John Leimgruber III's picture

John Leimgruber III

ubergarm

·

https://emptyduck.com

ubergarm

AI & ML interests

Open LLMs and Astrophotography image processing.

Recent Activity

new activity about 3 hours ago

unsloth/DeepSeek-R1-GGUF:Over 2 tok/sec agg backed by NVMe SSD on 96GB RAM + 24GB VRAM AM5 rig with llama.cpp

new activity about 21 hours ago

unsloth/DeepSeek-R1-GGUF:Got it running after downloading some RAM!

new activity 1 day ago

mradermacher/Qwen2.5-14B-Instruct-1M-i1-GGUF:Over 128k context on 1x 3090 TI FE 24GB VRAM!

View all activity

Organizations

None yet

ubergarm's activity

New activity in unsloth/DeepSeek-R1-GGUF about 3 hours ago

Over 2 tok/sec agg backed by NVMe SSD on 96GB RAM + 24GB VRAM AM5 rig with llama.cpp

#13 opened about 21 hours ago by

New activity in unsloth/DeepSeek-R1-GGUF about 21 hours ago

Got it running after downloading some RAM!

#7 opened 2 days ago by

New activity in mradermacher/Qwen2.5-14B-Instruct-1M-i1-GGUF 1 day ago

Over 128k context on 1x 3090 TI FE 24GB VRAM!

#1 opened 1 day ago by

New activity in unsloth/DeepSeek-R1-GGUF 1 day ago

Inference speed

#9 opened 2 days ago by

New activity in HKUSTAudio/Llasa-3B 3 days ago

Control over output

#12 opened 3 days ago by

TeachableMachine

New activity in srinivasbilla/llasa-3b-tts 4 days ago

Emotions

#3 opened 5 days ago by

New activity in jinaai/ReaderLM-v2 9 days ago

What advantage does this have over normal algorithmic ways of turning HTML to Markdown ?

#5 opened 14 days ago by

New activity in bartowski/DeepSeek-R1-Distill-Llama-70B-GGUF 9 days ago

FuseO1-DeekSeekR1-QwQ-SkyT1-32B-Preview

#1 opened 10 days ago by

New activity in deepseek-ai/DeepSeek-R1-Distill-Qwen-32B 10 days ago

System Prompt

#2 opened 10 days ago by

New activity in bartowski/DeepSeek-R1-Distill-Qwen-32B-GGUF 10 days ago

FIXED: Error with llama-server `unknown pre-tokenizer type: 'deepseek-r1-qwen'`

#1 opened 10 days ago by

New activity in unsloth/DeepSeek-R1-Distill-Qwen-32B-bnb-4bit 10 days ago

The `tokenizer_config.json` is missing the `chat_template` jinja?

#1 opened 10 days ago by

New activity in LatitudeGames/Wayfarer-12B-GGUF 13 days ago

Great RP model in only 12B! A few notes and sampler settings for llama.cpp server inside.

#2 opened 13 days ago by

New activity in hexgrad/Kokoro-82M 25 days ago

Nice ~90x real-time generation on 3090TI. Quickstart provided.

#20 opened 25 days ago by

New activity in bartowski/Llama-3.1-Nemotron-70B-Instruct-HF-GGUF 4 months ago

Observation: 4-bit quantization can't answer the Strawberry prompt

#2 opened 4 months ago by

New activity in bartowski/SuperNova-Medius-GGUF 4 months ago

63.17 MMLU-Pro Computer Science with `Q8_0`

#2 opened 4 months ago by

New activity in bartowski/qwen2.5-7b-ins-v3-GGUF 4 months ago

Benchmarks worse than Qwen2.5-7B-Instruct on MMLU-Pro Computer Science in limited testing.

#1 opened 4 months ago by

New activity in bartowski/Qwen2.5-32B-Instruct-GGUF 4 months ago

Promising looking results on 24GB VRAM folks!

#3 opened 4 months ago by

New activity in deepseek-ai/DeepSeek-V2.5 4 months ago

Awesome model

#5 opened 5 months ago by

New activity in bartowski/DeepSeek-V2.5-GGUF 5 months ago

vram usage of each?

#1 opened 5 months ago by

New activity in bartowski/Mistral-Large-Instruct-2407-GGUF 6 months ago

Works good generating python on my 64GB RAM w/ 3090TI 24GB VRAM dev box

#2 opened 6 months ago by