Clément Moutet

Results 3 comments of Clément Moutet

@orangetin I can confirm that the inference of togethercomputer/Pythia-Chat-Base-7B works on Google Colab Pro +, but not the togethercomputer/GPT-NeoXT-Chat-Base-20B (this model can get loaded but consumes 39.4 GB of vRAM...

@thomasjv799 No difference for the moment unfortunately A major hardware limitation for playing with LLMs is the vRAM, the memory of the GPU, on which the model needs to be...

I share all files generated during this test (in 2 parts) [Archive1.zip](https://github.com/user-attachments/files/16496145/Archive1.zip) [Archive2.zip](https://github.com/user-attachments/files/16496146/Archive2.zip)