File size: 10,460 Bytes
f80be72 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 |
---
language:
- en
license: mit
library_name: transformers
tags:
- axolotl
- finetune
- dpo
- microsoft
- phi
- pytorch
- phi-3
- nlp
- code
- chatml
base_model: microsoft/Phi-3-mini-4k-instruct
model_name: Phi-3-mini-4k-instruct-v0.3
pipeline_tag: text-generation
inference: false
model_creator: MaziyarPanahi
model-index:
- name: Phi-3-mini-4k-instruct-v0.3
results:
- task:
type: text-generation
name: Text Generation
dataset:
name: AI2 Reasoning Challenge (25-Shot)
type: ai2_arc
config: ARC-Challenge
split: test
args:
num_few_shot: 25
metrics:
- type: acc_norm
value: 63.48
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Phi-3-mini-4k-instruct-v0.3
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: HellaSwag (10-Shot)
type: hellaswag
split: validation
args:
num_few_shot: 10
metrics:
- type: acc_norm
value: 80.86
name: normalized accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Phi-3-mini-4k-instruct-v0.3
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: MMLU (5-Shot)
type: cais/mmlu
config: all
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 69.24
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Phi-3-mini-4k-instruct-v0.3
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: TruthfulQA (0-shot)
type: truthful_qa
config: multiple_choice
split: validation
args:
num_few_shot: 0
metrics:
- type: mc2
value: 60.66
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Phi-3-mini-4k-instruct-v0.3
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: Winogrande (5-shot)
type: winogrande
config: winogrande_xl
split: validation
args:
num_few_shot: 5
metrics:
- type: acc
value: 72.77
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Phi-3-mini-4k-instruct-v0.3
name: Open LLM Leaderboard
- task:
type: text-generation
name: Text Generation
dataset:
name: GSM8k (5-shot)
type: gsm8k
config: main
split: test
args:
num_few_shot: 5
metrics:
- type: acc
value: 74.53
name: accuracy
source:
url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=MaziyarPanahi/Phi-3-mini-4k-instruct-v0.3
name: Open LLM Leaderboard
quantized_by: bartowski
---
## Llamacpp imatrix Quantizations of Phi-3-mini-4k-instruct-v0.3
Using <a href="https://github.com/ggerganov/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggerganov/llama.cpp/releases/tag/b3086">b3086</a> for quantization.
Original model: https://huggingface.co/MaziyarPanahi/Phi-3-mini-4k-instruct-v0.3
All quants made using imatrix option with dataset from [here](https://gist.github.com/bartowski1182/eb213dccb3571f863da82e99418f81e8)
## Prompt format
```
<|im_start|> system
{system_prompt}<|im_end|>
<|im_start|> user
{prompt}<|im_end|>
<|im_start|> assistant
```
## Download a file (not the whole branch) from below:
| Filename | Quant type | File Size | Description |
| -------- | ---------- | --------- | ----------- |
| [Phi-3-mini-4k-instruct-v0.3-Q8_0.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-Q8_0.gguf) | Q8_0 | 4.06GB | Extremely high quality, generally unneeded but max available quant. |
| [Phi-3-mini-4k-instruct-v0.3-Q6_K.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-Q6_K.gguf) | Q6_K | 3.13GB | Very high quality, near perfect, *recommended*. |
| [Phi-3-mini-4k-instruct-v0.3-Q5_K_M.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-Q5_K_M.gguf) | Q5_K_M | 2.81GB | High quality, *recommended*. |
| [Phi-3-mini-4k-instruct-v0.3-Q5_K_S.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-Q5_K_S.gguf) | Q5_K_S | 2.64GB | High quality, *recommended*. |
| [Phi-3-mini-4k-instruct-v0.3-Q4_K_M.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-Q4_K_M.gguf) | Q4_K_M | 2.39GB | Good quality, uses about 4.83 bits per weight, *recommended*. |
| [Phi-3-mini-4k-instruct-v0.3-Q4_K_S.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-Q4_K_S.gguf) | Q4_K_S | 2.18GB | Slightly lower quality with more space savings, *recommended*. |
| [Phi-3-mini-4k-instruct-v0.3-IQ4_XS.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-IQ4_XS.gguf) | IQ4_XS | 2.05GB | Decent quality, smaller than Q4_K_S with similar performance, *recommended*. |
| [Phi-3-mini-4k-instruct-v0.3-Q3_K_L.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-Q3_K_L.gguf) | Q3_K_L | 2.08GB | Lower quality but usable, good for low RAM availability. |
| [Phi-3-mini-4k-instruct-v0.3-Q3_K_M.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-Q3_K_M.gguf) | Q3_K_M | 1.95GB | Even lower quality. |
| [Phi-3-mini-4k-instruct-v0.3-IQ3_M.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-IQ3_M.gguf) | IQ3_M | 1.85GB | Medium-low quality, new method with decent performance comparable to Q3_K_M. |
| [Phi-3-mini-4k-instruct-v0.3-Q3_K_S.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-Q3_K_S.gguf) | Q3_K_S | 1.68GB | Low quality, not recommended. |
| [Phi-3-mini-4k-instruct-v0.3-IQ3_XS.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-IQ3_XS.gguf) | IQ3_XS | 1.62GB | Lower quality, new method with decent performance, slightly better than Q3_K_S. |
| [Phi-3-mini-4k-instruct-v0.3-IQ3_XXS.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-IQ3_XXS.gguf) | IQ3_XXS | 1.51GB | Lower quality, new method with decent performance, comparable to Q3 quants. |
| [Phi-3-mini-4k-instruct-v0.3-Q2_K.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-Q2_K.gguf) | Q2_K | 1.41GB | Very low quality but surprisingly usable. |
| [Phi-3-mini-4k-instruct-v0.3-IQ2_M.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-IQ2_M.gguf) | IQ2_M | 1.31GB | Very low quality, uses SOTA techniques to also be surprisingly usable. |
| [Phi-3-mini-4k-instruct-v0.3-IQ2_S.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-IQ2_S.gguf) | IQ2_S | 1.21GB | Very low quality, uses SOTA techniques to be usable. |
| [Phi-3-mini-4k-instruct-v0.3-IQ2_XS.gguf](https://huggingface.co/bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF/blob/main/Phi-3-mini-4k-instruct-v0.3-IQ2_XS.gguf) | IQ2_XS | 1.15GB | Very low quality, uses SOTA techniques to be usable. |
## Downloading using huggingface-cli
First, make sure you have hugginface-cli installed:
```
pip install -U "huggingface_hub[cli]"
```
Then, you can target the specific file you want:
```
huggingface-cli download bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF --include "Phi-3-mini-4k-instruct-v0.3-Q4_K_M.gguf" --local-dir ./
```
If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:
```
huggingface-cli download bartowski/Phi-3-mini-4k-instruct-v0.3-GGUF --include "Phi-3-mini-4k-instruct-v0.3-Q8_0.gguf/*" --local-dir Phi-3-mini-4k-instruct-v0.3-Q8_0
```
You can either specify a new local-dir (Phi-3-mini-4k-instruct-v0.3-Q8_0) or download them all in place (./)
## Which file should I choose?
A great write up with charts showing various performances is provided by Artefact2 [here](https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9)
The first thing to figure out is how big a model you can run. To do this, you'll need to figure out how much RAM and/or VRAM you have.
If you want your model running as FAST as possible, you'll want to fit the whole thing on your GPU's VRAM. Aim for a quant with a file size 1-2GB smaller than your GPU's total VRAM.
If you want the absolute maximum quality, add both your system RAM and your GPU's VRAM together, then similarly grab a quant with a file size 1-2GB Smaller than that total.
Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'.
If you don't want to think too much, grab one of the K-quants. These are in format 'QX_K_X', like Q5_K_M.
If you want to get more into the weeds, you can check out this extremely useful feature chart:
[llama.cpp feature matrix](https://github.com/ggerganov/llama.cpp/wiki/Feature-matrix)
But basically, if you're aiming for below Q4, and you're running cuBLAS (Nvidia) or rocBLAS (AMD), you should look towards the I-quants. These are in format IQX_X, like IQ3_M. These are newer and offer better performance for their size.
These I-quants can also be used on CPU and Apple Metal, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide.
The I-quants are *not* compatible with Vulcan, which is also AMD, so if you have an AMD card double check if you're using the rocBLAS build or the Vulcan build. At the time of writing this, LM Studio has a preview with ROCm support, and other inference engines have specific builds for ROCm.
Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski
|