[Qwen2_5] support Qwen2.5-3B

models/Qwen2_5/compile/README.md

@@ -33,8 +33,15 @@ python3 -m dfss --url=open@sophgo.com:/ext_model_information/LLM/LLM-TPU/qwen2.5
 
 ### 全流程编译脚本
 以上步骤可以通过运行compile文件夹下的run_compile.sh完成，具体命令是：
+
+迁移Qwen2.5-7B
+```shell
+./run_compile.sh --model_name qwen2.5-7b --seq_length 512 --model_path your_model_path --tpu_mlir_path your_tpu_mlir_path
+```
+
+迁移Qwen2.5-3B，且输入+输出最大长度为4096
 ``` shell
-./run_compile.sh --model_name qwen2.5-7b --seq_length 512 --model_path your model path --tpu_mlir_path your tpu_mlir path
+./run_compile.sh --model_name qwen2.5-3b --seq_length 4096
 ```
 如果没有填写model_path，脚本会从modelscope下载模型；
 如果没有填写tpu_mlir_path，脚本会通过dfss下载对应的tpu_mlir压缩包并解压

models/Qwen2_5/compile/compile.sh

@@ -59,20 +59,24 @@ if [[ -z "$seq_length" ]]; then
     exit 1
 fi
 
-if [ "$name" = "qwen2.5-7b" ]; then
-  num_layers=28
-  hidden_size=3584
-  echo "Compile Qwen2.5-7B"
-elif [ "$name" = "qwen2.5-14b" ]; then
+if [ "$name" = "qwen2.5-14b" ]; then
   num_layers=48
   hidden_size=5120
   echo "Compile Qwen2.5-14B"
+elif [ "$name" = "qwen2.5-7b" ]; then
+  num_layers=28
+  hidden_size=3584
+  echo "Compile Qwen2.5-7B"
+elif [ "$name" = "qwen2.5-3b" ]; then
+  num_layers=36
+  hidden_size=2048
+  echo "Compile Qwen2.5-3B"
 elif [ "$name" = "qwen2.5-1.5b" ]; then
   num_layers=28
   hidden_size=1536
   echo "Compile Qwen2.5-1.5B"
 else
-  >&2 echo -e "Error: Invalid name $name, the input name must be \033[31mqwen2.5-7b|qwen2.5-1.5b|qwen2.5-14b\033[0m"
+  >&2 echo -e "Error: Invalid name $name, the input name must be \033[31mqwen2.5-14b|qwen2.5-7b|qwen2.5-3b|qwen2.5-1.5b\033[0m"
   exit 1
 fi
 

models/Qwen2_5/compile/files/Qwen2.5-3B-Instruct/config.json

@@ -0,0 +1,27 @@
+{
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 70,
+  "model_type": "qwen2",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 36,
+  "num_key_value_heads": 2,
+  "rms_norm_eps": 1e-06,
+  "rope_theta": 1000000.0,
+  "sliding_window": 32768,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.43.1",
+  "use_cache": true,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}