Skip to content

Support masked LMs #9

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
aalok-sathe wants to merge 3 commits into
base: main
Choose a base branch
from
Draft

Support masked LMs #9

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
1473cbc
small modification in basic test driver script
aalok-sathe Jan 24, 2023
5ef4ce4
unfinished intermediate commit to start implementing pseudo log-likel…
carina-kauf Nov 15, 2023
e893aa2
Merge branch 'main' into feature-support-masked-lm-1
aalok-sathe Nov 15, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 16 additions & 1 deletion surprisal/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -321,7 +321,22 @@ def surprise(
mask_mask = torch.eye(n, n)[1:, :].repeat(b, 1).bool()
ids_with_bos_token[mask_mask] = self.tokenizer.mask_token_id

import IPython
# below is from ckauf and neuranna?
# if "within_word_l2r" == PLL_metric:
# """
# Future tokens belonging to the same word as the target token are masked during token inference as well.
# """
# mask_indices = [
# [mask_pos]
# + [
# j
# for j in range(mask_pos + 1, effective_length + 2)
# if word_ids[j] == word_ids[mask_pos]
# ]
# if word_ids[mask_pos] is not None
# else [mask_pos]
# for mask_pos in range(effective_length + 2)
# ]

raise NotImplementedError

Expand Down
7 changes: 4 additions & 3 deletions test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,11 +5,12 @@

import surprisal

g = surprisal.AutoHuggingFaceModel.from_pretrained(model_id="gpt2")
# b = surprisal.AutoHuggingFaceModel.from_pretrained(model_id="bert-base-uncased")
# m = surprisal.AutoHuggingFaceModel.from_pretrained(model_id="gpt2")
m = surprisal.AutoHuggingFaceModel.from_pretrained(model_id="bert-base-uncased")


stims = [
"The hooligan wrecked my new car!",
"I am a cat on the mat",
# "The cat sat on the mat.",
# "The cat sat on the pizza.",
Expand All @@ -20,7 +21,7 @@
# "How likely is a spicy burrito?",
]

surps = [*g.surprise(stims), *g.surprise(stims, use_bos_token=False)]
surps = [*m.surprise(stims), *m.surprise(stims, use_bos_token=False)]


f, a = plt.subplots()
Expand Down