update

.gitattributes

@@ -33,5 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
-tokenizer.json filter=lfs diff=lfs merge=lfs -textfigures/z1_32b_benchmark.png filter=lfs diff=lfs merge=lfs -text
-figures/z1_9b_benchmark.png filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -22,14 +22,15 @@ Finally, **GLM-Z1-9B-0414** is a surprise. We employed the aforementioned series
 ## Performance
 
 <p align="center">
-  <img width="100%" src="figures/z1_32b_benchmark.png">
+  <img width="100%" src="https://raw.githubusercontent.com/THUDM/GLM-4/refs/heads/main/resources/Bench-Z1-32B.png">
 </p>
 
 <p align="center">
-  <img width="100%" src="figures/z1_9b_benchmark.png">
+  <img width="100%" src="https://raw.githubusercontent.com/THUDM/GLM-4/refs/heads/main/resources/Bench-Z1-9B.png">
 </p>
 
 ## Model Usage Guidelines
+
 ### I. Sampling Parameters
 
 | Parameter    | Recommended Value | Description                                  |

chat_template.jinja

@@ -41,4 +41,4 @@
 {{ visible }}
     {%- endif %}
 {%- endfor %}
-{% if add_generation_prompt %}<|assistant|>{% endif %}
+{% if add_generation_prompt %}<|assistant|>\n<think>{% endif %}