aalok-sathe/surprisal issues and pull requests

#23 - CLI wasn't actually using the BOS token arg; make sure to pass it in case of gpt-like models

Pull Request - State: open - Opened by aalok-sathe 4 months ago

#22 - Dependency issues in Python 3.12

Issue - State: closed - Opened by lpetley 8 months ago - 1 comment

#21 - Adding support for OpenAI chat models

Issue - State: closed - Opened by jennhu 9 months ago - 1 comment

#20 - Fix #19: Conflating causal LM and "gpt" model class (and masked LM and 'bert')

Pull Request - State: closed - Opened by aalok-sathe 10 months ago

#19 - Conflating causal LM and "gpt" model class

Issue - State: closed - Opened by jennhu 10 months ago
Labels: enhancement

#18 - Indexing into SurprisalArray using singletons fails in NGramSurprisal.

Issue - State: open - Opened by aalok-sathe 11 months ago
Labels: bug

#17 - Make the CI stuff work

Issue - State: closed - Opened by aalok-sathe 11 months ago
Labels: enhancement

#16 - Are surprisal values across different batch sizes slightly different?

Issue - State: open - Opened by aalok-sathe 11 months ago

#15 - Add deprecation warning to OpenAI models: OpenAI has deprecated their logprobs API for user input

Issue - State: closed - Opened by aalok-sathe 11 months ago - 5 comments
Labels: documentation, wontfix

#14 - support kenlm models and surprisal from them

Pull Request - State: closed - Opened by aalok-sathe 11 months ago - 1 comment

#13 - Updates and bugfixes from Ben: support different precisions, always right-pad, attn masks, etc.

Pull Request - State: closed - Opened by aalok-sathe 11 months ago - 1 comment

#12 - cyclical python version reqs causing `pip install` to fail

Issue - State: closed - Opened by aalok-sathe over 1 year ago

#11 - compute surprisal for Chinese characters

Issue - State: closed - Opened by hezaoke over 1 year ago - 6 comments

#10 - towards supporting Distributed Bloom-Petals models

Pull Request - State: closed - Opened by aalok-sathe over 1 year ago

#9 - Support masked LMs

Pull Request - State: open - Opened by aalok-sathe over 1 year ago - 1 comment
Labels: help wanted

#8 - reduced requirements to Python 3.8 and fixed GPU device specification for GPT-type models

Pull Request - State: closed - Opened by smeylan over 1 year ago

#7 - add minimal unit tests using pytest

Pull Request - State: closed - Opened by aalok-sathe almost 2 years ago

#6 - Slicing in SurprisalArray is not fully Pythonic

Issue - State: open - Opened by aalok-sathe almost 2 years ago
Labels: documentation, enhancement

#5 - support GPU

Pull Request - State: closed - Opened by aalok-sathe almost 2 years ago - 1 comment

#4 - add support to use GPT3 for surprisal using the OpenAI API

Pull Request - State: closed - Opened by aalok-sathe about 2 years ago

#3 - Support BERT-like bidirectional encoders for surprisal computation

Issue - State: open - Opened by aalok-sathe about 2 years ago
Labels: enhancement

#2 - Support GPU

Issue - State: closed - Opened by aalok-sathe about 2 years ago - 2 comments
Labels: enhancement, good first issue

#1 - Error when using Python-based tokenizers

Issue - State: open - Opened by aalok-sathe about 2 years ago - 1 comment

Ecosyste.ms: Issues

GitHub / aalok-sathe/surprisal issues and pull requests