34 17

Quazimoto PRO

Quazim0t0

https://discord.gg/eeDfVF4FNe

SeanceTable

AI & ML interests

the hunchback of huggingface 🔙 joined: 1-20-2025 🦥unsloth user 4️⃣ Phi User 🔨 ai hobbyist 📫 On Leaderboards Top 100-200

Recent Activity

reacted to AtAndDev's post with 👍 about 6 hours ago

There seems to multiple paid apps shared here that are based on models on hf, but some ppl sell their wrappers as "products" and promote them here. For a long time, hf was the best and only platform to do oss model stuff but with the recent AI website builders anyone can create a product (really crappy ones btw) and try to sell it with no contribution to oss stuff. Please dont do this, or try finetuning the models you use... Sorry for filling yall feed with this bs but yk...

reacted to burtenshaw's post with 🤗 1 day 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

posted an update 3 days ago

Thank you to the Open LLM Leaderboard's team for offering it to the community for as long as they did. I only recently joined HF, and it provided a lot of incentive and information to make better models. Always will remember getting to #112 :D Anyone have a solid way to test my models privately? Please let me know!

View all activity

Organizations

Posts 4

Post

966

Thank you to the Open LLM Leaderboard's team for offering it to the community for as long as they did. I only recently joined HF, and it provided a lot of incentive and information to make better models.

Always will remember getting to #112 :D

Anyone have a solid way to test my models privately? Please let me know!

Post

541

Update to the Imagine side-project.
Just uploaded the 16Bit & Q4

Samples: (Used a base Microsoft Phi4 model)
*You may experience bugs with either the model or the Open WebUI function*
Open WebUI function: https://openwebui.com/f/quaz93/imagine_phi
Quazim0t0/Imagine-v0.5-16bit - Haven't tested
Quazim0t0/ImagineTest-v0.5-GGUF - Tested (Pictures)

Dataset: Quazim0t0/Amanita-Imagine
Small Dataset of 500+ entries, still working on it here and there when I can.
Pictures use the Open Web UI function I provided.

View all Posts