Article
LeRobot goes to driving school: World’s largest open-source self-driving dataset
•
40
Brigitte Tousignant
BrigitteTousi
AI & ML interests
None yet
Recent Activity
replied to
burtenshaw's
post
about 10 hours ago
everybody and their dog is fine-tuning Gemma 3 today, so I thought I'd do a longer post on the tips and sharp edges I find. let's go!
1. has to be install everything form main and nightly. this is what I'm working with to get unsloth and TRL running
```txt
git+https://github.com/huggingface/transformers@main
git+https://github.com/huggingface/trl.git@main
bitsandbytes
peft
```
plus this with `--no-deps`
```txt
git+https://github.com/unslothai/unsloth-zoo.git@nightly
git+https://github.com/unslothai/unsloth.git@nightly
```
2. will brown's code to turn GSM8k into a reasoning dataset is a nice toy experiment https://gist.github.com/willccbb/4676755236bb08cab5f4e54a0475d6fb
3. with a learning rate of 5e-6 rewards and loss stayed flat for the first 100 or so steps.
4. so far none of my runs have undermined the outputs after 1 epoch. therefore, I'm mainly experimenting with bigger LoRA adapters.
```python
from trl import GRPOConfig
training_args = GRPOConfig(
learning_rate = 5e-6,
adam_beta1 = 0.9,
adam_beta2 = 0.99,
weight_decay = 0.1,
warmup_ratio = 0.1,
lr_scheduler_type = "cosine",
optim = "adamw_8bit",
logging_steps = 1,
per_device_train_batch_size = 2,
gradient_accumulation_steps = 1,
num_generations = 2,
max_prompt_length = 256,
max_completion_length = 1024 - 256,
num_train_epochs = 1,
max_steps = 250,
save_steps = 250,
max_grad_norm = 0.1,
report_to = "none",
)
```
5. vision fine-tuning isn't available in TRL's GRPOTrainer, so stick to text datasets. but no need to load the model differently in transformers or Unsloth
```python
from transformers import AutoModelForImageTextToText
model = AutoModelForImageTextToText.from_pretrained("google/gemma-3-4b-it)
```
if you want an introduction to GRPO, check out the reasoning course, it walks you through the algorithm, theory, and implementation in a smooth way.
https://huggingface.co/reasoning-course
reacted
to
burtenshaw's
post
with 🔥
about 10 hours ago
everybody and their dog is fine-tuning Gemma 3 today, so I thought I'd do a longer post on the tips and sharp edges I find. let's go!
1. has to be install everything form main and nightly. this is what I'm working with to get unsloth and TRL running
```txt
git+https://github.com/huggingface/transformers@main
git+https://github.com/huggingface/trl.git@main
bitsandbytes
peft
```
plus this with `--no-deps`
```txt
git+https://github.com/unslothai/unsloth-zoo.git@nightly
git+https://github.com/unslothai/unsloth.git@nightly
```
2. will brown's code to turn GSM8k into a reasoning dataset is a nice toy experiment https://gist.github.com/willccbb/4676755236bb08cab5f4e54a0475d6fb
3. with a learning rate of 5e-6 rewards and loss stayed flat for the first 100 or so steps.
4. so far none of my runs have undermined the outputs after 1 epoch. therefore, I'm mainly experimenting with bigger LoRA adapters.
```python
from trl import GRPOConfig
training_args = GRPOConfig(
learning_rate = 5e-6,
adam_beta1 = 0.9,
adam_beta2 = 0.99,
weight_decay = 0.1,
warmup_ratio = 0.1,
lr_scheduler_type = "cosine",
optim = "adamw_8bit",
logging_steps = 1,
per_device_train_batch_size = 2,
gradient_accumulation_steps = 1,
num_generations = 2,
max_prompt_length = 256,
max_completion_length = 1024 - 256,
num_train_epochs = 1,
max_steps = 250,
save_steps = 250,
max_grad_norm = 0.1,
report_to = "none",
)
```
5. vision fine-tuning isn't available in TRL's GRPOTrainer, so stick to text datasets. but no need to load the model differently in transformers or Unsloth
```python
from transformers import AutoModelForImageTextToText
model = AutoModelForImageTextToText.from_pretrained("google/gemma-3-4b-it)
```
if you want an introduction to GRPO, check out the reasoning course, it walks you through the algorithm, theory, and implementation in a smooth way.
https://huggingface.co/reasoning-course
Organizations
BrigitteTousi's activity
upvoted
an
article
2 days ago
upvoted
an
article
9 days ago
Article
A Deepdive into Aya Vision: Advancing the Frontier of Multilingual Multimodality
•
65
upvoted
an
article
10 days ago
Article
SmolVLM - small yet mighty Vision Language Model
•
220
upvoted
an
article
13 days ago
Article
Smol but Mighty: Can Small Models Reason well? 🤔
By
•
•
9
upvoted
a
paper
16 days ago
upvoted
an
article
16 days ago
Article
FastRTC: The Real-Time Communication Library for Python
•
141
upvoted
3
articles
about 1 month ago
Article
Open-source DeepResearch – Freeing our search agents
•
1.16k
Article
Open-R1: Update #1
By
and 7 others
•
•
295Article
Open-R1: a fully open reproduction of DeepSeek-R1
•
803
upvoted
an
article
about 2 months ago
Article
Hugging Face and FriendliAI partner to supercharge model deployment on the Hub
•
36
upvoted
3
articles
2 months ago
Article
🌁#82: AI and ML in Real Life
By
•
•
16Article
Crowd-sourced Open Preference Dataset for Text-to-Image Generation
By
and 4 others
•
•
18Article
🌁#81: Key AI Concepts to Follow in 2025
By
•
•
24
upvoted
2
articles
3 months ago
Article
Unveiling CIVICS: A New Dataset for Examining Cultural Values in Language Models
By
•
•
11Article
🇪🇺✍️ EU AI Act: Systemic Risks in the First CoP Draft Comments ✍️🇪🇺
By
and 1 other
•
•
14
upvoted
2
papers
3 months ago
upvoted
an
article
3 months ago
Article
Use Models from the Hugging Face Hub in LM Studio
By
•
•
138
upvoted
a
collection
5 months ago
upvoted
an
article
6 months ago
Article
Getty Images Brings High-Quality, Commercially Safe Dataset to Hugging Face
By
•
•
16