Update README.md
Browse files
README.md
CHANGED
|
@@ -14,6 +14,9 @@ tags:
|
|
| 14 |
|
| 15 |
ChemQ3MTP-base is a lightweight generative model for chemistry, trained on 2.33 million valid bioactive and natural product molecules dataset curated from ChemBL34, COCONUTDB, and SuperNatural3. It is built on a compact Qwen2-like transformer backbone and employs a multi-horizon predictive (MTP) loss objective to model molecular structures in SELFIES representation.
|
| 16 |
|
|
|
|
|
|
|
|
|
|
| 17 |
Current version (0.1) (Lic for Code: MIT; Weights: Apache 2.0)
|
| 18 |
|
| 19 |
A custom Qwen2-style language model, adapted for molecular generation:
|
|
|
|
| 14 |
|
| 15 |
ChemQ3MTP-base is a lightweight generative model for chemistry, trained on 2.33 million valid bioactive and natural product molecules dataset curated from ChemBL34, COCONUTDB, and SuperNatural3. It is built on a compact Qwen2-like transformer backbone and employs a multi-horizon predictive (MTP) loss objective to model molecular structures in SELFIES representation.
|
| 16 |
|
| 17 |
+
Train Loss: 1.1720 → Perplexity: ~3.23
|
| 18 |
+
Validation Loss: 1.0448 → Perplexity: ~2.84
|
| 19 |
+
|
| 20 |
Current version (0.1) (Lic for Code: MIT; Weights: Apache 2.0)
|
| 21 |
|
| 22 |
A custom Qwen2-style language model, adapted for molecular generation:
|