1/ GLM-130B outperforms OpenAI's GPT-3 175B and Google's PALM 540B on critical benchmarks.
AND it's open sourced, which means — you can run this model on your own machine, for free.
AND it's open sourced, which means — you can run this model on your own machine, for free.
2/ GLM-130B is instruction-finetuned, leverages Chinchilla scaling laws, and has bells and whistles like 4-bit quantization and bidirectional attention.
With 4-bit quantization, the model can run on 1 x 80 GB A100 or a consumer GPU rig.
With 4-bit quantization, the model can run on 1 x 80 GB A100 or a consumer GPU rig.
3/ The paper, written by a team at a public university, is well-written, and methodology well-documented:
arxiv.org/pdf/2210.02414…
arxiv.org/pdf/2210.02414…
4/ In short, there's an open-source model that beats GPT-175B and is competitive with Google latest language models.
You can download the weights and run tests yourself:
github.com/THUDM/GLM-130B
You can download the weights and run tests yourself:
github.com/THUDM/GLM-130B
• • •
Missing some Tweet in this thread? You can try to
force a refresh
