Use Series.str.extract with casting to floats:
df = pd.DataFrame({'A':'4.5 mg, 5.8 mg, 6.3 mg'.split(', ')})
df['new'] = df['A'].str.extract(r'(\d\.\d)+').astype(float)
If possible some integers values:
df['new'] = df['A'].str.extract(r"(\d*\.?\d+|\d+)").astype(float)
print (df)
A new
0 4.5 mg 4.5
1 5.8 mg 5.8
2 6.3 mg 6.3
If possible use split by first whitespace use Series.str.split with str for indexing first values:
df['val'] = df['A'].str.split().str[0].astype(float)