Djuunaa
djuna
AI & ML interests
None yet
Recent Activity
reacted
to
danielhanchen's
post
with π₯
2 days ago
π¦₯ Introducing Unsloth Dynamic v2.0 GGUFs!
Our v2.0 quants set new benchmarks on 5-shot MMLU and KL Divergence, meaning you can now run & fine-tune quantized LLMs while preserving as much accuracy as possible.
Llama 4: https://huggingface.co/unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF
DeepSeek-R1: https://huggingface.co/unsloth/DeepSeek-R1-GGUF-UD
Gemma 3: https://huggingface.co/unsloth/gemma-3-27b-it-GGUF
We made selective layer quantization much smarter. Instead of modifying only a subset of layers, we now dynamically quantize all layers so every layer has a different bit. Now, our dynamic method can be applied to all LLM architectures, not just MoE's.
Blog with Details: https://docs.unsloth.ai/basics/dynamic-v2.0
All our future GGUF uploads will leverage Dynamic 2.0 and our hand curated 300Kβ1.5M token calibration dataset to improve conversational chat performance.
For accurate benchmarking, we built an evaluation framework to match the reported 5-shot MMLU scores of Llama 4 and Gemma 3. This allowed apples-to-apples comparisons between full-precision vs. Dynamic v2.0, QAT and standard iMatrix quants.
Dynamic v2.0 aims to minimize the performance gap between full-precision models and their quantized counterparts.
new activity
10 days ago
zed-industries/zeta:Need Example for inference code
liked
a model
10 days ago
aipgpt/Txt-Polisher-Douyin-Style
Organizations
Collections
1
spaces
6
Running
Gradient Array Slider
π’
Gradient Array Slider (for mergekit)
Paused
1
Replete Qwen2 1.5B Instruct
π¬
Running
Llama Pro Dry Run
β‘
Calculate duplicated neural network layers
Running
3
Chat Template Tester
π
Experiment with Chat Template
Running
16
mergekit-gui
π
Merge and upload models using a YAML config
Running
3
Generate Mergkit For Upscale LLM
π
Generate YAML config from model parameters and layers
models
98
djuna/DeepSeek-R1-Distill-Qwen-14B-abliterated-v2-remap
Text Generation
β’
Updated
β’
5
β’
2
djuna/DeepSeek-R1-Distill-Qwen-14B-abliterated-remap
Text Generation
β’
Updated
β’
6
β’
1
djuna/MN-Chinofun-12B-4-4bit
Text Generation
β’
Updated
β’
2
djuna/TEST3-Q2.5-Lenned-14B-Q5_K_M-GGUF
Updated
β’
4
djuna/TEST3-Q2.5-Lenned-14B
Text Generation
β’
Updated
β’
4
β’
1
djuna/TEST2-Q2.5-Lenned-14B-Q5_K_M-GGUF
Updated
β’
19
β’
1
djuna/TEST2-Q2.5-Lenned-14B
Text Generation
β’
Updated
β’
8
β’
4
djuna/TEST-Q2.5-Lenned-14B
Text Generation
β’
Updated
β’
20
β’
1
djuna/MN-Chinofun-12B-4
Text Generation
β’
Updated
β’
7
β’
3
djuna/MN-Chinofun-12B-4.1-Q6_K-GGUF
Updated
β’
10
β’
1
datasets
0
None public yet