microsoft / simxns Goto Github PK

I am writing to inform you that I have been facing an issue with downloading the checkpoints for the past 4 days. The download stops when it reaches 6-7 GB in size, and I am not sure if others are facing the same problem.

Could you please look into this and let me know if there is a solution or if anyone else is encountering a similar issue?

According to the code
random.choices(neg_candidates, weights=neg_scores, k=num_hard_negatives)

If it's true, why it is better than without replacement?

Hello,this is a really impressive work.
I've got some problems with the checkpoint.

you said that you initialized the model with bert, and I'm confused about which two layers you used to initialized the other heads,like c_head,query_head

FileNotFoundError: [Errno 2] No such file or directory: '/home/socialab/Desktop/Kailash/SimXNS-main/SimANS/output/co_training_tq_SimANS_test/log.txt'

Datasets and ckpts are not accessible now!!!

Traceback (most recent call last):
File "co_training/co_training_marco_generate.py", line 401, in
main()
File "co_training/co_training_marco_generate.py", line 392, in main
get_new_dataset(args, model, global_step, renew_tools)
File "co_training/co_training_marco_generate.py", line 354, in get_new_dataset
renew_tools.get_question_topk(train_q, train_q_embed, train_q_embed2id, ground_truth_path,
File "/SimXNS/SimANS/co_training/co_training_generate.py", line 439, in get_question_topk
train_pos_qp, train_pos_qp_add = load_pos_examples(mode)
File "/SimXNS/SimANS/co_training/co_training_generate.py", line 141, in load_pos_examples
with open(file_add) as inp:
FileNotFoundError: [Errno 2] No such file or directory: 'data/MS-Pas/qrels.train.addition.tsv'

@lx865712528

In project ALLIES，How can I get the "passage_embedding.pb" and "passage_embedding2id.pb" ?

In BLOB, I did not see the relevant resources, if I need to embeding by myself, in the source code I did not see the relevant part of the content?

Looking forward to your reply !

why there is no def SimANS(pos_pair, neg_pairs_list) in the code?

OSError: Can't load the configuration of 'ckpt/MS-Doc/adore-star'. If you were trying to load it from 'https://huggingface.co/models', make sure you don't have a local directory with the same name. Otherwise, make sure 'ckpt/MS-Doc/adore-star' is the correct path to a directory containing a config.json file
ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: 1) local_rank: 0 (pid: 18826) of binary: /home/socialab/miniconda3/envs/IR_SimANS-env/bin/python

orch.cuda.set_device(args.local_rank)
File "/home/socialab/miniconda3/envs/IR_SimANS-env/lib/python3.10/site-packages/torch/cuda/init.py", line 314, in set_device
torch._C._cuda_setDevice(device)
RuntimeError: CUDA error: invalid device ordinal

Hi, 麻烦问下。

对Fig.1的梯度均值，统计的是整体loss对模型的梯度向量的L2范数吗？
对Fig.1的梯度方差，统计的是[正样本doc的loss的梯度的L2范数, 负样本doc的loss梯度的L2范数】这个向量的方差吗？

如果可以的话，可以提供一个简单的作图代码。感谢！