4 points | by mirzap 9 hours ago ago
1 comments
You can do same with Harbor in a couple of commands :)
# Pull Qwen 3.5 35B
harbor pull unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_XL
# Set llama.cpp image to an optimised one for performance on Strix Halo
harbor config set llamacpp.image_rocm docker.io/kyuz0/amd-strix-halo-toolboxes:rocm-7.2
# Set default model for the claw
harbor config set openclaw.model unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_XL
# Start both
harbor up openclaw llamacpp
You can do same with Harbor in a couple of commands :)
# Pull Qwen 3.5 35B
harbor pull unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_XL
# Set llama.cpp image to an optimised one for performance on Strix Halo
harbor config set llamacpp.image_rocm docker.io/kyuz0/amd-strix-halo-toolboxes:rocm-7.2
# Set default model for the claw
harbor config set openclaw.model unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_XL
# Start both
harbor up openclaw llamacpp