88plug/Kimi-VL-A3B-Thinking-2506-W4A16

fix(config): dual-write torch_dtype from dtype for loader compat

P15 aux: add preprocessor_config.json from base moonshotai/Kimi-VL-A3B-Thinking-2506

Sync model card README (tier, quant, FLAC, benchmarks)

Sync model card README (tier row + footer)

Add model.safetensors

Add tokenizer_config.json

Add tokenization_moonshot.py

Add tiktoken.model

Add recipe.yaml

Add quality_stamp.json

Add processor_config.json

Add processing_kimi_vl.py

Add modeling_kimi_vl.py

Add image_processing_kimi_vl.py

Add generation_config.json

Add configuration_kimi_vl.py

Add config.json

Add chat_template.jinja

Add README.md

initial commit