BlazingCustoms/acrostic-4b-v4a

WITHDRAW onnx/ — defective export (wrong hidden_size, duplicated tied lm_head, single 3.10GB external-data file over the 2GiB ArrayBuffer wall); card updated

card: ONNX q4 measured (367/380); q4f16 confirmed smaller AND better

fix: publish intact tokenizer (onnx/merges.txt) — merged/ copy could not be loaded

fix: publish intact tokenizer (merges.txt) — merged/ copy could not be loaded

fix: publish intact tokenizer (vocab.json) — merged/ copy could not be loaded

fix: publish intact tokenizer (tokenizer.json) — merged/ copy could not be loaded

fix: publish intact tokenizer (onnx/tokenizer_config.json) — merged/ copy could not be loaded

fix: publish intact tokenizer (tokenizer_config.json) — merged/ copy could not be loaded

add onnx/merges.txt

add onnx/vocab.json

add onnx/special_tokens_map.json

add onnx/tokenizer_config.json

add onnx/tokenizer.json

add onnx/generation_config.json

add onnx/config.json

add onnx/model.onnx_data

add onnx/model.onnx

add gguf/acrostic-4b-v4a-Q4_K_M.gguf

add gguf/acrostic-4b-v4a-Q5_K_M.gguf

add gguf/acrostic-4b-v4a-Q6_K.gguf

add adapter/adapter_model.safetensors

add adapter/adapter_config.json

add model.safetensors

add chat_template.jinja

add tokenizer.json

add tokenizer_config.json

add generation_config.json

add config.json

add NOTICE

add LICENSE

add README.md

initial commit