Fix typo (#1671)

docs/docs/roadmap.md

@@ -75,7 +75,7 @@ Objectives in this space include refining the abstractions and implementations f
 
 In February 2023, DSPy introduced the notion of compiling to optimize the weights of an LM program. (To understand just how long ago that was in AI terms, this was before the Alpaca training project at Stanford had even started and a month before the first GPT-4 was released.) Since then, we have shown in October 2023 and, much more expansively, in July 2024, that the fine-tuning flavor of DSPy can deliver large gains for small LMs, especially when composed with prompt optimization.
 
-Overall, though, most DSPy users in practice explore prompt optimization and not weight optimization and most of our examples do the same. The primary reason for a lot of this is infrastructure. Fine-tuning in the DSPy flavor is more than just training a model: ultimately, we need to bootstrap training data for serveral different modules in a program, train multiple models and handle model selection, and then load and plug in those models into the program's modules. Doing this robustly at the level of abstraction DSPy offers requires a level of resource management that is not generally supported by external existing tools. Major efforts in this regard are currently led by Dilara Soylu and Isaac Miller.
+Overall, though, most DSPy users in practice explore prompt optimization and not weight optimization and most of our examples do the same. The primary reason for a lot of this is infrastructure. Fine-tuning in the DSPy flavor is more than just training a model: ultimately, we need to bootstrap training data for several different modules in a program, train multiple models and handle model selection, and then load and plug in those models into the program's modules. Doing this robustly at the level of abstraction DSPy offers requires a level of resource management that is not generally supported by external existing tools. Major efforts in this regard are currently led by Dilara Soylu and Isaac Miller.
 
 
 ### On Optimizers & Assertions