---
library_name: lucid
license: mit
tags:
  - causal-lm
  - gpt2
  - lucid
datasets:
  - webtext
pipeline_tag: text-generation
---

# GPT-2 Medium (causal-LM)

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

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

## Available weights

| Tag | Params | GFLOPs | Size | Source |
|---|---|---|---|---|
| `WEBTEXT` *(default)* | 354.8M | — | 1549.89 MB | transformers |

## Usage

```python
import lucid
import lucid.models as models
from lucid.models.weights import GPT2MediumLMWeights

# default tag
model = models.gpt2_medium_lm(pretrained=True)

# explicit tag (enum or string)
model = models.gpt2_medium_lm(weights=GPT2MediumLMWeights.WEBTEXT)
model = models.gpt2_medium_lm(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)
logits = out.logits  # (B, T, vocab_size)
```

## Conversion

Converted from `transformers/gpt2-medium` via
`python -m tools.convert_weights gpt2_medium_lm --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).
```