]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 3e945cc)
convert.py : fix llama/llama2 conversion due to vocab_size=-1 (#5019)
authorDavid Sommers <redacted>
Thu, 18 Jan 2024 17:20:59 +0000 (12:20 -0500)
committerGitHub <redacted>
Thu, 18 Jan 2024 17:20:59 +0000 (19:20 +0200)
commitb46757735d30f5c6ed4f20ebeccc684e02d4f3bf
tree7327dd01fecf2f45007ae84e9d09345f317083bbtree
parent3e945cc1e9c06d2001031360e4e303e9548fb02ccommit | diff
convert.py : fix llama/llama2 conversion due to vocab_size=-1 (#5019)

PR #4818 (merged last week) reintroduced a config check for vocab_size that was addressed in PR #4258 (merged 2023-11-30).

Without the fix, llama2 models can't be converted. The error is:

`ValueError: The model's vocab size is set to -1 in params.json. Please update it manually. Maybe 32000?`
convert.py diff | blob | history
Packaging of ggml-org/llama.cpp