**Note**: The LoRA configuration folder should be specified by exporting `VLLM_TUNED_CONFIG_FOLDER=/path/to/configs`. Without this, the shrink/expand kernels will use default configurations.
**Note**: The LoRA configuration folder should be specified by exporting `VLLM_TUNED_CONFIG_FOLDER=/path/to/configs`.
Without this, the shrink/expand kernels will use default configurations.
## Tuning Process
## Tuning Process
Multi-lora shrink/expand Triton kernel tuning follows a similar methodology from [Triton MoE tuning](https://github.com/vllm-project/vllm/blob/main/benchmarks/kernels/benchmark_moe.py).
Multi-lora shrink/expand Triton kernel tuning follows a similar methodology from
1. Define the searching space. Here is an example of searching space:
Define the searching space. An example searching space:
```python
```python
block_m_range=[16,32,64,128,256]
block_m_range=[16,32,64,128,256]
block_n_range=[32,64,128,256]
block_n_range=[32,64,128,256]
block_k_range=[32,64,128,256]
block_k_range=[32,64,128,256]
num_warps_range=[4,8]
num_warps_range=[4,8]
num_stage_range=[2,3,4,5]
num_stage_range=[2,3,4,5]
num_ctas_range=[1]
num_ctas_range=[1]
split_k_range=[4,8,16,32,64]
split_k_range=[4,8,16,32,64]
```
```
**Step 2**
2. Get all hidden_state sizes and num_slices that the target model uses for a specific TP size.
Get all hidden_state sizes and num_slices that the target model uses for a specific TP size.
For example, we can aquire those info by simply checking [add_lora_linear](https://github.com/li2haipeng/vllm/blob/multi_lora_v01011/vllm/lora/punica_wrapper/punica_gpu.py#L192):
For example, you can acquire the info by simply checking
3. Benchmark the shrink/expand kernel runtime with different kernel configurations generated from the pre-defined search space
Benchmark the shrink/expand kernel runtime with different kernel configurations generated from the pre-defined search space by performing a grid search to find the optimal kernel configuration. vLLM's [benchmark_lora.py](https://github.com/vllm-project/vllm/blob/main/benchmarks/kernels/benchmark_lora.py) can be used to search for configurations for different shapes.
by performing a grid search to find the optimal kernel configuration.
For `expand`, the config fileis named as `{gpu_name}_EXPAND_{add_input}.json`, e.g. `NVIDIA_H200_EXPAND_TRUE.json`.
The `gpu_name` can be automatically detected by calling `torch.cuda.get_device_name()`.
For `fused_moe_lora_w13_shrink`, the config file is named as `{gpu_name}_FUSED_MOE_LORA_W13_SHRINK.json`, e.g. `NVIDIA_H200_FUSED_MOE_LORA_W13_SHRINK.json`.
### JSON Structure
For `fused_moe_lora_w13_expand`, the config file is named as `{gpu_name}_FUSED_MOE_LORA_W13_EXPAND.json`, e.g. `NVIDIA_H200_FUSED_MOE_LORA_W13_EXPAND.json`.
Optimal kernel configuration files are saved as JSON files with the structure `config_data[max_loras][num_slices][m][k][n][i]`,
For `fused_moe_lora_w2_shrink`, the config file is named as `{gpu_name}_FUSED_MOE_LORA_W2_SHRINK.json`, e.g. `NVIDIA_H200_FUSED_MOE_LORA_W2_SHRINK.json`.
For `fused_moe_lora_w2_expand`, the config file is named as `{gpu_name}_FUSED_MOE_LORA_W2_EXPAND.json`, e.g. `NVIDIA_H200_FUSED_MOE_LORA_W2_EXPAND.json`.
The `gpu_name` can be automatically detected by calling `torch.cuda.get_device_name()`
### Json Structure
Optimal kernel configuration files are saved as JSON files with the structure `config_data[max_loras][num_slices][m][k][n][i]`
where `i` is an optional dimension in the `fused_moe_lora` configuration, representing the intermediate size of the MoE layer.
where `i` is an optional dimension in the `fused_moe_lora` configuration, representing the intermediate size of the MoE layer.
windsonsea <haifeng.yao@daocloud.io>