AtomicChat/Ling-3.0-flash-GGUF
:ledger: fix(readme): Bump up release version with fixes
:ledger: docs(readme): Update building release
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
Upload folder using huggingface_hub
:ledger: docs(readme): Update graphs
:ledger: docs(readme): Add graphs, update documentation & measurements
docs: use the release archives, no build needed
chore: name the importance matrix so it is not mistaken for a second bf16 model
add NVFP4_V5
add NVFP4_V5
chore: drop the broken v4 build
add NVFP4_V4
add NVFP4_V4
chore: drop the broken v3 build
add NVFP4_V3
add NVFP4_V3
chore: NVFP4_STOCK is the round-to-nearest control, ours stays AD
docs: measured grid, speed, naming
docs: measured grid, speed, naming
docs: measured grid, speed, naming
add NVFP4_AI2
add NVFP4_AI2
chore: the importance matrix is not a quant, keep its folder plain
chore: brand the folder too, the hub labels a quant by its folder
chore: one folder per rung, AD prefix on the dynamic layouts
add NVFP4_AI
