File size: 1,103 Bytes
64eff2e
94a9749
 
 
 
 
64eff2e
94a9749
 
64eff2e
94a9749
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
language:
- en
tags:
- pytorch
- causal-lm
license: apache-2.0
datasets:
- EleutherAI/pile
---
This is a d-Matrix functional reference of the [GPTJ-6B](https://huggingface.co/EleutherAI/gpt-j-6b) model.

The reference provides the following functional *configurations*:
  Configuration | Explanation
  :-- | :-- 
  **`BASELINE`** | a reference functionally equivalent to the original model
  **`BASIC`** | all linear algebraic operands quantized to `BFP16-64`, and all other operations transformed to approximated kernel simulations


### Usage

Install d-Matrix [ML Tools](https://github.com/d-matrix-ai/dmx-mltools) first.

```sh
pip install dmx-mltools
```

The following is an example model and its evaluation.  

```python
from mltools.dmx import pipeline

pipe = pipeline(
    task="text-generation",
    model="d-matrix/gpt-j-6b",
    dmx_config="BASELINE",  # see above for other variants
    trust_remote_code=True,
    # device_map="auto",  # enabling model parallel on multi-GPU nodes
)

results = pipe.evaluate(metric=["d-matrix/perplexity", "d-matrix/accuracy"], dataset="LAMBADA")
```