From 57f85add585f8c55d52bfe1481c49dda042208a7 Mon Sep 17 00:00:00 2001
From: hiyouga <hiyouga@buaa.edu.cn>
Date: Wed, 28 Feb 2024 21:11:23 +0800
Subject: [PATCH] update chatglm3 template

Former-commit-id: 38d8b2cef8d70ce8c390de0317559df7f04b4a5d
---
 README.md                        |  2 +-
 README_zh.md                     |  2 +-
 src/llmtuner/data/template.py    | 15 +++++++++++++++
 src/llmtuner/extras/constants.py |  9 ++++++++-
 4 files changed, 25 insertions(+), 3 deletions(-)

diff --git a/README.md b/README.md
index a34736b3..df64aa75 100644
--- a/README.md
+++ b/README.md
@@ -43,7 +43,7 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/6ba60acc-e2e2-4bec-b846
 - **Various models**: LLaMA, Mistral, Mixtral-MoE, Qwen, Yi, Gemma, Baichuan, ChatGLM, Phi, etc.
 - **Integrated methods**: (Continuous) pre-training, supervised fine-tuning, reward modeling, PPO and DPO.
 - **Scalable resources**: 32-bit full-tuning, 16-bit freeze tuning, 16-bit LoRA tuning, 2/4/8-bit QLoRA with AQLM/AWQ/GPTQ/LLM.int8.
-- **Advanced algorithms**: DoRA, LongLoRA, LLaMA Pro, agent tuning.
+- **Advanced algorithms**: DoRA, LongLoRA, LLaMA Pro, LoftQ, agent tuning.
 - **Intriguing tricks**: FlashAttention-2, Unsloth, RoPE scaling, NEFTune, rsLoRA.
 
 ## Benchmark
diff --git a/README_zh.md b/README_zh.md
index af8216de..c8db1485 100644
--- a/README_zh.md
+++ b/README_zh.md
@@ -43,7 +43,7 @@ https://github.com/hiyouga/LLaMA-Factory/assets/16256802/6ba60acc-e2e2-4bec-b846
 - **多种模型**：LLaMA、Mistral、Mixtral-MoE、Qwen、Yi、Gemma、Baichuan、ChatGLM、Phi 等等。
 - **集成方法**：（增量）预训练、指令监督微调、奖励模型训练、PPO 训练、DPO 训练。
 - **多种精度**：32 比特全参数训练、16 比特部分参数训练、16比特 LoRA 训练、基于 AQLM/AWQ/GPTQ/LLM.int8 的 2/4/8 比特 LoRA 训练。
-- **先进算法**: DoRA、LongLoRA、LLaMA Pro、agent tuning。
+- **先进算法**: DoRA、LongLoRA、LLaMA Pro、LoftQ、agent tuning。
 - **新鲜技巧**：FlashAttention-2、Unsloth、RoPE scaling、NEFTune、rsLoRA。
 
 ## 性能指标
diff --git a/src/llmtuner/data/template.py b/src/llmtuner/data/template.py
index 07ecf72b..428188ba 100644
--- a/src/llmtuner/data/template.py
+++ b/src/llmtuner/data/template.py
@@ -360,6 +360,21 @@ _register_template(
     name="chatglm3",
     format_user=StringFormatter(slots=[{"token": "<|user|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]),
     format_assistant=StringFormatter(slots=["\n", "{{content}}"]),
+    format_system=StringFormatter(slots=[{"token": "[gMASK]"}, {"token": "sop"}]),
+    format_function=FunctionFormatter(slots=["{{name}}\n{{arguments}}"]),
+    format_observation=StringFormatter(
+        slots=[{"token": "<|observation|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]
+    ),
+    stop_words=["<|user|>", "<|observation|>"],
+    efficient_eos=True,
+    force_system=True,
+)
+
+
+_register_template(
+    name="chatglm3_system",
+    format_user=StringFormatter(slots=[{"token": "<|user|>"}, "\n", "{{content}}", {"token": "<|assistant|>"}]),
+    format_assistant=StringFormatter(slots=["\n", "{{content}}"]),
     format_system=StringFormatter(
         slots=[{"token": "[gMASK]"}, {"token": "sop"}, {"token": "<|system|>"}, "\n", "{{content}}"]
     ),
diff --git a/src/llmtuner/extras/constants.py b/src/llmtuner/extras/constants.py
index 62221346..2d2740f9 100644
--- a/src/llmtuner/extras/constants.py
+++ b/src/llmtuner/extras/constants.py
@@ -328,15 +328,19 @@ register_model_group(
     models={
         "Gemma-2B": {
             DownloadSource.DEFAULT: "google/gemma-2b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/gemma-2b",
         },
         "Gemma-7B": {
             DownloadSource.DEFAULT: "google/gemma-7b",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/gemma-2b-it",
         },
         "Gemma-2B-Chat": {
             DownloadSource.DEFAULT: "google/gemma-2b-it",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/gemma-7b",
         },
         "Gemma-7B-Chat": {
             DownloadSource.DEFAULT: "google/gemma-7b-it",
+            DownloadSource.MODELSCOPE: "AI-ModelScope/gemma-7b-it",
         },
     },
     template="gemma",
@@ -562,7 +566,10 @@ register_model_group(
             DownloadSource.DEFAULT: "Qwen/Qwen-1_8B-Chat",
             DownloadSource.MODELSCOPE: "qwen/Qwen-1_8B-Chat",
         },
-        "Qwen-7B-Chat": {DownloadSource.DEFAULT: "Qwen/Qwen-7B-Chat", DownloadSource.MODELSCOPE: "qwen/Qwen-7B-Chat"},
+        "Qwen-7B-Chat": {
+            DownloadSource.DEFAULT: "Qwen/Qwen-7B-Chat",
+            DownloadSource.MODELSCOPE: "qwen/Qwen-7B-Chat",
+        },
         "Qwen-14B-Chat": {
             DownloadSource.DEFAULT: "Qwen/Qwen-14B-Chat",
             DownloadSource.MODELSCOPE: "qwen/Qwen-14B-Chat",