Downloads · 30 days
21
40% of all-time downloads
ExpertiaSWE-Q4
OscarFeMa/ExpertiaSWE-Q4
ExpertiaSWE-Q4 is a machine learning model from OscarFeMa. 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 cc-by-nc-4.0.
Spanish-first software engineering specialist: formal definitions + parameters + minimal examples. Fine-tuned from microsoft/Phi-4-mini-reasoning (MIT) with QLoRA r16 on ~45k pairs (StackOverflow + softwareengineering…
All-time downloads
52
Public
Repo size
17.8 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.gguf10.2 GB · 57%
At a glance
- License
- cc-by-nc-4.0
- Access
- Public
- Created
- Sep 19, 2026
- Updated
- Sep 19, 2026
- SHA
- 0742022b
Base models
Model card
From the Hugging Face model README
ExpertiaSWE (Q4_K_M)
Spanish-first software engineering specialist: formal definitions + parameters + minimal examples.
Fine-tuned from microsoft/Phi-4-mini-reasoning (MIT) with QLoRA r16 on ~45k pairs
(StackOverflow + softwareengineering.se + codereview.se + Wikipedia + Wikidata definitional bulk),
3 epochs on RTX 3070 (seq1024, BF16). structured_knowledge was cleaned of the
Entity:/Properties:/P-code scaffolding before training (anti-regurgitation).
Evaluation (500 held-out samples)
| Perplexity | |
|---|---|
| Base | 14892.9 |
| + adapter | 63.7 (-99.6%) |
Canary (10 fixed prompts via Ollama): 10/10 correct definitions, 0 template
regurgitations. Note: answers in English to Spanish prompts (shared with sibling
specialists) — production use is distillation with SYSTEM + think:false, not chat.
Variants in this repo
| File | Size | Needs | Use |
|---|---|---|---|
expertia-swe-q4_k_m.gguf | 2.4GB | 6GB VRAM | Daily inference (Ollama) |
expertia-swe-f16.gguf | 7.3GB | 16GB VRAM | Max quality inference |
fp16/ | ~7GB | — | Base for further fine-tuning |
Usage (Ollama)
ollama create expertia-swe -f Modelfile-ExpertiaSWE-Q4F
Note: GGUF uses gpt-2 pre-tokenizer; set num_ctx 8192.
Limitations
Narrow domain specialist (software engineering). Adapter weights: CC-BY-NC-4.0. Base: MIT.
- License
- cc-by-nc-4.0
- Languages
- es, en
- Created
- Sep 19, 2026
- Updated
- Sep 19, 2026