File size: 1,245 Bytes
07c718f
 
f5e8289
 
 
 
 
 
 
 
 
 
 
07c718f
f5e8289
 
 
b70666d
 
 
 
f5e8289
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: mit
datasets:
- databricks/databricks-dolly-15k
language:
- en
pipeline_tag: text-generation
tags:
- dolly
- dolly-v2
- instruct
- sharded
inference: False
---

# dolly-v2-12b: sharded checkpoint 

<a href="https://colab.research.google.com/gist/pszemraj/6eb7ccce28ea6aa07b8ec86388ac010e/sharded-instruction-model-playground.ipynb">
  <img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"/>
</a>

This is a sharded checkpoint (with ~4GB shards) of the `databricks/dolly-v2-12b` model. Refer to the [original model](https://huggingface.co/databricks/dolly-v2-12b) for all details.

- this enables low-RAM loading, i.e. Colab :)

## Basic Usage


install `transformers`, `accelerate`, and `bitsandbytes`.

```bash
pip install -U -q transformers bitsandbytes accelerate
```

Load the model in 8bit, then [run inference](https://huggingface.co/docs/transformers/generation_strategies#contrastive-search):

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

model_name = "ethzanalytics/dolly-v2-12b-sharded"
tokenizer = AutoTokenizer.from_pretrained(model_name)

model = AutoModelForCausalLM.from_pretrained(
          model_name, load_in_8bit=True, device_map="auto",
        )
```