tysam-code / hlb-gptLinks
Minimalistic, extremely fast, and hackable researcher's toolbench for GPT models in 307 lines of code. Reaches <3.8 validation loss on wikitext-103 on a single A100 in <100 seconds. Scales to larger models with one parameter change (feature currently in alpha).
☆345Updated 10 months ago
Alternatives and similar repositories for hlb-gpt
Users that are interested in hlb-gpt are comparing it to the libraries listed below
Sorting:
- a small code base for training large models☆301Updated last month
- Puzzles for exploring transformers☆349Updated 2 years ago
- Fast bare-bones BPE for modern tokenizer training☆159Updated 2 months ago
- A crude RLHF layer on top of nanoGPT with Gumbel-Softmax trick☆290Updated last year
- ☆92Updated last year
- An interactive exploration of Transformer programming.☆264Updated last year
- JAX implementation of the Llama 2 model☆218Updated last year
- ☆143Updated 2 years ago
- Simple Transformer in Jax☆137Updated last year
- A puzzle to learn about prompting☆128Updated 2 years ago
- ☆270Updated 11 months ago
- Language Modeling with the H3 State Space Model☆519Updated last year
- Extract full next-token probabilities via language model APIs☆247Updated last year
- ☆415Updated last year
- Visualize the intermediate output of Mistral 7B☆367Updated 5 months ago
- Visualizing the internal board state of a GPT trained on chess PGN strings, and performing interventions on its internal board state and …☆206Updated 7 months ago
- Ungreedy subword tokenizer and vocabulary trainer for Python, Go & Javascript☆584Updated 11 months ago
- Fully fine-tune large models like Mistral, Llama-2-13B, or Qwen-14B completely for free☆231Updated 7 months ago
- Fine-tune mistral-7B on 3090s, a100s, h100s☆714Updated last year
- Inference code for Persimmon-8B☆415Updated last year
- A MAD laboratory to improve AI architecture designs 🧪☆120Updated 6 months ago
- batched loras☆343Updated last year
- Fast & Simple repository for pre-training and fine-tuning T5-style models☆1,005Updated 10 months ago
- Draw more samples☆191Updated last year
- Legible, Scalable, Reproducible Foundation Models with Named Tensors and Jax☆595Updated this week
- A really tiny autograd engine☆94Updated 3 weeks ago
- Erasing concepts from neural representations with provable guarantees☆228Updated 4 months ago
- git extension for {collaborative, communal, continual} model development☆213Updated 7 months ago
- Full finetuning of large language models without large memory requirements☆94Updated last year
- seqax = sequence modeling + JAX☆159Updated last week