---
license: other
license_name: deepseek-coder-33b
license_link: https://huggingface.co/deepseek-ai/deepseek-coder-33b-base/blob/main/LICENSE
quantized_by: bartowski
pipeline_tag: text-generation
---

## Exllama v2 Quantizations of Trinity-33B-v1.0

Using <a href="https://github.com/turboderp/exllamav2/releases/tag/v0.0.13">turboderp's ExLlamaV2 v0.0.13</a> for quantization.

<b>The "main" branch only contains the measurement.json, download one of the other branches for the model (see below)</b>

Each branch contains an individual bits per weight, with the main one containing only the meaurement.json for further conversions.

Original model: https://huggingface.co/WhiteRabbitNeo/Trinity-33B-v1.0

| Branch | Bits | lm_head bits | VRAM (4k) | VRAM (16k) | VRAM (32k) | Description |
| ------ | ---- | ------------ | ---- | ---- | ---- | ----------- |
| [6_5](https://huggingface.co/bartowski/Trinity-33B-v1.0-exl2/tree/6_5)   | 6.5  | 8.0 | 28.9 GB | 31.6 GB | 35.6 GB | Near unquantized performance at vastly reduced size, **recommended**.       |
| [4_25](https://huggingface.co/bartowski/Trinity-33B-v1.0-exl2/tree/4_25) | 4.25 | 6.0 | 19.5 GB | 22.2 GB | 26.2 GB | GPTQ equivalent bits per weight, slightly higher quality.                   |
| [3_5](https://huggingface.co/bartowski/Trinity-33B-v1.0-exl2/tree/3_5)   | 3.5  | 6.0 | 16.5 GB | 19.2 GB | 23.2 GB | Lower quality, only use if you have to.                                     |
| [3_0](https://huggingface.co/bartowski/Trinity-33B-v1.0-exl2/tree/3_0)   | 3.0  | 6.0 | 14.3 GB | 17.0 GB | 21.0 GB | Very low quality, usable with 16gb of VRAM. |
| [2_4](https://huggingface.co/bartowski/Trinity-33B-v1.0-exl2/tree/3_0)   | 2.4  | 6.0 | 12.0 GB | 14.7 GB | 18.7 GB | Extremely low quality, only recommended if trying to fit into 12GB with under 4k context. |

## Download instructions

With git:

```shell
git clone --single-branch --branch 6_5 https://huggingface.co/bartowski/Trinity-33B-v1.0-exl2 Trinity-33B-v1.0-exl2-6_5
```

With huggingface hub (credit to TheBloke for instructions):

```shell
pip3 install huggingface-hub
```

To download the `main` (only useful if you only care about measurement.json) branch to a folder called `Trinity-33B-v1.0-exl2`:

```shell
mkdir Trinity-33B-v1.0-exl2
huggingface-cli download bartowski/Trinity-33B-v1.0-exl2 --local-dir Trinity-33B-v1.0-exl2 --local-dir-use-symlinks False
```

To download from a different branch, add the `--revision` parameter:

Linux:

```shell
mkdir Trinity-33B-v1.0-exl2-6_5
huggingface-cli download bartowski/Trinity-33B-v1.0-exl2 --revision 6_5 --local-dir Trinity-33B-v1.0-exl2-6_5 --local-dir-use-symlinks False
```

Windows (which apparently doesn't like _ in folders sometimes?):

```shell
mkdir Trinity-33B-v1.0-exl2-6.5
huggingface-cli download bartowski/Trinity-33B-v1.0-exl2 --revision 6_5 --local-dir Trinity-33B-v1.0-exl2-6.5 --local-dir-use-symlinks False
```

Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski