RuntimeError: a leaf Variable that requires grad is being used in an in-place operation

🤗Transformers
1

Hi there, i’am trying to finetune JAIS-13B and LLM for arabic on a custom summarization dataset and when i run trainer.train() i got the ERROR below ! any help plz

RuntimeError Traceback (most recent call last)
in <cell line: 2>()
1 trainer.deprecated=True
----> 2 trainer.train()

21 frames
~/.cache/huggingface/modules/transformers_modules/inception-mbzuai/jais-13b-chat/96080d1c163804428c4792b8618c2d39661e9d7f/modeling_jais.py in forward(self, input_ids, past_key_values, attention_mask, token_type_ids, position_ids, head_mask, inputs_embeds, encoder_hidden_states, encoder_attention_mask, use_cache, output_attentions, output_hidden_states, return_dict)
867 else:
868 hidden_states = inputs_embeds
→ 869 hidden_states *= torch.tensor(
870 float(self.embeddings_scale), dtype=hidden_states.dtype, device=hidden_states.device
871 )

RuntimeError: a leaf Variable that requires grad is being used in an in-place operation.

2 Likes
2

@sgugger any help from your side plz !?

3

You can change the code of modeling_jais.py from:
hidden_states *= torch.tensor(
float(self.embeddings_scale), dtype=hidden_states.dtype, device=hidden_states.device
)
to:
scale_factor_hidden = torch.tensor(float(self.embeddings_scale), dtype=hidden_states.dtype, device=hidden_states.device)
hidden_states = hidden_states * scale_factor_hidden

1 Like
4

i will make a pull request and hopefully the authors would accept it, Thank you for your help

1 Like
6

i remember it worked ! otherwise you can check the Asas AI version : https://huggingface.co/asas-ai/jais-13b-chat-8bit