mirror of
https://github.com/dogkeeper886/ollama37.git
synced 2025-12-12 00:37:04 +00:00
deprecate ggml
- remove ggml runner - automatically pull gguf models when ggml detected - tell users to update to gguf in the case automatic pull fails Co-Authored-By: Jeffrey Morgan <jmorganca@gmail.com>
This commit is contained in:
committed by
Daniel Hiltgen
parent
ed195f3562
commit
811b1f03c8
16
llm/llm.go
16
llm/llm.go
@@ -76,16 +76,8 @@ func New(workDir, model string, adapters, projectors []string, opts api.Options)
|
||||
}
|
||||
}
|
||||
|
||||
switch ggml.Name() {
|
||||
case "gguf":
|
||||
// TODO: gguf will load these options automatically from the model binary
|
||||
opts.NumGQA = 0
|
||||
opts.RopeFrequencyBase = 0.0
|
||||
opts.RopeFrequencyScale = 0.0
|
||||
return newLlama(model, adapters, projectors, chooseRunners(workDir, "gguf"), ggml.NumLayers(), opts)
|
||||
case "ggml", "ggmf", "ggjt", "ggla":
|
||||
return newLlama(model, adapters, projectors, chooseRunners(workDir, "ggml"), ggml.NumLayers(), opts)
|
||||
default:
|
||||
return nil, fmt.Errorf("unknown ggml type: %s", ggml.ModelFamily())
|
||||
}
|
||||
opts.NumGQA = 0
|
||||
opts.RopeFrequencyBase = 0.0
|
||||
opts.RopeFrequencyScale = 0.0
|
||||
return newLlama(model, adapters, projectors, chooseRunners(workDir), ggml.NumLayers(), opts)
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user