2_available():
     from flash_attn import flash_attn_func
     from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input  # noqa