diff --git a/CHANGELOG.md b/CHANGELOG.md
index 68020d0..956e69e 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,5 +1,12 @@
 # Changelog
 *********************************
+## CellTypist 1.2.0 (August 22, 2022)
+- Report all model url request errors (including timeout error) [#28](https://github.com/Teichlab/celltypist/issues/28)
+- Add `with_mean` parameter in [celltypist.train](https://celltypist.readthedocs.io/en/latest/celltypist.train.html) to optimize RAM usage by not subtracting the mean [#29](https://github.com/Teichlab/celltypist/issues/29)
+- Raise warning instead of error during prediction when the input expression matrix contains only a subset of genes
+- Reformat [docs](https://celltypist.readthedocs.io/en/latest/?badge=latest#) with independent pages for main functions
+- Support dot plot when the prediction comes from the 'prob match' mode [#33](https://github.com/Teichlab/celltypist/issues/33)
+- Disable hard-coded `max_iter` in [celltypist.train](https://celltypist.readthedocs.io/en/latest/celltypist.train.html) with default values varying according to dataset sizes
 ## CellTypist 1.1.0 (June 09, 2022)
 - Add citation (PMID: 35549406)
 - Add Docker/Singularity container
diff --git a/README.md b/README.md
index 38d9e77..265547b 100644
--- a/README.md
+++ b/README.md
@@ -480,7 +480,7 @@ Currently, there is no plan for R compatibility. Try to convert R objects into A
   The downstream workflow is the same as that from one-pass data training.
 
   ### General parameters relating to runtime and RAM usage
-  `max_iter`: when `celltypist.train` does not converge for a long time, setting `max_iter` to a lower number (default to 1000) can reduce runtime at a possible cost of a suboptimal model.  
+  `max_iter`: when `celltypist.train` does not converge for a long time, setting `max_iter` to a lower number can reduce runtime at a possible cost of a suboptimal model.  
     
   `with_mean`: when the training data is a sparse matrix, setting `with_mean = False` will preserve sparsity by skipping the step of subtraction by the mean during scaling, and thus lower the RAM usage at the cost of a suboptimal model.  
     
diff --git a/celltypist/__init__.py b/celltypist/__init__.py
index dff1cf6..c0e9339 100644
--- a/celltypist/__init__.py
+++ b/celltypist/__init__.py
@@ -3,4 +3,4 @@
 from .train import train
 from .plot import dotplot
 
-__version__ = "1.1.0"
+__version__ = "1.2.0"