And of course, lots of usual docs/bug fixes! s/o @eltociear @tangjeff0 (x2) @concaption @kevin_w_huo Jon Luo, Peng Qu Happy weekend!
OPEN SOURCE
-
LangChain PagedPDF Tracks Page Numbers for Source Citations
By
–
PDFs with pages @andrewwhite01 added a PagedPDF that keeps track of page numbers in the metadata This metadata is useful when doing question answering with sources! Docs: https://
langchain.readthedocs.io/en/latest/modu
les/document_loaders/examples/pdf.html
… -
Tiktoken Improvements: Enhanced Special Token Parameters
By
–
Tiktoken improvements Thanks to @jamescalam you can now specify `allowed_special` and `disallowed_special` as parameters for the tiktoken tokenizer PR: https://
github.com/hwchase17/lang
chain/pull/924
… -
LangChain Adds Batch Embeddings Support for OpenAI
By
–
Batch embeddings While we have support for batch processing of LLM calls, we didn't have support for batch processing of embeddings until now. @jmdagdelen added it for the @OpenAI embedding module PR: https://
github.com/hwchase17/lang
chain/pull/969/files
… -
LangChain 0.0.82 Release: New Document Loaders and Embeddings
By
–
End of week release! 0.0.82 Some new/improved document loaders:
URLs – @UnstructuredIO Matt Robinson
EveryNote
PDFs @andrewwhite01 YouTube (Alex Zhangji) Some new features:
Batch embeddings @jmdagdelen Tiktoken improvements And the usual docs/bug fixes! -
LangChain URL Document Loader Integration with UnstructuredIO
By
–
URL document loader Using @UnstructuredIO
, now load data from URLs directly into the LangChain document format s/o Matt Robinson for adding! Docs: https://
langchain.readthedocs.io/en/latest/modu
les/document_loaders/examples/url.html
… -
Open Source AI Alternatives to Paid Closed Source APIs
By
–
My point here was that you can replicate the same with a bit more effort here and there with Open Source alternatives instead of paying a closed source API
-
SpeechT5 vs FastSpeech2 vs VITS: TTS Model Comparison
By
–
Ofcourse. It depends on what do you optimise for SpeechT5 TTS (AR) does sound better than vanilla FastSpeech2 (NAR) however worse than VITS/PortaSpeech. Significantly worse than TorToise TTS. However it is considerably slow.
-
Stability AI Open-Sources Checkpoints Thanks
By
–
Also Big Thanks to @StabilityAI and @RiversHaveWings for open-sourcing the checkpoints!
-
Open-Assistant Project: LAION-AI’s Open Source Initiative
By
–
El proyecto no es ese, es este. https://
github.com/LAION-AI/Open-
Assistant
…