---
base_model: [codellama/CodeLlama-70b-Instruct-hf]
tags:
- mergekit
- merge
- code
license: mit
pipeline_tag: conversational
---
# BigCodeLLama 92b GGUF files 🚀

## Experimental 92B CodeLlaMA that should be better than stock

### Models Merged with base ```codellama/CodeLlama-70b-Instruct-hf```

Full model here: https://huggingface.co/nisten/BigCodeLlama-92b

### Models Merged

The following models were included in the merge:
* ../CodeLlama-70b-Python-hf
* ../CodeLlama-70b-Instruct-hf

### Configuration

The following YAML configuration was used to produce this model:

```yaml
dtype: bfloat16
merge_method: passthrough
slices:
- sources:
  - layer_range: [0, 69]
    model:
      model:
        path: ../CodeLlama-70b-Instruct-hf
- sources:
  - layer_range: [42, 80]
    model:
      model:
        path: ../CodeLlama-70b-Python-hf
```

To merge together the 6bit for example download both parts then do
```bash
cat BigCodeLlama-92b-q6.gguf.part0 BigCodeLlama-92b-q6.gguf.part1 > BigCodeLlama-92b-q6.gguf
```

Comparison over stock with question:

```bash
Plan and write code for building a city on mars via calculating aldrin cycler orbits in js for cargo shipments starting in year 2030, and after coding it in python and c++ output a table of calendar of deliver dates.

Don't ask for clarification just do the work smartly.
```

![image/png](https://cdn-uploads.huggingface.co/production/uploads/6379683a81c1783a4a2ddba8/59mxGRA4GrqPGX3vqmzMa.png)

and our 6bit quant


![image/png](https://cdn-uploads.huggingface.co/production/uploads/6379683a81c1783a4a2ddba8/okdhnUzTCTMYhCKlEEcOn.png)