🐞 Bug: Unable to spit models to remote host

[tristan-k]

Description

Models downloaded through huggingface.co can't be spit to a remote host.

How to reproduce

$ ollama pull hf.co/lmstudio-community/Llama-3.3-70B-Instruct-GGUF:Q4_K_M
$ gollama --spit hf.co/lmstudio-community/Llama-3.3-70B-Instruct-GGUF:Q4_K_M --remote http://192.168.10.127:11434 -ollama-dir /usr/share/ollama/.ollama/models
Error copying model(s) to remote host: model not found in /usr/share/ollama/.ollama/models/manifests/registry.ollama.ai/hf.co/lmstudio-community/Llama-3.3-70B-Instruct-GGUF/Q4_K_M

Environment

• Linux distribution: Debian GNU/Linux 12 (bookworm)

• Install source: Go install
• Go version: go version go1.24.3 linux/amd64

[sammcj]

Hey, not sure if you're still trying to do this or not but could you please let me know what the actual file path on disk for that model is? (I'm assuming it's not /usr/share/ollama/.ollama/models/manifests/registry.ollama.ai/hf.co/lmstudio-community/Llama-3.3-70B-Instruct-GGUF/Q4_K_M)

[tristan-k]

I just tried this again on another system and it still fails:

$ ollama ls | grep codellama
codellama:34b          685be00e1532    19 GB     4 weeks ago

$ gollama -v
1.35.1
$ gollama -H -l | grep codellama
codellama:34b                                   17.74GB     34B         Q4_0        llama           2025-07-16  685be00

$ gollama -H --spit codellama:34b --remote http://192.168.178.111:11434 -ollama-dir /Users/admin/.ollama/models
Copying model codellama:34b to remote host http://192.168.178.111:11434...
Error copying model(s) to remote host: model not found in /usr/share/ollama/models/manifests/registry.ollama.ai/library/codellama/34b

In this case the path to the models is /root/.ollama/models/ and not /usr/share/ollama/models/manifests/registry.ollama.ai/library/.

Specifying the model path on the host as well as on the remote host, also throws an error:

$ gollama -H -ollama-dir /root/.ollama/models --spit codellama:34b --remote http://192.168.178.111:11434 -ollama-dir /Users/admin/.ollama/models
Copying model codellama:34b to remote host http://192.168.178.111:11434...
Error copying model(s) to remote host: model not found in /Users/admin/.ollama/models/manifests/registry.ollama.ai/library/codellama/34b

On the remote host:

$ ls /Users/admin/.ollama/models/manifests/registry.ollama.ai/library
all-minilm          deepseek-r1         gemma2              gemma3n             llama3.2-vision     magistral           phi4-mini-reasoning qwen2.5-coder       qwen3-coder
bge-m3              devstral            gemma3              gpt-oss             llava               phi4                phi4-reasoning      qwen3

Why does gollama expects the model to be on the remote host if the purpose is to transfer the model to the remote host in the first place?

Same thing with spitter:

$ spitter codellama:34b http://192.168.178.111:11434 --model-dir /Users/admin/.ollama/models
Error: model not found in /Users/admin/.ollama/models/manifests/registry.ollama.ai/library/codellama/34b
Usage:
  spitter [local_model] [remote_server] [flags]

Flags:
  -a, --all                 Push all models to the remote host
  -h, --help                help for spitter
  -d, --model-dir string    Custom Ollama model directory path
  -c, --ollama-cmd string   Custom Ollama command (e.g., "docker exec -it ollama ollama")

model not found in /Users/admin/.ollama/models/manifests/registry.ollama.ai/library/codellama/34b

[sammcj]

Hello, I'm sorry to say I'm decommissioning the model spitting functionality in Gollama.

For more information, please see #225 (comment)