File size: 4,983 Bytes
64e5534
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
da4b672
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
64e5534
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
---
license: apache-2.0
library_name: peft
tags:
- generated_from_trainer
base_model: unsloth/tinyllama-bnb-4bit
metrics:
- accuracy
model-index:
- name: tinyllama-qwantz-coherent
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# tinyllama-qwantz-coherent

This model is a fine-tuned version of [unsloth/tinyllama-bnb-4bit](https://huggingface.co/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