i tested the q8 in lm studio( getting wrong output)

#9
by gopi87 - opened

hey guys i tested the result in lm studio but its showing wrong answer is it because prompt issues ?

I tested int4 gguf in llama.cpp, checked the conversation template but still output seems not right

yeah in mine it seems like the tags are messed up? I'm still trying to determine if I can fix this on my end:

me:

write me a super tiny rust replacement for chrony

solar pro preview:

system<|endoftext|>; D., & Fuchs, M. A. (2015). The effect of teacher training on outcomes in instructional practices and student achievement: A meta-analysis. Psychological Bulletin, 47(3), 368–393.Find ArticleGoogle SchSchooler, K., Pfeiffer, J., Mullins, D., & Glassman, T. (2015). Teachers who can do: A case study of how teachers learned math skills in a university setting to improve their teaching effectiveness and students’ achievement. Journal of Research on Educational Development, 68(3), 9-24.Find ArticleGoogle SchSchott, D., & Gleicher, J. (1997). An investigation of the impacts of a university mathematics course for teachers upon teacher instructional practices and student achievement. Mathematics Education Journal, 10:18–36.Find ArticleGoogle
Searching for Categories
Looking for something else? Check out these categories that have relevant resources to technology-fposed problems or related closely to them.
This category is filled with articles about the development and testing of new technologies as well as case studies of existing technologies used in a unique manner. See all articles from this category.
These articles use one type of data collection methodology, designing their research according to the parameters that it dictates. See all articles from this methodology.<|endoftext|>s (23%) and the second most frequent site is the upper jaw with 57 dental implants in 18 patients (79%). The right side was more affected than left side both for location on face (68%) vs 3

me too, full of hallucination and wrong thinks.
Tested: RTX 3090 - LM Studio - 4q_k_m

looks like there is a issues with model i think!!

gopi87 changed discussion title from i tested the q8 in lm studio to i tested the q8 in lm studio( getting wrong output)

You have to disable the system prompt and only use user. Can you confirm if the system prompt is not being used?

Me too. Called marble, a marpie and then performed unrelated arithmetic, gguf q8

me too

完全无法正常使用。有人测试过f32模型也会有类似的问题吗?我试过不使用system,结果一样是胡乱输出

Sign up or log in to comment