47 lines
1.0 KiB
Markdown
47 lines
1.0 KiB
Markdown
---
|
|
language:
|
|
- en
|
|
license: apache-2.0
|
|
base_model: Qwen/Qwen2.5-Coder-7B-Instruct
|
|
tags:
|
|
- synoema
|
|
- code-generation
|
|
- functional-programming
|
|
- qlora
|
|
- fine-tuned
|
|
datasets:
|
|
- custom
|
|
pipeline_tag: text-generation
|
|
---
|
|
|
|
# Synoema Coder 7B v6 (0.1.0-alpha.3)
|
|
|
|
Fine-tuned Qwen/Qwen2.5-Coder-7B-Instruct for Synoema code generation.
|
|
|
|
## Model Details
|
|
|
|
| Property | Value |
|
|
|----------|-------|
|
|
| Base model | Qwen/Qwen2.5-Coder-7B-Instruct |
|
|
| Fine-tune method | QLoRA (4-bit, rank 8) |
|
|
| Training corpus | v6 (5,907 examples, Alpaca format) |
|
|
| Semantic examples | 870 validated |
|
|
| Synoema version | 0.1.0-alpha.3 |
|
|
| Epochs | 3 |
|
|
| Hardware | AMD RX 7900 GRE 16GB (ROCm) |
|
|
|
|
## Eval Results (semantic suite, 40 tests)
|
|
|
|
| Metric | Baseline | This model |
|
|
|--------|----------|------------|
|
|
| syntax_pass | 100% | 100% |
|
|
| run_pass | 100% | 100% |
|
|
| constructs_pass | 82.5% | 87.5% |
|
|
|
|
## Naming Convention
|
|
|
|
synoema-coder-SIZE-vCORPUS-COMPILER
|
|
|
|
- v6 = corpus version 6 (5037 original + 870 semantic)
|
|
- 0.1.0a3 = Synoema compiler 0.1.0-alpha.3
|