Views
No views yet
| Parameter | Value |
|---|---|
| Base Model | Qwen2.5-Coder-14B |
| Method | LoRA (rank 32, alpha 64) |
| Training GPU | NVIDIA A100-SXM4-40GB |
| Dataset | 881 expert examples |
| Epochs | 3 |
| Final Loss | 0.17 |
| Sequence Length | 1024 |
| Target Modules | q, k, v, o, gate, up, down |
| File | Size | Quantization | Use Case |
|---|---|---|---|
jarvis-14b-cnc-expert.Q4_K_M.gguf | 8.4 GB | Q4_K_M (4.87 BPW) | Recommended — best quality/size ratio |
1# Download and create model
2ollama create jarvis-cnc -f Modelfile
3
4# Run
5ollama run jarvis-cnc "Write Fanuc G-code for drilling 4 M6 holes in a 40x40mm square pattern, depth 15mm"FROM jarvis-14b-cnc-expert.Q4_K_M.gguf
SYSTEM "You are JARVIS, a CNC and IoT expert. You generate production-ready G-code for Fanuc, Siemens and Heidenhain controllers."
PARAMETER temperature 0.7
PARAMETER num_ctx 4096This model is gated. All access requires approval.