Lightning-AI / litgpt

Pretrain, finetune, deploy 20+ LLMs on your own data. Uses state-of-the-art techniques: flash attention, FSDP, 4-bit, LoRA, and more.

https://lightning.ai

Lightning-AI/litgpt Issues

Continually pretrained Llama2-7B-hf model inference is not working on 16GB GPU machine
Updated 3 days ago4
Stream option
Updated 3 days ago2
Resolve output characters garbled
Updated 4 days ago1
how to pretrain llama2?
Updated 4 days ago1
Python API
Updated 5 days ago
prompt_style
Updated 5 days ago1
Continue pre-training got RuntimeError: Failed processing /tmp/data
Updated 5 days ago4
Address frozen parameter warning with FSDP on nightly torch
Updated 5 days ago1
Lora recipes use lots of memory because of not wrapping parameters with gradient in separate FSDP unit
Closed 5 days ago2
Pretraining example from readme fails in Colab
Updated 6 days ago3
support for qwen2 and baichuan
Updated 6 days ago
'Phi-3-mini-4k-instruct' is not a supported config name
Updated 6 days ago1
Redundancy?
Closed 8 days ago2
Streamline LitGPT API
Updated 9 days ago7
test_tinyllama issue with LitData and `iterate_over_all`
Closed 9 days ago2
Will CycleIterator forward to dataset on resume for pretrain?
Updated 10 days ago1
Remove old and unused LLMs
Updated 11 days ago
Add release workflow
Closed 11 days ago2
LoRA matrices dropout
Updated 11 days ago
how to solve this debug
Updated 12 days ago
LoRA multi-GPU no longer works if applying LoRA selectively
Closed 12 days ago3
Cannot copy out of meta tensor; no data!
Closed 12 days ago2
LR scheduler can result in a division by 0
Updated 12 days ago
fabric.print only works on sys.stderr, does not print inference result
Updated 13 days ago1
Customizable loss function & inference step?
Closed 13 days ago2
ValueError: Cannot attend to 3063, block size is only 2048
Closed 13 days ago1
How to use custom dataset for evaluate?
Updated 15 days ago
tokenizer.py
Closed 16 days ago2
litgpt download doesn't work
Closed 19 days ago7
How to specify which GPU to use？
Closed 18 days ago2
After some iteration in pretraining a LLM, IndexError is raised related to dataset chunking
Updated 18 days ago
Why FSDPStrategy is so slow-down when I use multi-machine
Updated 19 days ago4
A potential bug for multi-GPU training
Updated 19 days ago5
Failed to load the finetuned model with `AutoModelForCausalLM.from_pretrained(name, state_dict=state_dict)`
Updated 19 days ago4
Add support for memory-efficient and faster optimizers
Updated 19 days ago1
combine FSDP with selective activation checkpointing
Updated 21 days ago
OOM Error: CUDA out of memory when finetuning llama3-8b
Closed 22 days ago3
The `litgpt evaluate` command attempts to download config files from gated repos
Closed 23 days ago1
Conversion to HF checkpoint should generate a checkpoint format that can be loaded directly
Updated 23 days ago1
ValueError: 'Meta-Llama-3-8B-Instruct' is not a supported config name
Closed 23 days ago4
--checkpoint-dir 'xx' is missing the files: ['model_config.yaml']
Closed 24 days ago2
merging intermediate adapter
Closed 24 days ago
Nucleus (top-p) sampling
Updated 25 days ago2
Continual pretraining for custom data is not working. Not recognizing TextFiles as a data attribute.
Closed 25 days ago3
Add support for phi-3-mini
Closed 25 days ago2
Finetune and Pretrain Path Discrepancy
Closed 25 days ago2
Blockwise quantization only supports 16/32-bit floats, but got torch.uint8 ( `bnb.nf4` quantisation is not working)
Updated 25 days ago15
Qwen1.5 Family Support
Updated a month ago
Pretraining example is not working
Updated a month ago1
simple API interface for initializing and running model inference
Updated a month ago