-
Notifications
You must be signed in to change notification settings - Fork 883
Insights: Lightning-AI/litgpt
Overview
Could not load contribution data
Please try again later
1 Release published by 1 person
-
v0.4.4
published
Jul 11, 2024
12 Pull requests merged by 6 people
-
Add slow interconnect warning
#1577 merged
Jul 15, 2024 -
Chat: dynamically set kv-cache size
#1583 merged
Jul 15, 2024 -
chat/base.py: extend checkpoint_dir before accessing it
#1575 merged
Jul 13, 2024 -
Bump lightning version
#1579 merged
Jul 13, 2024 -
✅ Issue #1557: Fixed warning during model download and conversion.
#1574 merged
Jul 12, 2024 -
Update LitData to latest version 0.2.16
#1573 merged
Jul 11, 2024 -
Bumb version for 0.4.4 release
#1572 merged
Jul 10, 2024 -
Add automatic downloading to CLI
#1571 merged
Jul 10, 2024 -
Fix multi-turn prompting error handling and extra newline
#1569 merged
Jul 10, 2024 -
Add Python API section to 0 to LitGPT docs
#1570 merged
Jul 10, 2024 -
Fix SFTDataset sample generation
#1568 merged
Jul 10, 2024 -
Remove duplicated bos_token for CodeLlama
#1566 merged
Jul 10, 2024
2 Pull requests opened by 1 person
-
Add FlashAttention v3 support
#1578 opened
Jul 12, 2024 -
Make sure quantization is used correctly
#1585 opened
Jul 16, 2024
12 Issues closed by 4 people
-
Extreme slow-down when using FSDP
#607 closed
Jul 15, 2024 -
Why FSDPStrategy is so slow-down when I use multi-machine
#1369 closed
Jul 15, 2024 -
`Chat` consumes more VRAM than `Generate`
#1558 closed
Jul 15, 2024 -
Deal with warnings during model download and conversion
#1557 closed
Jul 15, 2024 -
Upgrade LitData
#1441 closed
Jul 11, 2024 -
Current way masking training data creates issues with the Tokenizer
#1567 closed
Jul 10, 2024 -
`bos_token` is added twice when using CodeLlama prompting style
#1565 closed
Jul 10, 2024 -
Convert bnb pretrained models to Lit-GPT format
#367 closed
Jul 9, 2024 -
Will NTK RoPE be supported?
#482 closed
Jul 9, 2024 -
Add Evaluation with HELM
#349 closed
Jul 9, 2024
2 Issues opened by 2 people
-
Chat: doesn't work with enabled `compilation`
#1584 opened
Jul 15, 2024 -
Out of memory error using Python API but not with CLI
#1582 opened
Jul 15, 2024
7 Unresolved conversations
Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.
-
Gemma 2: `9b` and `27b` versions
#1545 commented on
Jul 12, 2024 • 2 new comments -
how to change dataset path or download url when evaluating
#1556 commented on
Jul 10, 2024 • 0 new comments -
validation output during finetuning
#1443 commented on
Jul 10, 2024 • 0 new comments -
LR scheduler can result in a division by 0
#1393 commented on
Jul 10, 2024 • 0 new comments -
Sample packing for pretraining/fine-tuning
#620 commented on
Jul 15, 2024 • 0 new comments -
Wrong epoch number on last line
#1110 commented on
Jul 15, 2024 • 0 new comments -
Using custom data for `Continue pretraining an LLM`
#1450 commented on
Jul 16, 2024 • 0 new comments