What is the version of yt-dlp?
a670531899 opened this issue · 7 comments
What is the version of yt-dlp?
When I use the dwn_yt.py scripts to download videos, it says "Unrecognized option '-cookies'.".
Is it a version problem?
Or can I just use youtube-dl to download videos by myself?
@a670531899 could you let me know what command you used?
you need to use "--cookies=/path/to/cookiesfile"
yt-dlp (as far as I know) is better maintained youtube-dl. Using latest version is recommended.
I used this cmd
python prep_data/dwn_yt.py --task_type='dwn_vids' --max_processes=30 --cookies_file=cookies.txt
And I get this error:
Unrecognized option '-cookies'.
Error splitting the argument list: Option not found
@a670531899 I recall now, I think with yt-dlp you don't need to use cookies (they mention this in their readme). cookies was for an earlier version which relied on youtube-dl
Let me know if it works out
@TheShadow29 Thanks for your reply! And I have fixed this problem.
However, I have a doubt when I want to predict the arguments with the predicted verbs.
Why does the argument generation process need the annotation?
Specifically, it is the code in mdl_sf_base.py:
inp_prep = self.prepare_prev_toks_inp(inp)
inp["src_tokens"] = inp_prep["dst_toks"][..., :1]
inp["src_lengths"] = inp_prep["dst_lens"]
inp_ids = inp_prep["dst_toks"][..., :1]
There you use the dst_toks which is loaded from the annotation file.
That's really confused me, looking for your reply.
@a670531899 Great!
I take up until the first token (which is the given). In our task, you are given the verb, and you need to predict the semantic roles.
inp_ids = inp_prep["dst_toks"][..., :1] just takes the first verb token as the input ids. You can consider these to be the prefixes for the decoder.
Does that answer your question?
@TheShadow29 Thank you! That's very helpful.
@a670531899 feel free to create a new issue if you have any questions!