A LLM trained only on data from certain time periods to reduce modern bias
Python
Update README.md
Hayk Grigorianfa1c00822e95
diff --git a/README.md b/README.mdindex f215e19..b9689dc 100644--- a/README.md+++ b/README.md@@ -81,7 +81,7 @@ I will train this dataset today and it should take around 4-5 hours. Once it's d## August 12, 2025-Version 1 is complete, this version uses Phi 1.5 (700M parameters) and is trained on a dataset that has almost 7000 texts. The main improvements with this model is it's ability to reference real names and historical events or at least kind of. Version 0 and 0.5 could not do this, whenever I prompted those models with something like "In the year 1821" the output would just be hallucinations, stuff that never happened. An example of this with Version 1:+Version 1 is complete, this version uses Phi 1.5 (700M parameters) and is trained on a dataset that has almost 7000 texts. The main improvements with this model is it's ability to sometimes reference real names and historical events. Version 0 and 0.5 could not do this, whenever I prompted those models with something like "In the year 1821" the output would just be hallucinations, stuff that never happened. An example of this with Version 1:My prompt: It was the year of our Lord 1834,@@ -190,3 +190,4 @@ Ram: 16GB DDR5.+