Spaces:
Sleeping
Sleeping
Devdan Schretlen
Claude Opus 4.8
Add MicroGPT chat demo: FastAPI + React app, Dockerfile, gpt3.1 model
8b12a0a | """ | |
| Model architectures and the chat tokenizer. | |
| Re-exports the public surface so callers can write `from models import GPT, | |
| GPT3, get_encoding, ...` without reaching into the individual submodules: | |
| * gpt2.py -- GPT-2 (124M): GPT / GPTConfig | |
| * gpt3.py -- GPT-3-style (254M): GPT3 / GPT3Config | |
| * tokenizer.py -- GPT-2 tiktoken encoding extended with chat special tokens | |
| """ | |
| from .gpt2 import GPT, GPTConfig | |
| from .gpt3 import GPT3, GPT3Config | |
| from .tokenizer import SPECIAL_TOKENS, get_encoding | |
| __all__ = [ | |
| "GPT", | |
| "GPTConfig", | |
| "GPT3", | |
| "GPT3Config", | |
| "SPECIAL_TOKENS", | |
| "get_encoding", | |
| ] | |