philipp-zettl/inaturalist-tool
2.96 MB
PZ
philipp-zettl
AI & ML interests
NLP/CV/Multimodal learning
Recent Activity
reacted to danielhanchen's post with 🔥 1 day ago
A new way to use Unsloth.
Coming soon... posted an update 1 day ago
I'm unemployed, I have a gaming GPU, and I just published a German LLM.
qwen3-0.6b-german - fine-tuned Qwen3-0.6B in ~40h on an RTX 4070 Ti, using the exact same instruct datasets as the LLäMmlein paper (ACL 2025).
HellaSwag-DE: 0.3111 → 0.3193 ✅
ARC-DE: 0.2352 → 0.2575 ✅
MMlu-DE: 0.3600 → 0.2475 🔻 (alignment tax - known trade-off)
Instruction fine-tuning trades some factual breadth for better reasoning and format following. The model is more useful, even if not better on every metric.
Weights, LoRA adapter, full training script and logs all public.
https://huggingface.co/philipp-zettl/qwen3-0.6b-german
It ain't much, but it's honest work. updated a dataset 1 day ago
philipp-zettl/qwen3-0.6b-german-trainingOrganizations
Posts 6
Post
99
I'm unemployed, I have a gaming GPU, and I just published a German LLM.
qwen3-0.6b-german - fine-tuned Qwen3-0.6B in ~40h on an RTX 4070 Ti, using the exact same instruct datasets as the LLäMmlein paper (ACL 2025).
HellaSwag-DE: 0.3111 → 0.3193 ✅
ARC-DE: 0.2352 → 0.2575 ✅
MMlu-DE: 0.3600 → 0.2475 🔻 (alignment tax - known trade-off)
Instruction fine-tuning trades some factual breadth for better reasoning and format following. The model is more useful, even if not better on every metric.
Weights, LoRA adapter, full training script and logs all public.
philipp-zettl/qwen3-0.6b-german
It ain't much, but it's honest work.
qwen3-0.6b-german - fine-tuned Qwen3-0.6B in ~40h on an RTX 4070 Ti, using the exact same instruct datasets as the LLäMmlein paper (ACL 2025).
HellaSwag-DE: 0.3111 → 0.3193 ✅
ARC-DE: 0.2352 → 0.2575 ✅
MMlu-DE: 0.3600 → 0.2475 🔻 (alignment tax - known trade-off)
Instruction fine-tuning trades some factual breadth for better reasoning and format following. The model is more useful, even if not better on every metric.
Weights, LoRA adapter, full training script and logs all public.
philipp-zettl/qwen3-0.6b-german
It ain't much, but it's honest work.
Post
1206
alias rm='rm -i'Better be safe than sorry.
spaces 35
pinned
Sleeping
Tiny Dy Svit CIFAR10
📊
Classify images and see importance heatmap
pinned
Running
Structured Output Generator
📚
Generate Your Structured Output HERE
pinned
Sleeping
Roast My Model
💬
Retrieve a naive baseline model for your ML use case.
pinned
Sleeping
Replace Newlines
🔥
Replace newlines with escape sequences
pinned
Running
food-dora
🧑
Describe what you want to cook, receive the recipe for it.
pinned
Runtime error
QA/FAQ Generator
📈
Generates Questions and Answers from given text content.
models 72
philipp-zettl/qwen3-0.6b-german
Text Generation • Updated • 400
philipp-zettl/qwen3-0.6b-german-merged
0.6B • Updated • 14
philipp-zettl/qwen3.5-0.8b-german
Updated
philipp-zettl/modernbert-diffusion-openwebtext
Fill-Mask • 0.1B • Updated
philipp-zettl/Dy-SViT-CIFAR10
Image Classification • Updated
philipp-zettl/modernbert-diffusion-universal
Fill-Mask • 0.1B • Updated
philipp-zettl/modernbert-diffusion-refactor
Fill-Mask • 0.1B • Updated
philipp-zettl/modernbert-diffusion-alpaca-ft
Fill-Mask • 0.1B • Updated
philipp-zettl/modernbert-diffusion-code
Fill-Mask • 0.1B • Updated
philipp-zettl/modernbert-diffusion-instruct
Fill-Mask • 0.1B • Updated
datasets 16
philipp-zettl/qwen3-0.6b-german-training
Viewer • Updated • 1.6k • 6
philipp-zettl/inaturalist-bbs
Updated • 57
philipp-zettl/inaturalist-enriched
Viewer • Updated • 1.49M • 145
philipp-zettl/inaturalist-s3-massive
Viewer • Updated • 1.49M • 1.16k
philipp-zettl/pile-of-law_atticus_contracts
Viewer • Updated • 488k • 82
philipp-zettl/pile-of-law
Updated • 206
philipp-zettl/DeepJSONEval
Viewer • Updated • 2.1k • 14
philipp-zettl/MTGEmb-small-embs
Viewer • Updated • 59.8k • 7
philipp-zettl/my_first_lora_v1-dataset
Viewer • Updated • 7 • 6
philipp-zettl/NibbleNix-DE
Viewer • Updated • 4M • 15