Bunch of good models
Sawyer Bowerman
soyrsoyr
AI & ML interests
cv, optimization
Recent Activity
updated a collection 11 days ago
LLM Compressor testing updated a collection 11 days ago
LLM Compressor testing updated a model 13 days ago
inference-optimization/Inkling-0.92B-MTPOrganizations
Bunch of bad models
GPTQ quantized across W4A16, W8A8, FP8, NVFP4 using llm-compressor.
-
soyrsoyr/Llama-3.2-1B-Instruct-W4A16-GPTQ
Text Generation • 1B • Updated • 20 -
soyrsoyr/Llama-3.2-1B-Instruct-W8A8-GPTQ
Text Generation • 1B • Updated • 26 -
soyrsoyr/Llama-3.2-1B-Instruct-FP8-GPTQ
Text Generation • 1B • Updated • 16 -
soyrsoyr/Llama-3.2-1B-Instruct-NVFP4-GPTQ
Text Generation • 0.8B • Updated • 40
models I'm using for testing
-
soyrsoyr/gemma-4-unified-0.8B-tiny
Image-Text-to-Text • 0.8B • Updated • 48 -
soyrsoyr/Nemotron-3.5-Lightning-1.4B-A0.1B-MTP-FP8-Dynamic
1B • Updated • 14 -
soyrsoyr/Qwen3-VL-Reranker-0.1B-tiny
Image-Text-to-Text • 98.7M • Updated • 34 -
soyrsoyr/Nemotron-3.5-Lightning-1.4B-A0.1B-MTP-NVFP4
Text Generation • 1B • Updated • 91 • 1
Quantized across W4A16, FP8, NVFP4 using llm-compressor.
DeepSeek-MoE-16B-Chat quantized with GPTQ via llm-compressor: W8A8, W4A16, FP8, NVFP4.
-
soyrsoyr/deepseek-moe-16b-chat-W8A8-GPTQ
Text Generation • 16B • Updated • 29 -
soyrsoyr/deepseek-moe-16b-chat-W4A16-GPTQ
Text Generation • 3B • Updated • 82 -
soyrsoyr/deepseek-moe-16b-chat-FP8-GPTQ
Text Generation • 16B • Updated • 22 -
soyrsoyr/deepseek-moe-16b-chat-NVFP4-GPTQ
Text Generation • 9B • Updated • 30
tiny models for development
Bunch of good models
models I'm using for testing
-
soyrsoyr/gemma-4-unified-0.8B-tiny
Image-Text-to-Text • 0.8B • Updated • 48 -
soyrsoyr/Nemotron-3.5-Lightning-1.4B-A0.1B-MTP-FP8-Dynamic
1B • Updated • 14 -
soyrsoyr/Qwen3-VL-Reranker-0.1B-tiny
Image-Text-to-Text • 98.7M • Updated • 34 -
soyrsoyr/Nemotron-3.5-Lightning-1.4B-A0.1B-MTP-NVFP4
Text Generation • 1B • Updated • 91 • 1
Bunch of bad models
Quantized across W4A16, FP8, NVFP4 using llm-compressor.
GPTQ quantized across W4A16, W8A8, FP8, NVFP4 using llm-compressor.
-
soyrsoyr/Llama-3.2-1B-Instruct-W4A16-GPTQ
Text Generation • 1B • Updated • 20 -
soyrsoyr/Llama-3.2-1B-Instruct-W8A8-GPTQ
Text Generation • 1B • Updated • 26 -
soyrsoyr/Llama-3.2-1B-Instruct-FP8-GPTQ
Text Generation • 1B • Updated • 16 -
soyrsoyr/Llama-3.2-1B-Instruct-NVFP4-GPTQ
Text Generation • 0.8B • Updated • 40
DeepSeek-MoE-16B-Chat quantized with GPTQ via llm-compressor: W8A8, W4A16, FP8, NVFP4.
-
soyrsoyr/deepseek-moe-16b-chat-W8A8-GPTQ
Text Generation • 16B • Updated • 29 -
soyrsoyr/deepseek-moe-16b-chat-W4A16-GPTQ
Text Generation • 3B • Updated • 82 -
soyrsoyr/deepseek-moe-16b-chat-FP8-GPTQ
Text Generation • 16B • Updated • 22 -
soyrsoyr/deepseek-moe-16b-chat-NVFP4-GPTQ
Text Generation • 9B • Updated • 30
tiny models for development