Junlei Zhang comments

Results 38 comments of


                                            Junlei Zhang

Would you release the code of finetuning BLIP-2-T5?

Thank you for your work. Would you release code of how to reimplement the results o BLIP-2-T5 in your paper?

Error installation

same error

Could I do multi-thread evaluation?

Hello, do you have any advise on how to set up multiple dockers for the same website. For example, we can set up 10 shoping weisite with different port. So...

Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1580, OpType=ALLREDUCE, NumelIn=466119168, NumelOut=466119168, Timeout(ms)=600000) ran for 600004 milliseconds before timing out.

训练开始后，一会就会报watch dog错误

Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1580, OpType=ALLREDUCE, NumelIn=466119168, NumelOut=466119168, Timeout(ms)=600000) ran for 600004 milliseconds before timing out.

请问有什么解决方案么@hiyouga

Watchdog caught collective operation timeout: WorkNCCL(SeqNum=1580, OpType=ALLREDUCE, NumelIn=466119168, NumelOut=466119168, Timeout(ms)=600000) ran for 600004 milliseconds before timing out.

请问有什么解决方案么

Many invalid links in the shopping_admin env

I have tried both the http://ec2-3-131-244-37.us-east-2.compute.amazonaws.com:7780/admin/admin/dashboard/ and my host website. They have the same issue until now.

OSError: [Errno 7] Argument list too long

### model model_name_or_path: /storage/home/westlakeLab/zhangjunlei/models/Qwen2-VL-7B-Instruct/Qwen2-VL-7B-Instruct ### method stage: sft do_train: true finetuning_type: full deepspeed: examples/deepspeed/ds_z2_config.json ### dataset dataset: phone_web_0131_fix_merge_1500_wait_scroll template: qwen2_vl cutoff_len: 8192 max_samples: 1000000000 overwrite_cache: true preprocessing_num_workers: 1 dataset_dir: /storage/home/westlakeLab/zhangjunlei/dataset...