Hallucinates too much
#6
by
WajihUllahBaig
- opened
With a typo error in the word, I tried the model - seems like it hallucinates alot.
I wasnt even talking about an article. It seems that it generated a question by itself in the answer
asking about opinion
Query: What is wrong with too much sugare consumption?:
- Obesity
-. - Diabetes
- High Blood Pressure
- High Cholesterol
- Depression
- Heart Attacks & Strokes
- Kidney disease
- Tooth decay
- Skin problems
- Sleep Apnea
- Osteoporosis
- Gall stones.
- 1.1k
- 2.1k
- 2.3k
- 2.1k
This is a great article.
It is good to see people are aware of how sugar can affect the body.
But what is your opinion on the new study that suggests sugar does not cause obesity?
Thank You
The new study, published in the journal Nature, has found that sugar intake and body fatness may share similar genes.
I have a question about this.
Is that because people who have a high sugar diet are eating less of other things, and therefore are losing more weight?
Or is there
definitely not useful model. need more pretrain or RLHF
Maybe because this is the base model and you have to finetune it on a dataset. Try doing that.
I wonder if lora is useful on finetune the model to a useful model, or must use full-weight finetune.
@cekal is right, this sort of behaviour is expected from the base model has it not been finetuned to take in instructions or have conversations. You should check-out Falcon-7B-Instruct instead π.
FalconLLM
changed discussion status to
closed