Fix: AttributeError when `input_ids` is None during multimodal LLM training

support transformers 4.44

fix

transformer 4.42.4 upgrade

Merge branch 'main' of https://huggingface.co/THUDM/glm-4-9b-chat

Fix past_key_values

fix import error of flash attn

flash_attention_2 (#51)

Add output_attentions for ChatGLMForSequenceClassification

Merge branch 'main' of https://huggingface.co/THUDM/glm-4-9b-chat

Fix default dtype for classification head

disable `torch.jit` when using Ascend NPUs (#20)

Fix classification model

fix tensor shape error when torch version less than 2 #4 (#9)

fixes https://github.com/THUDM/GLM-4/issues/22 (#5)

first commit