RuntimeError: Internal: D:\a\sentencepiece\sentencepiece\src\sentencepiece_processor.cc(1102) [model_proto->ParseFromArray(serialized.data(), serialized.size())]
Hi everyone, I'm really excited to try and install this uncensored ChatGPT alternative on my computer, being tired of the creative limitations the extreme censorship seems to enforce (couldn't even create a hate-poem with clever puns for the Skaven species from Warhammer without triggering some content policy violation...). Unfortunately, I seem to be hitting a wall trying to install the AI.
I seem to be getting a different Runtime Error than the other people here in the comments, when trying to run the start-webui (following the recent video guide from Aitrepeneur):
RuntimeError: Internal: D:\a\sentencepiece\sentencepiece\src\sentencepiece_processor.cc(1102) [model_proto->ParseFromArray(serialized.data(), serialized.size())]
No idea what it means or what I'm supposed to do about it. Tried messing around with the start-webui parameters as suggested various places (e.g. --gpu-memory and --pre_layer etc.), but it has no effect at all, always the same error and traceback files. Wondering if the install itself is messed up somehow, so maybe I'll have to download and install everything from scratch again?
I did notice some deviations from Aitrepeneur's guide in my installation files, such as a lack of a git folder (to delete according to his instructions), and instead of --chat I had --cai chat in the parameters, which the command window itself told me to edit to --chat upon trying to launch - though ultimately it still led to the same runtime error.
My GPU trying to run it is a 10GB RTX 3080 (desktop) if that matters - I know it should ideally be 16+GB, but apparently people with even less powerful cards than mine are having some success?
Any help or tips is welcome, and thank you in advance.
I solved this problem on my machine, for some reason the tokenizor is stored using github LFS despite being less than a megabyte, you likely have a 1kb file pointer instead of the real tokenizor
I solved this problem on my machine, for some reason the tokenizer is stored using GitHub LFS despite being less than a megabyte, you likely have a 1kb file pointer instead of the real tokenizer
How can we fix this, because I'm running into the same issues, and can't seem to find the tokenizer.
Try to download tokenizer.model from "Files and versions" and put it into the folder of the model. If there is already a file with only 1KB replace it with the correct tokenizer.model.
I solved this problem on my machine, for some reason the tokenizor is stored using github LFS despite being less than a megabyte, you likely have a 1kb file pointer instead of the real tokenizor
Thanks mm04926412, you're a godsend! This was the fix for me as well. Would've thanked you last night already if my newly made account wasn't restricted 🙏
godsend! indeed