Release DXA-LM specialized deployment intelligence weights, Ollama Modelfile, and Phase 28-30 scientific validation artifacts
2523b1a verified Download quantization_comparison.json from deployxa/dxa-lm: direct link, hf CLI and curl.
- Browser
- Download file 1.23 kB
-
https://huggingface.co/deployxa/dxa-lm/resolve/main/quantization_comparison.json
- Command line
-
hf download hf://deployxa/dxa-lm/quantization_comparison.json
-
curl -L -o quantization_comparison.json https://huggingface.co/deployxa/dxa-lm/resolve/main/quantization_comparison.json
1.23 kB
| [ | |
| { | |
| "quant_type": "FP16", | |
| "file_suffix": "fp16.gguf", | |
| "size_mb": 3100, | |
| "ram_required_mb": 3800, | |
| "core_accuracy_pct": 94.77, | |
| "real_accuracy_pct": 60.42, | |
| "recovery_rate_pct": 53.85, | |
| "tokens_per_sec": 14.2, | |
| "p50_latency_ms": 320.0, | |
| "p95_latency_ms": 480.0 | |
| }, | |
| { | |
| "quant_type": "Q8_0", | |
| "file_suffix": "q8_0.gguf", | |
| "size_mb": 1650, | |
| "ram_required_mb": 2100, | |
| "core_accuracy_pct": 94.72, | |
| "real_accuracy_pct": 60.35, | |
| "recovery_rate_pct": 53.8, | |
| "tokens_per_sec": 22.5, | |
| "p50_latency_ms": 210.0, | |
| "p95_latency_ms": 310.0 | |
| }, | |
| { | |
| "quant_type": "Q6_K", | |
| "file_suffix": "q6_k.gguf", | |
| "size_mb": 1280, | |
| "ram_required_mb": 1650, | |
| "core_accuracy_pct": 94.6, | |
| "real_accuracy_pct": 60.1, | |
| "recovery_rate_pct": 53.5, | |
| "tokens_per_sec": 28.1, | |
| "p50_latency_ms": 175.0, | |
| "p95_latency_ms": 250.0 | |
| }, | |
| { | |
| "quant_type": "Q4_K_M", | |
| "file_suffix": "q4_k_m.gguf", | |
| "size_mb": 940, | |
| "ram_required_mb": 1200, | |
| "core_accuracy_pct": 94.45, | |
| "real_accuracy_pct": 59.85, | |
| "recovery_rate_pct": 53.1, | |
| "tokens_per_sec": 34.8, | |
| "p50_latency_ms": 140.0, | |
| "p95_latency_ms": 195.0, | |
| "recommended_production": true | |
| } | |
| ] |