README.md

---
pipeline_tag: text-generation
inference: true
widget:
- text: 'def print_hello_world():'
  example_title: Hello world
  group: Python
datasets:
- bigcode/the-stack-v2-train
license: bigcode-openrail-m
library_name: transformers
tags:
- code
model-index:
- name: starcoder2-15b
  results:
  - task:
      type: text-generation
    dataset:
      name: CruxEval-I
      type: cruxeval-i
    metrics:
    - type: pass@1
      value: 48.1
  - task:
      type: text-generation
    dataset:
      name: DS-1000
      type: ds-1000
    metrics:
    - type: pass@1
      value: 33.8
  - task:
      type: text-generation
    dataset:
      name: GSM8K (PAL)
      type: gsm8k-pal
    metrics:
    - type: accuracy
      value: 65.1
  - task:
      type: text-generation
    dataset:
      name: HumanEval+
      type: humanevalplus
    metrics:
    - type: pass@1
      value: 37.8
  - task:
      type: text-generation
    dataset:
      name: HumanEval
      type: humaneval
    metrics:
    - type: pass@1
      value: 46.3
  - task:
      type: text-generation
    dataset:
      name: RepoBench-v1.1
      type: repobench-v1.1
    metrics:
    - type: edit-smiliarity
      value: 74.08
quantized_by: bartowski
---

## Exllama v2 Quantizations of starcoder2-15b

Using <a href="https://github.com/turboderp/exllamav2/">turboderp's ExLlamaV2 v0.0.15 preview</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/bigcode/starcoder2-15b


| Branch | Bits | lm_head bits | VRAM (4k) | VRAM (16k) | VRAM (32k) | Description |
| ----- | ---- | ------- | ------ | ------ | ------ | ------------ |
| [8_0](https://huggingface.co/bartowski/starcoder2-15b-exl2/tree/8_0) | 8.0 | 8.0 | 16.6 GB | 17.5 GB | 18.8 GB | Maximum quality that ExLlamaV2 can produce, near unquantized performance. |
| [6_5](https://huggingface.co/bartowski/starcoder2-15b-exl2/tree/6_5) | 6.5  | 8.0 | 13.9 GB | 14.9 GB | 16.2 GB | Near unquantized performance at vastly reduced size, **recommended**. |
| [5_0](https://huggingface.co/bartowski/starcoder2-15b-exl2/tree/5_0) | 5.0  | 6.0 | 11.2 GB | 12.2 GB | 13.5 GB | Slightly lower quality vs 6.5. |
| [4_25](https://huggingface.co/bartowski/starcoder2-15b-exl2/tree/4_25) | 4.25 | 6.0 | 9.8 GB | 10.7 GB | 12.0 GB | GPTQ equivalent bits per weight. |
| [3_5](https://huggingface.co/bartowski/starcoder2-15b-exl2/tree/3_5) | 3.5  | 6.0 | 8.4 GB | 9.3 GB | 10.6 GB | Lower quality, not recommended. |


## Download instructions

With git:

```shell
git clone --single-branch --branch 6_5 https://huggingface.co/bartowski/starcoder2-15b-exl2
```

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 `starcoder2-15b-exl2`:

```shell
mkdir starcoder2-15b-exl2
huggingface-cli download bartowski/starcoder2-15b-exl2 --local-dir starcoder2-15b-exl2 --local-dir-use-symlinks False
```

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

Linux:

```shell
mkdir starcoder2-15b-exl2-6_5
huggingface-cli download bartowski/starcoder2-15b-exl2 --revision 6_5 --local-dir starcoder2-15b-exl2-6_5 --local-dir-use-symlinks False
```

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

```shell
mkdir starcoder2-15b-exl2-6.5
huggingface-cli download bartowski/starcoder2-15b-exl2 --revision 6_5 --local-dir starcoder2-15b-exl2-6.5 --local-dir-use-symlinks False
```