Update requirements.txt

[RDXiaoLu]

solve some requirement problems.... such as peft version.... accelerate version.

I can run success...

[kellenf]

@RDXiaoLu 您好，我跑实验的时候看到有问题：
Some weights of the model checkpoint at ckpts/pretrain_qformer/ were not used when initializing LlavaLlamaForCausalLM: ['model.vision_tower.vision_to
wer.blocks.17.attn.rope.freqs_sin', 'model.vision_tower.vision_tower.blocks.22.mlp.ffn_ln.bias', 'model.mm_projector.query_decoder.layers.4.transfor
mer_layers.5.bias', 'model.vision_tower.vision_tower.blocks.19.attn.rope.freqs_sin', 'model.vision_tower.vision_tower.blocks.18.attn.q_bias', 'model.
vision_tower.vision_tower.blocks.1.attn.proj.weight', 'model.vision_tower.vision_tower.blocks.8.attn.v_proj.weight', 'model.vision_tower.vision_tower
.blocks.23.attn.q_bias', 'model.vision_tower.vision_tower.blocks.13.mlp.w2.bias', 'model.vision_tower.vision_tower.blocks.18.mlp.w2.bias', 'model.mm
projector.query_decoder._layers.3.transformer_layers.0.attn.out_proj.weight', 'model.mm_projector.query_decoder._layers.4.transformer_layers.4._layer
s.0.bias', 'model.vision_tower.vision_tower.blocks.12.attn.proj.bias', 'model.mm_projector.query_decoder._layers.5.transformer_layers.2.attn.out_proj
.bias', 'model.vision_tower.vision_tower.blocks.
你那边有这样的情况吗？

[RDXiaoLu]

@RDXiaoLu 您好，我跑实验的时候看到有问题： Some weights of the model checkpoint at ckpts/pretrain_qformer/ were not used when initializing LlavaLlamaForCausalLM: ['model.vision_tower.vision_to wer.blocks.17.attn.rope.freqs_sin', 'model.vision_tower.vision_tower.blocks.22.mlp.ffn_ln.bias', 'model.mm_projector.query_decoder.layers.4.transfor mer_layers.5.bias', 'model.vision_tower.vision_tower.blocks.19.attn.rope.freqs_sin', 'model.vision_tower.vision_tower.blocks.18.attn.q_bias', 'model. vision_tower.vision_tower.blocks.1.attn.proj.weight', 'model.vision_tower.vision_tower.blocks.8.attn.v_proj.weight', 'model.vision_tower.vision_tower .blocks.23.attn.q_bias', 'model.vision_tower.vision_tower.blocks.13.mlp.w2.bias', 'model.vision_tower.vision_tower.blocks.18.mlp.w2.bias', 'model.mm projector.query_decoder._layers.3.transformer_layers.0.attn.out_proj.weight', 'model.mm_projector.query_decoder._layers.4.transformer_layers.4._layer s.0.bias', 'model.vision_tower.vision_tower.blocks.12.attn.proj.bias', 'model.mm_projector.query_decoder._layers.5.transformer_layers.2.attn.out_proj .bias', 'model.vision_tower.vision_tower.blocks. 你那边有这样的情况吗？

Hello, I guess it may be that the file under your /pretrain_qformer has not been fully downloaded, and the eva02_petr_proj.pth file needs to be placed under/ckpts.