Support InternLM

[vansinhu]

Dear ollama developer,

Greetings! I am vansinhu, a community developer and volunteer at InternLM. InternLM is a large language model similar to llama2, and we look forward to InternLM being supported in ollama. If there are any challenges or inquiries regarding support for InternLM, please feel free to join our Discord discussion at https://discord.gg/gF9ezcmtM3.

Best regards,
vansinhu

[mchiang0610]

@vansinhu I would love to support this. Will take a look through your huggingface repos. Sorry for the delay!!

[mchiang0610]

Possible to ask you to quantize the models, and then upload it to ollama?

[jmorganca]

Hi @vansinhu. We'd love to support this. Ollama currently runs llama.cpp under the hood and we'd need some help getting that in ggerganov/llama.cpp#3133. After that, this should be no problem to add

[CausalLM]

See ggerganov/llama.cpp#4283

[BruceMacD]

Thanks for the request, the model is now supported, and it looks like some community members have added it to Ollama:
https://ollama.com/search?q=intern&p=1

Resolving this for now, depending on the popularity we might add it to the official library namespace.

[QiuZiXian]

@BruceMacD May you give me a template example？
I follow fun-tunning，get the merge model，
then follow ollama to convert to gguf
create from mainfest is ok，
but conversion is not.
replay can't stop

my template refer :
{{ if .System }}<|im_start|>system
{{ .System }}<|im_end|>
{{ end }}{{ if .Prompt }}<|im_start|>user
{{ .Prompt }}<|im_end|>
{{ end }}<|im_start|>assistant
{{ .Response }}<|im_end|>