ollama - 💡(How to fix) Fix Error: 400 (empty body) when pulling hf.co GGUF models after successful download

pulling manifest pulling fe4c25ea2f4e: 100% ▕████████████████████████████████████▏ 8.0 GB Error: 400:

TL;DR

Manually creating the model from the downloaded blob using the provided workaround script may resolve the issue.

Guidance

• The error occurs after a successful download of the GGUF model, suggesting an issue with the model registration process.
• The workaround provided in the issue body can be used to manually create the model from the downloaded blob, which may help to identify if the issue is specific to the ollama run command.
• Verify that the downloaded blob is a valid GGUF v3 file, as confirmed in the issue body, to rule out any issues with the model file itself.
• Check the server log for any additional error details, although the current log output only shows a 200 POST /api/pull response.

Example

The provided workaround script can be used as an example:

cat > Modelfile << 'MODELFILE'
FROM /path/to/.ollama/models/blobs/sha256-<hash>

TEMPLATE """<bos>{{ if .System }}<start_of_turn>user
{{ .System }}<end_of_turn>
{{ end }}{{ range .Messages }}{{ if eq .Role "user" }}<start_of_turn>user
{{ .Content }}<end_of_turn>
<start_of_turn>model
{{ else if eq .Role "assistant" }}{{ .Content }}<end_of_turn>
{{ end }}{{ end }}<start_of_turn>model
"""

PARAMETER stop "<end_of_turn>"
PARAMETER stop "<eos>"
MODELFILE

ollama create my-model -f Modelfile

Replace <hash> with the actual hash of the downloaded blob.

Notes

The issue may be specific to the ollama run command or the model registration process, and further investigation is needed to determine the root cause.

Recommendation

Apply the workaround by manually creating the model from the downloaded blob using the provided script, as it has been confirmed to work in the issue body. This may help to resolve the issue until a more permanent fix is available.