[Bug] InternLM2.5-20b-chat 长文本推理启动报错 `Illegal instruction`

### Checklist

- [X] 1. I have searched related issues but cannot get the expected help.
- [X] 2. The bug has not been fixed in the latest version.
- [X] 3. Please note that if the bug-related issue you submitted lacks corresponding environment info and a minimal reproducible demo, it will be challenging for us to reproduce and resolve the issue, reducing the likelihood of receiving feedback.

### Describe the bug

### LMDeploy 镜像:
```
openmmlab/lmdeploy:v0.6.0-cu12
```

### 命令

参考：https://lmdeploy.readthedocs.io/zh-cn/v0.6.0/advance/long_context.html。

使用k8s deployment部署。

```bash
lmdeploy serve api_server /model-cache/internlm2_5-20b-chat --model-name internlm/internlm2_5-20b-chat \
--server-port 80 --tp 4 --cache-max-entry-count 0.7 --session-len 327680 \
--rope-scaling-factor 2.5 --max-batch-size 2 --log-level INFO
```

启动后，报错：`Illegal instruction`后退出。pod CrashLoopBackOff。

### GPU 资源

```bash
NVIDIA GeForce RTX 4090(24564MiB) * 4

```

### Reproduction

```bash
lmdeploy serve api_server /model-cache/internlm2_5-20b-chat --model-name internlm/internlm2_5-20b-chat \
--server-port 80 --tp 4 --cache-max-entry-count 0.7 --session-len 327680 \
--rope-scaling-factor 2.5 --max-batch-size 2 --log-level INFO
```

### Environment

```Shell
### lmdeploy check_env

Python: 3.10.12 (main, Jul 29 2024, 16:56:48) [GCC 11.4.0]
CUDA available: True
MUSA available: False
numpy_random_seed: 2147483648
GPU 0,1,2,3: NVIDIA GeForce RTX 4090 D
CUDA_HOME: /usr/local/cuda
NVCC: Cuda compilation tools, release 12.4, V12.4.131
GCC: x86_64-linux-gnu-gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0
PyTorch: 2.3.0+cu121
PyTorch compiling details: PyTorch built with:
- GCC 9.3
- C++ Version: 201703
- Intel(R) oneAPI Math Kernel Library Version 2022.2-Product Build 20220804 for Intel(R) 64 architecture applications
- Intel(R) MKL-DNN v3.3.6 (Git Hash 86e6af5974177e513fd3fee58425e1063e7f1361)
- OpenMP 201511 (a.k.a. OpenMP 4.5)
- LAPACK is enabled (usually provided by MKL)
- NNPACK is enabled
- CPU capability usage: AVX512
- CUDA Runtime 12.1
- NVCC architecture flags: -gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86;-gencode;arch=compute_90,code=sm_90
- CuDNN 8.9.2
- Magma 2.6.1
- Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=12.1, CUDNN_VERSION=8.9.2, CXX_COMPILER=/opt/rh/devtoolset-9/root/usr/bin/c++, CXX_FLAGS= -D_GLIBCXX_USE_CXX11_ABI=0 -fabi-version=11 -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO -DLIBKINETO_NOROCTRACER -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -O2 -fPIC -Wall -Wextra -Werror=return-type -Werror=non-virtual-dtor -Werror=bool-operation -Wnarrowing -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-strict-overflow -Wno-strict-aliasing -Wno-stringop-overflow -Wsuggest-override -Wno-psabi -Wno-error=pedantic -Wno-error=old-style-cast -Wno-missing-braces -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=2.3.0, USE_CUDA=ON, USE_CUDNN=ON, USE_CUSPARSELT=1, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_GLOO=ON, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=1, USE_NNPACK=ON, USE_OPENMP=ON, USE_ROCM=OFF, USE_ROCM_KERNEL_ASSERT=OFF,

TorchVision: 0.18.0+cu121
LMDeploy: 0.6.0+e2aa4bd
transformers: 4.44.2
gradio: 4.44.0
fastapi: 0.114.1
pydantic: 2.9.1
triton: 2.3.0
NVIDIA Topology:
GPU0 GPU1 GPU2 GPU3 NIC0 NIC1 CPU Affinity NUMA Affinity GPU NUMA ID
GPU0 X NODE NODE NODE NODE NODE 32-63,96-127 1 N/A
GPU1 NODE X NODE NODE NODE NODE 32-63,96-127 1 N/A
GPU2 NODE NODE X NODE NODE NODE 32-63,96-127 1 N/A
GPU3 NODE NODE NODE X NODE NODE 32-63,96-127 1 N/A
NIC0 NODE NODE NODE NODE X PIX
NIC1 NODE NODE NODE NODE PIX X

Legend:

X = Self
SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
PIX = Connection traversing at most a single PCIe bridge
NV# = Connection traversing a bonded set of # NVLinks

NIC Legend:

NIC0: rocep171s0f0
NIC1: rocep171s0f1
```
```

### Error traceback

```Shell
2024-12-16 03:48:26,141 - lmdeploy - INFO - input backend=turbomind, backend_config=TurbomindEngineConfig(model_format=None, tp=4, session_len=327680, max_batch_size=2, cache_max_entry_count=0.7, cache_chunk_size=-1, cache_block_seq_len=64, enable_prefix_caching=False, quant_policy=0, rope_scaling_factor=2.5, use_logn_attn=False, download_dir=None, revision=None, max_prefill_token_num=8192, num_tokens_per_iter=0, max_prefill_iters=1)
2024-12-16 03:48:26,141 - lmdeploy - INFO - input chat_template_config=None
2024-12-16 03:48:26,250 - lmdeploy - INFO - updated chat_template_onfig=ChatTemplateConfig(model_name='internlm2', system=None, meta_instruction=None, eosys=None, user=None, eoh=None, assistant=None, eoa=None, separator=None, capability=None, stop_words=None)
2024-12-16 03:48:26,250 - lmdeploy - INFO - model_source: hf_model
2024-12-16 03:48:26,672 - lmdeploy - INFO - turbomind model config:

{
"model_config": {
"model_name": "",
"chat_template": "",
"model_arch": "InternLM2ForCausalLM",
"head_num": 48,
"kv_head_num": 8,
"hidden_units": 6144,
"vocab_size": 92544,
"num_layer": 48,
"inter_size": 16384,
"norm_eps": 1e-05,
"attn_bias": 0,
"start_id": 1,
"end_id": 2,
"size_per_head": 128,
"group_size": 128,
"weight_type": "bf16",
"session_len": 327680,
"tp": 4,
"model_format": "hf"
},
"attention_config": {
"rotary_embedding": 128,
"rope_theta": 50000000.0,
"max_position_embeddings": 32768,
"original_max_position_embeddings": 0,
"rope_scaling_type": "dynamic",
"rope_scaling_factor": 2.5,
"use_dynamic_ntk": 1,
"low_freq_factor": 1.0,
"high_freq_factor": 1.0,
"use_logn_attn": 0,
"cache_block_seq_len": 64
},
"lora_config": {
"lora_policy": "",
"lora_r": 0,
"lora_scale": 0.0,
"lora_max_wo_r": 0,
"lora_rank_pattern": "",
"lora_scale_pattern": ""
},
"engine_config": {
"model_format": null,
"tp": 4,
"session_len": 327680,
"max_batch_size": 2,
"cache_max_entry_count": 0.7,
"cache_chunk_size": -1,
"cache_block_seq_len": 64,
"enable_prefix_caching": false,
"quant_policy": 0,
"rope_scaling_factor": 2.5,
"use_logn_attn": false,
"download_dir": null,
"revision": null,
"max_prefill_token_num": 8192,
"num_tokens_per_iter": 8192,
"max_prefill_iters": 40
}
}
[TM][WARNING] [LlamaTritonModel] `max_context_token_num` is not set, default to 327680.
[TM][INFO] Model:
head_num: 48
kv_head_num: 8
size_per_head: 128
inter_size: 16384
num_layer: 48
vocab_size: 92544
attn_bias: 0
max_batch_size: 2
max_prefill_token_num: 8192
max_context_token_num: 327680
num_tokens_per_iter: 8192
max_prefill_iters: 40
session_len: 327680
cache_max_entry_count: 0.7
cache_block_seq_len: 64
cache_chunk_size: -1
enable_prefix_caching: 0
start_id: 1
tensor_para_size: 4
pipeline_para_size: 1
enable_custom_all_reduce: 0
model_name:
model_dir:
quant_policy: 0
group_size: 128

Illegal instruction
```