[model] add models (#7054)

* add qwen25vl awq models

* add moonlight

Former-commit-id: ec1a1bc1184d13188029e19c1d4e7de68707aaf6

examples/train_full/llama3_full_sft.yaml

@@ -34,6 +34,7 @@ bf16: true
 ddp_timeout: 180000000
 
 ### eval
+# eval_dataset: alpaca_en_demo
 # val_size: 0.1
 # per_device_eval_batch_size: 1
 # eval_strategy: steps

examples/train_lora/llama3_lora_dpo.yaml

@@ -37,6 +37,7 @@ bf16: true
 ddp_timeout: 180000000
 
 ### eval
+# eval_dataset: dpo_en_demo
 # val_size: 0.1
 # per_device_eval_batch_size: 1
 # eval_strategy: steps

examples/train_lora/llama3_lora_pretrain.yaml

@@ -34,6 +34,7 @@ bf16: true
 ddp_timeout: 180000000
 
 ### eval
+# eval_dataset: c4_demo
 # val_size: 0.1
 # per_device_eval_batch_size: 1
 # eval_strategy: steps

examples/train_lora/llama3_lora_reward.yaml

@@ -35,6 +35,7 @@ bf16: true
 ddp_timeout: 180000000
 
 ### eval
+# eval_dataset: dpo_en_demo
 # val_size: 0.1
 # per_device_eval_batch_size: 1
 # eval_strategy: steps

examples/train_lora/llama3_lora_sft.yaml

@@ -35,6 +35,7 @@ bf16: true
 ddp_timeout: 180000000
 
 ### eval
+# eval_dataset: alpaca_en_demo
 # val_size: 0.1
 # per_device_eval_batch_size: 1
 # eval_strategy: steps

src/llamafactory/data/template.py

@@ -985,6 +985,18 @@ register_template(
 )
 
 
+register_template(
+    name="moonlight",
+    format_user=StringFormatter(
+        slots=["<|im_user|>user<|im_middle|>{{content}}<|im_end|><|im_assistant|>assistant<|im_middle|>"]
+    ),
+    format_assistant=StringFormatter(slots=["{{content}}<|im_end|>"]),
+    format_system=StringFormatter(slots=["<|im_system|>system<|im_middle|>{{content}}<|im_end|>"]),
+    default_system="You are a helpful assistant provided by Moonshot-AI.",
+    stop_words=["<|im_end|>"],
+)
+
+
 # copied from vicuna template
 register_template(
     name="llava",

src/llamafactory/extras/constants.py

@@ -1315,6 +1315,21 @@ register_model_group(
 )
 
 
+register_model_group(
+    models={
+        "Moonlight-16B-A3B": {
+            DownloadSource.DEFAULT: "moonshotai/Moonlight-16B-A3B",
+            DownloadSource.MODELSCOPE: "moonshotai/Moonlight-16B-A3B",
+        },
+        "Moonlight-16B-A3B-Instruct": {
+            DownloadSource.DEFAULT: "moonshotai/Moonlight-16B-A3B-Instruct",
+            DownloadSource.MODELSCOPE: "moonshotai/Moonlight-16B-A3B-Instruct",
+        },
+    },
+    template="moonlight",
+)
+
+
 register_model_group(
     models={
         "OLMo-1B": {
@@ -2234,6 +2249,18 @@ register_model_group(
             DownloadSource.DEFAULT: "Qwen/Qwen2.5-VL-72B-Instruct",
             DownloadSource.MODELSCOPE: "Qwen/Qwen2.5-VL-72B-Instruct",
         },
+        "Qwen2.5-VL-3B-Instruct-AWQ": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2.5-VL-3B-Instruct-AWQ",
+            DownloadSource.MODELSCOPE: "Qwen/Qwen2.5-VL-3B-Instruct-AWQ",
+        },
+        "Qwen2.5-VL-7B-Instruct-AWQ": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2.5-VL-7B-Instruct-AWQ",
+            DownloadSource.MODELSCOPE: "Qwen/Qwen2.5-VL-7B-Instruct-AWQ",
+        },
+        "Qwen2.5-VL-72B-Instruct-AWQ": {
+            DownloadSource.DEFAULT: "Qwen/Qwen2.5-VL-72B-Instruct-AWQ",
+            DownloadSource.MODELSCOPE: "Qwen/Qwen2.5-VL-72B-Instruct-AWQ",
+        },
     },
     template="qwen2_vl",
     multimodal=True,