File size: 2,599 Bytes
1bf4a80
 
 
 
 
 
 
 
 
 
0d4c2a2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
68e49f9
0d4c2a2
cee8ea0
9e7e61d
e793e71
883600b
3d096d1
0d4c2a2
487b963
 
 
68e49f9
1c6adcc
cee8ea0
 
 
 
 
 
 
67ab3fc
1c6adcc
 
cee8ea0
 
 
 
 
 
 
 
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
---
base_model: https://huggingface.co/Phind/Phind-CodeLlama-34B-v2
inference: false
license: llama2
model_creator: https://huggingface.co/Phind
model_name: Phind-Codellama-34B-v2
model_type: llama
quantized_by: latimar
---

# Phind-CodeLlama-34B-v2 EXL2
 
Weights of [Phind-CodeLlama-34B-v2](https://huggingface.co/Phind/Phind-CodeLlama-34B-v2) converted
to [EXL2](https://github.com/turboderp/exllamav2#exl2-quantization) format.
 
Each separate quant is in a different branch, like in The Bloke's GPTQ repos.
 
```
export BRANCH=5_0-bpw-h8
git clone --single-branch --branch ${BRANCH} https://huggingface.co/latimar/Phind-Codellama-34B-v2-exl2
```
 
There are the following branches:
 
```
5_0-bpw-h8
5_0-bpw-h8-evol-ins
4_625-bpw-h6
4_4-bpw-h8
4_125-bpw-h6
3_8-bpw-h6
2_75-bpw-h6
2_55-bpw-h6
```

* Calibration dataset used for conversion: [wikitext-v2](https://huggingface.co/datasets/wikitext/blob/refs%2Fconvert%2Fparquet/wikitext-2-v1/test/0000.parquet)
* Evaluation dataset used to calculate perplexity: [wikitext-v2](https://huggingface.co/datasets/wikitext/blob/refs%2Fconvert%2Fparquet/wikitext-2-v1/validation/0000.parquet)
* Calibration dataset used for conversion of `5_0-bpw-h8-evol-ins`: [wizardLM-evol-instruct_70k](https://huggingface.co/datasets/WizardLM/WizardLM_evol_instruct_70k/blob/refs%2Fconvert%2Fparquet/default/train/0000.parquet)
* Evaluation dataset used to calculate ppl for `Evol-Ins`: : [nikrosh-evol-instruct](https://huggingface.co/datasets/nickrosh/Evol-Instruct-Code-80k-v1/blob/refs%2Fconvert%2Fparquet/default/train/0000.parquet)
* When converting `4_4-bpw-h8` quant, additional `-mr 32` arg was used.

PPL was measured with the [test_inference.py exllamav2 script](https://github.com/turboderp/exllamav2/blob/master/test_inference.py):

```  
python test_inference.py -m /storage/models/LLaMA/EXL2/Phind-Codellama-34B-v2 -ed /storage/datasets/text/evol-instruct/nickrosh-evol-instruct-code-80k.parquet
```

| BPW         | PPL on Wiki | PPL on Evol-Ins | File Size (Gb) | 
| ----------- | ----------- | --------------- | -------------- |
|  2.55-h6    |  11.0310    |     2.4542      |   10.56        |
|  2.75-h6    |  9.7902     |     2.2888      |   11.33        |
|  3.8-h6     |  6.7293     |     2.0724      |   15.37        |
|  4.125-h6   |  6.6713     |     2.0617      |   16.65        |
|  4.4-h8     |  6.6487     |     2.0509      |   17.76        |
|  4.625-h6   |  6.6576     |     2.0459      |   18.58        |
|  5.0-h8     |  6.6379     |     2.0419      |   20.09        |
|  5.0-h8-ev  |  6.7785     |     2.0445      |   20.09        |