Catalan @jordimash
's fine-tuned Whisper small checkpoint produced an astounding WER of 8.56 compared to 23.8 by the large checkpoint! Model: https://
huggingface.co/softcatala/whi
sper-small-ca
… Space: https://
huggingface.co/spaces/softcat
ala/whisper-demo-catalan
…
OPEN SOURCE
-
Whisper Small Model Achieves 8.56 WER for Catalan Speech
By
–
-
Fine-tuned Whisper Model Achieves Significant WER Improvement for Kannada
By
–
Kannada @saiteya10294's fine-tuned Whisper small checkpoint managed to get the WER down from 100.4 (large) -> 23.1! Model: https://
huggingface.co/steja/whisper-
small-kannada
… -
Lithuanian Whisper Model Achieves Major WER Improvement
By
–
Lithuanian @DeividasMat
's fine-tuned Whisper medium checkpoint managed to get the WER down from 35.2 (large) -> 20.44 Model: -
Researchers Win OGB-LSC Competition with Accelerated GNN Training
By
–
Super proud of our researchers for winning the MAG240M Track of the OGB-LSC competition at #NeurIPS2022! They used PGL and PaddlePaddle to accelerate GNN training on GPUs, reaching SOTA in only 1 hr (40 hrs before). Check out their code: https://
github.com/PaddlePaddle/P
GL/tree/main/examples/NeurIPS2022-OGB-Challenge/MAG240M
… -
Try Replicate’s Inpainter Tool on GitHub
By
–
Try it out yourself: https://t.co/d6dnzXsjR5 https://t.co/6X4w43bgJ1
— Replicate (@replicate) 15 décembre 2022Try it out yourself: https://
github.com/replicate/inpa
inter
… -
ERNIE-Code: Unified Multilingual Code-to-Text Model
By
–
Introducing ERNIE-Code, a unified pre-trained language model for 116 natural languages and 6 programming languages. ERNIE-Code outperforms previous models for multilingual code-to-text, text-to-code, code-to-code, and text-to-text generation. Paper: https://
arxiv.org/abs/2212.06742 -
LangChain Enables LLM Caching with Three Lines of Code
By
–
Caching now enabled With 3 lines of code, you can now enable caching for all LLM calls This makes it cheaper and easier to experiment with changing only parts of a chain Supports both a temporary InMemoryCache, as well as a persistent SQLiteCache
-
LangChain v0.0.38: New LLM Caching Feature Released
By
–
New in v0.0.38 Read more about this here: https://
langchain.readthedocs.io/en/latest/exam
ples/prompts/llm_functionality.html#Caching
… @AkashSamant4 for making LLM configurations serializable, a prerequisite for this @brucehammer and @DannyHabibs for highlighting this as a valuable feature -
Public and Private AI Training: Different Accuracy Standards
By
–
Also, my personal opinion is that both public pretraining (with certain constraints/limitations) as well as private training from scratch both have roles to play in the space. They of course can't be held to the same accuracy standard though.
-
Public vs Proprietary Datasets in AI Model Training
By
–
Thanks Alex. I like the papers that do this, but I also have some concern when this is done on a dataset that is proprietary and only Google has access (JFT). I would like to see a version that is pretrained on LAION. This still has privacy issues, but is at least all public.