Fix kenlm loading in readme (#62)

* Fix kenlm loading in readme.

* Update __init__.py

Co-authored-by: Jeremy Lopez <lopez86@users.noreply.github.com>

README.md

@@ -25,12 +25,8 @@ pip install pyctcdecode
 ### Quick Start:
 
 ``` python
-import kenlm
 from pyctcdecode import build_ctcdecoder
 
-# load trained kenlm model
-kenlm_model = kenlm.Model("/my/dir/kenlm_model.arpa")
-
 # specify alphabet labels as they appear in logits
 labels = [
     " ", "a", "b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l",
@@ -40,7 +36,7 @@ labels = [
 # prepare decoder and decode logits via shallow fusion
 decoder = build_ctcdecoder(
     labels,
-    kenlm_model,
+    kenlm_model_path="/my/dir/kenlm_model.arpa",  # either .arpa or .bin file
     alpha=0.5,  # tuned on a val set
     beta=1.0,  # tuned on a val set
 )
@@ -56,7 +52,7 @@ labels = ["<unk>", "▁bug", "s", "▁bunny"]
 
 decoder = build_ctcdecoder(
     labels,
-    kenlm_model,
+    kenlm_model_path,
 )
 text = decoder.decode(logits)
 ```

pyctcdecode/__init__.py

@@ -5,4 +5,4 @@
 
 
 __package_name__ = "pyctcdecode"
-__version__ = "0.3.3"
+__version__ = "0.3.4"