set "use_flash_attn" to false

config.json

@@ -17,7 +17,7 @@
     "architectures": [
       "InternLM2ForCausalLM"
     ],
-    "attn_implementation": "flash_attention_2",
+    "attn_implementation": "eager",
     "auto_map": {
       "AutoConfig": "configuration_internlm2.InternLM2Config",
       "AutoModel": "modeling_internlm2.InternLM2ForCausalLM",
@@ -138,6 +138,6 @@
     "torch_dtype": "bfloat16",
     "transformers_version": "4.37.2",
     "use_bfloat16": true,
-    "use_flash_attn": true
+    "use_flash_attn": false
   }
 }