tinyllama-qwantz-coherent

This model is a fine-tuned version of unsloth/tinyllama-bnb-4bit on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5015
• Accuracy: 0.8650

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.4508	1.0	1428	0.3631	0.8509
0.4223	2.0	2856	0.4421	0.8617
0.147	3.0	4284	0.5015	0.8650

Can save 90% of coherent strings by discarding 79% of dp strings (cutoff is -60.36113500595093)
Can save 95% of coherent strings by discarding 70% of dp strings (cutoff is -76.16859674453735)
Can save 98% of coherent strings by discarding 58% of dp strings (cutoff is -91.0932719707489)
Can save 99% of coherent strings by discarding 40% of dp strings (cutoff is -96.55501246452332)

My only problem (s) have to do with  ==>  coherent: 95.82%
My only problem (s) to cheer them personally  ==>  dp: 99.89%
(in small text) crazy utahraptor  ==>  coherent: 60.62%
(in small text) ".  ==>  dp: 58.77%
Well, I've made up my own joke to get him today. All I need to do is "  ==>  coherent: 52.12%
Well, I've made up my own joke to get him today. All I need a father and gentlemen  ==>  dp: 99.77%
I will be immortalized by kicking an evil  ==>  coherent: 50.01%
I will be immortalized by kicking other punches  ==>  dp: 58.54%
Aw shoot, I was supposed to  ==>  coherent: 92.77%
Aw shoot, I was APOCALYPSE PORN  ==>  dp: 79.21%
Get it? Because CRIME DOESN'T PAY!! Listen, my story has  ==>  dp: 59.16%
Get it? Because CRIME DOESN'T PAY!! Listen, transcriptions of it  ==>  dp: 80.30%
Utahraptor!! DON'T LISTEN TO  ==>  coherent: 90.59%
Utahraptor! This is sort of  ==>  dp: 88.15%
Doesn't exist in my mouth, that is!! Because it's too big  ==>  coherent: 95.82%
Doesn't exist in my mouth, that is!! Because if Superman.  ==>  dp: 99.44%
Now, HERE'S how  ==>  coherent: 94.16%
Now, guys would  ==>  dp: 73.45%
But I am a rock star  ==>  coherent: 98.57%
But I am a guy come  ==>  dp: 96.79%
But I have a solution to make them interesting again: all you need is stories where not  ==>  dp: 64.66%
But I have a solution to make them interesting again: all you need is gonna! Diseases  ==>  dp: 99.92%
At that point, there's a sequence of six nines in a row, and his joke was that he'd like to memorize pi up to that point, so that when reciting he could end with "9,9,9,9,9,9... and so on. " Others  ==>  dp: 77.18%
At that point, there's a sequence of six nines in a row, and his joke was that he'd like to memorize pi up to that point, so that when reciting he could end with "9,9,9,9,9,9... and so it's great he looks  ==>  dp: 99.98%
This is definitely called " T -Rex's Hilarious e joke ", okay  ==>  coherent: 77.15%
This is definitely called " T -Rex's Hilarious e joke AND IN THE  ==>  dp: 99.15%
" Your mouth is full of cockroaches:  ==>  coherent: 72.25%
" Your mouth is full of smooches.  ==>  coherent: 80.86%
Excuse me, sexual congress? Everyone else on the planet is dead, and  ==>  coherent: 93.15%
Excuse me, sexual congress? Everyone else on the planet without syntactic ambiguity!  ==>  dp: 96.98%
Sony is going to write swears on my bathroom  ==>  dp: 50.96%
Sony is going to write their babies need to  ==>  dp: 99.89%
Beginning with the most modest: why am I  ==>  coherent: 90.87%
Beginning with the most modest: why T -  ==>  dp: 84.62%
Is there any greater meaning -to anything  ==>  dp: 93.76%
Is there any greater meaning? When you  ==>  coherent: 57.11%
I've also got steaks AND  ==>  coherent: 83.50%
I've also cold -deterministic  ==>  dp: 99.30%
I had a friend (female) who dated her roommate (also female)  ==>  coherent: 99.24%
I had a friend (female) who dated her roommate, je suis grand  ==>  dp: 99.30%
Yes... TOO BAD INDEED  ==>  coherent: 70.86%
Yes... TOO MANY YEARS  ==>  coherent: 64.13%

Framework versions

• PEFT 0.10.0
• Transformers 4.40.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1