uv-scripts/training
Streaming LLM Training with Unsloth Train on massive datasets without downloading anything - data streams directly from the Hub. 🦥 Latin LLM Example Teaches Qwen Latin using 1.47M texts from FineWeb-2, streamed directly from the Hub. Blog post: Train on Massive Datasets Without Downloading Quick Start # Run on HF Jobs (recommended - 2x faster streaming) hf jobs uv run latin-llm-streaming.py \ --flavor a100-large \ --timeout 2h \ --secrets… See the full description on the dataset page: https://huggingface.co/datasets/uv-scripts/training.
Upload vlm-test-no-trackio.py with huggingface_hub
Upload vlm-streaming-sft-unsloth-qwen.py with huggingface_hub
Upload vlm-streaming-sft-unsloth-qwen.py with huggingface_hub
Update VLM docs with verified Qwen3-VL config and performance
Fix VLM detection error: match official Unsloth notebook config
Upload vlm-streaming-sft-unsloth-qwen.py with huggingface_hub
Upload vlm-streaming-sft-unsloth-qwen.py with huggingface_hub
Upload vlm-streaming-sft-unsloth-qwen.py with huggingface_hub
Upload vlm-streaming-sft-unsloth-qwen.py with huggingface_hub
Upload vlm-streaming-sft-unsloth-qwen.py with huggingface_hub
Upload vlm-streaming-sft-unsloth-qwen.py with huggingface_hub
Upload vlm-streaming-sft-unsloth-qwen.py with huggingface_hub
Upload vlm-streaming-sft-unsloth.py with huggingface_hub
Upload vlm-streaming-sft-unsloth.py with huggingface_hub
Upload vlm-streaming-sft-unsloth.py with huggingface_hub
Upload vlm-streaming-sft-unsloth.py with huggingface_hub
Upload vlm-streaming-sft-unsloth.py with huggingface_hub
Update README for streaming training script
Add Latin LLM streaming training script
Update dependency comments for clarity and consistency
Remove unused dependencies from the script
Reorder unsloth import for optimization clarity
deps
deps
Fix typo in dependency name for huggingface_hub
deps
deps
deps
deps
Strip dependencies for Unsloth Docker image
update deps
Add xformers dependency
Install transformers from git for Qwen3-VL support
Fix default eval split to 'test'
Fix dependency versions and import order
Switch to pure UV approach without Docker image
Add dataset metadata to disable viewer
Add Unsloth VLM training script for Iconclass
initial commit