File size: 1,407 Bytes
facb666
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
---
library_name: lucid
license: mit
tags:
  - base
  - gpt2
  - lucid
datasets:
  - webtext
pipeline_tag: feature-extraction
---

# GPT-2 XL

> https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf

[Lucid](https://github.com/ChanLumerico/lucid) port of `transformers/gpt2-xl`,
converted to Lucid-native safetensors.

## Available weights

| Tag | Params | GFLOPs | Size | Source |
|---|---|---|---|---|
| `WEBTEXT` *(default)* | 1557.6M | — | 5941.87 MB | transformers |

## Usage

```python
import lucid
import lucid.models as models
from lucid.models.weights import GPT2XLargeWeights

# default tag
model = models.gpt2_xlarge(pretrained=True)

# explicit tag (enum or string)
model = models.gpt2_xlarge(weights=GPT2XLargeWeights.WEBTEXT)
model = models.gpt2_xlarge(pretrained="WEBTEXT")

# feed token ids (tokenize with the matching lucid.utils.tokenizer)
input_ids = lucid.tensor([[101, 7592, 2088, 102]], dtype=lucid.int64)
out = model(input_ids)
hidden = out.last_hidden_state  # (B, T, hidden_size)
```

## Conversion

Converted from `transformers/gpt2-xl` via
`python -m tools.convert_weights gpt2_xlarge --tag WEBTEXT`.
Key mapping + numerical parity verified against the source.

## License

`mit` — inherited from the original weights.

## Citation

```
Radford et al., "Language Models are Unsupervised Multitask Learners", 2019 (GPT-2).
```