Defetya
/

qwen-1.8B-jax

Model card Files Files and versions

3.67 GB

1 contributor

History: 6 commits

Defetya's picture

fix the damn proj bias

4a06227 verified about 2 years ago

.gitattributes

1.52 kB

initial commit about 2 years ago
README.md

28 Bytes

initial commit about 2 years ago
qwen-1.8B-JAX.pickle
Detected Pickle imports (10)
- "lib.llama.decoder_block.DecoderBlock",
- "jax._src.array._reconstruct_array",
- "numpy.core.multiarray._reconstruct",
- "ml_dtypes.bfloat16",
- "lib.llama.llama.Llama",
- "lib.llama.llama_model.LlamaModel",
- "lib.llama.attention.Attention",
- "numpy.ndarray",
- "lib.llama.attention.AttentionProj",
- "numpy.dtype"
How to fix it?
3.67 GB
xet

fix the damn proj bias about 2 years ago