Andrei-Aksionov

`open in google colab` integration

3

HI there 👋 As a way to make this repo more accessible how about to add notebooks with  button? For now there is a `howto` folder with...

Adapter small fix

4

Hi there 👋 As @carmocca mentioned in [PR #352](https://github.com/Lightning-AI/lit-llama/pull/352) some code changes need to be done: 1. Change `self.n_embd` --> `C` since this value is extracted from the shape of...

Hi there 👋 While I was working on #1341, I noticed a couple of strange things with the tokenizer: 1. If to make a prompt a bit more complex (by...

### System Info > [!Important] > The issue only exists in MacOS Hi there 👋 If I want to install `hf-transfer` (0.1.6) in a venv with Python 3.9.19 I have...

Hi there, A year or so ago I tried to implement my own transformer module and used this repository as a reference. Recently I decided to revisit it in order...

AutoTokenizer: Phi-3 drops spaces when decodes a token at a time

1

### System Info - `transformers` version: 4.41.2 - Platform: macOS-14.5-x86_64-i386-64bit - Python version: 3.11.6 - Huggingface_hub version: 0.23.4 - Safetensors version: 0.4.3 - Accelerate version: 0.31.0 - Accelerate config: not...

Gemma 2: `9b` and `27b` versions

8

Hi there 👋 Fixes #1535 Google [released](https://blog.google/technology/developers/google-gemma-2/) the latest and greatest Gemma model - v2. This time it comes in three sizes: - 2b (not yet released) - 9b -...

Deal with warnings during model download and conversion

3

When we download (and automatically convert) a model, we have these three warnings: 1. .../huggingface_hub/file_download.py:1132: FutureWarning: `resume_download` is deprecated and will be removed in version 1.0.0. Downloads always resume when...

`Chat` consumes more VRAM than `Generate`

7

### Bug description Hi there 👋 While working on integration of Gemma 2 (9b variant) I noticed that while a regular generate script barely fits into L4 with 24 GB,...

chat/base.py: extend checkpoint_dir before accessing it

7

Hi there 👋 Inside `auto_download_checkpoint` the code also extends the path by prepending `checkpoints/` if it's not provided. Since currently it's done after `check_file_size_on_cpu_and_warn`, the command ```bash litgpt chat google/gemma-2-9b-it...