125d0a013a
ggufv3 adds support for big endianness, mainly for s390x architecture. while that's not currently supported for ollama, the change is simple. loosen version check to be more forward compatible. unless specified, gguf versions other v1 will be decoded into v2. |
||
|---|---|---|
| .. | ||
| llama.cpp | ||
| falcon.go | ||
| ggml.go | ||
| gguf.go | ||
| llama.go | ||
| llm.go | ||
| starcoder.go | ||
| utils.go | ||