Downloads · 30 days
4
2% of all-time downloads
phila-ft-v3-GGUF
onblueroses/phila-ft-v3-GGUF
phila-ft-v3-GGUF is a machine learning model from onblueroses. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
Gate-only fine-tune of Llama 3.2 3B for phila, a silence-first group chat agent.
All-time downloads
251
Public
Repo size
2 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.gguf2 GB · 100%
At a glance
- License
- mit
- Access
- Public
- Created
- Apr 6, 2026
- Updated
- Apr 6, 2026
- SHA
- 0cf259a1
Base models
Model card
From the Hugging Face model README
phila-ft-v3 (Q4_K_M)
Gate-only fine-tune of Llama 3.2 3B for phila, a silence-first group chat agent.
Training
- Method: QLoRA r=16 a=32, 4-bit quantization
- Examples: 3,799 gate-only (59.3% speak, 40.7% silent)
- GPU: Vast.ai RTX 4090
- Base model: meta-llama/Llama-3.2-3B-Instruct
Results
| Config | Original Suite | Independent Suite | Gap |
|---|---|---|---|
| Base llama3.2 | 86.3% | 67.0% | -19.3pp |
| phila-ft-v2 | 90.3% | 76.7% | -13.6pp |
| phila-ft-v3 | 93.6% | 93.3% | -0.3pp |
Independent suite: 174 scenarios generated by Claude Opus from category definitions only. Zero contamination (SHA-256 verified).
Key improvement: recall jumped from 0.612 to 0.890 on independent suite (+27.8pp) while precision improved from 0.984 to 0.994.
Usage
# Create Ollama model
ollama create phila-ft-v3 -f Modelfile
Modelfile:
FROM phila-ft-v3.Q4_K_M.gguf
TEMPLATE "{{- if .System }}<|start_header_id|>system<|end_header_id|>
{{ .System }}<|eot_id|>{{ end }}{{- range .Messages }}<|start_header_id|>{{ .Role }}<|end_header_id|>
{{ .Content }}<|eot_id|>{{ end }}<|start_header_id|>assistant<|end_header_id|>
"
PARAMETER top_p 0.52
PARAMETER num_predict 64
PARAMETER stop <|start_header_id|>
PARAMETER stop <|end_header_id|>
PARAMETER stop <|eot_id|>
PARAMETER temperature 0.1
Project
- License
- mit
- Created
- Apr 6, 2026
- Updated
- Apr 6, 2026