abacaj / train-with-fsdpLinks
☆94Updated last year
Alternatives and similar repositories for train-with-fsdp
Users that are interested in train-with-fsdp are comparing it to the libraries listed below
Sorting:
- Comprehensive analysis of difference in performance of QLora, Lora, and Full Finetunes.☆83Updated 2 years ago
- Fully fine-tune large models like Mistral, Llama-2-13B, or Qwen-14B completely for free☆232Updated 10 months ago
- Full finetuning of large language models without large memory requirements☆94Updated last year
- ☆46Updated last year
- ☆22Updated 2 years ago
- Minimal example scripts of the Hugging Face Trainer, focused on staying under 150 lines☆197Updated last year
- Just a bunch of benchmark logs for different LLMs☆119Updated last year
- QLoRA with Enhanced Multi GPU Support☆37Updated 2 years ago
- Multipack distributed sampler for fast padding-free training of LLMs☆199Updated last year
- Large scale 4D parallelism pre-training for 🤗 transformers in Mixture of Experts *(still work in progress)*☆87Updated last year
- experiments with inference on llama☆104Updated last year
- ☆416Updated last year
- ☆96Updated 2 years ago
- Manage scalable open LLM inference endpoints in Slurm clusters☆271Updated last year
- Exploring finetuning public checkpoints on filter 8K sequences on Pile☆116Updated 2 years ago
- Fast & more realistic evaluation of chat language models. Includes leaderboard.☆188Updated last year
- Multi-Domain Expert Learning☆67Updated last year
- Reimplementation of the task generation part from the Alpaca paper☆119Updated 2 years ago
- Doing simple retrieval from LLM models at various context lengths to measure accuracy☆102Updated last year
- Helpers and such for working with Lambda Cloud☆51Updated last year
- Simplex Random Feature attention, in PyTorch☆74Updated last year
- batched loras☆345Updated 2 years ago
- Code repository for the c-BTM paper☆107Updated last year
- ☆199Updated last year
- Finetune Falcon, LLaMA, MPT, and RedPajama on consumer hardware using PEFT LoRA☆104Updated 3 months ago
- git extension for {collaborative, communal, continual} model development☆216Updated 9 months ago
- Minimalistic, extremely fast, and hackable researcher's toolbench for GPT models in 307 lines of code. Reaches <3.8 validation loss on wi…☆349Updated last year
- ModuleFormer is a MoE-based architecture that includes two different types of experts: stick-breaking attention heads and feedforward exp…☆223Updated last year
- NeurIPS Large Language Model Efficiency Challenge: 1 LLM + 1GPU + 1Day☆256Updated last year
- An introduction to LLM Sampling☆80Updated 8 months ago