[BugFix][Models] Add tie_word_embeddings for lmhead

[DrRyanHuang]

Motivation

动态图运行 Qwen2-0.5B 报错：

RuntimeError: (PreconditionNotMet) Tensor's dimension is out of bound.Tensor's dimension must be equal or less than the size of its memory.But received Tensor's dimension is 272269312, memory's size is 0.
  [Hint: Expected numel() * SizeOf(dtype()) <= memory_size(), but received numel() * SizeOf(dtype()):272269312 > memory_size():0.] (at /workspace/Paddle/paddle/phi/core/dense_tensor_impl.cc:48)
  [operator < linear > error]

定位到是 self.lm_head.weight 权重没有初始化

Qwen2-0.5B / 1.5B 等小模型为了减少总参数量，提升小模型核心部分的参数占比，开启 tie_word_embeddings=True 将输入 Embedding 参数与输出 lm_head 的参数共享

#3502 没有考虑 Qwen2-0.5B / 1.5B 这类小模型，没有像 #3255 一样给 lm_head 做初始化，故本PR添加

Modifications

给 Qwen2 组网部分添加 lm_head 的权重初始化

Usage or Command

MODEL=/workspace/MODELS/Qwen2-0.5B-Instruct
rm -rf log/*

export export FD_USE_MACHETE=0
export CUDA_VISIBLE_DEVICES=0
export PORT=39905

python -m fastdeploy.entrypoints.openai.api_server \
  --model $MODEL \
  --metrics-port 39717 \
  --port $PORT \
  --engine-worker-queue-port 39719 \
  --max-model-len 32768 \
  --max-num-seqs 256 \
  --kv-cache-ratio 0.75 \
  --tensor-parallel-size 1 \
  --graph-optimization-config '{"graph_opt_level": 0, "use_cudagraph": false, "full_cuda_graph": false}' \

Accuracy Tests

Checklist

• Add at least a tag in the PR title.
  • Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
  • You can add new tags based on the PR content, but the semantics must be clear.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[paddle-bot]

Thanks for your contribution!

[gongshaotian]

LGTM