Add MiniCPM-V-2_6 to iGPU Perf (#11810)

* Add MiniCPM-V-2_6 to iGPU Perf

* keep last model in yaml

* fix MINICPM_V_IDS

* Restore tested model list

* Small fix

---------

Co-authored-by: Yuwen Hu <yuwen.hu@intel.com>

python/llm/dev/benchmark/all-in-one/run.py

@@ -992,6 +992,13 @@ def run_transformer_int4_gpu_win(repo_id,
                                                      trust_remote_code=True, use_cache=True, cpu_embedding=cpu_embedding).eval()
         tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
         model = model.to('xpu')
+    elif repo_id in MINICPM_V_IDS:
+        model = AutoModel.from_pretrained(model_path, optimize_model=True, load_in_low_bit=low_bit,
+                                          modules_to_not_convert=["vpm", "resampler"],
+                                          trust_remote_code=True, use_cache=True, cpu_embedding=cpu_embedding).eval()
+        tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+        model = model.to('xpu')
+        model = model.llm
     else:
         model = AutoModelForCausalLM.from_pretrained(model_path, optimize_model=True, load_in_low_bit=low_bit,
                                                      trust_remote_code=True, use_cache=True, cpu_embedding=cpu_embedding).eval()
@@ -1108,6 +1115,14 @@ def run_transformer_int4_fp16_gpu_win(repo_id,
                                                      torch_dtype=torch.float16).eval()
         tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
         model = model.to('xpu')
+    elif repo_id in MINICPM_V_IDS:
+        model = AutoModel.from_pretrained(model_path, optimize_model=True, load_in_low_bit=low_bit,
+                                          modules_to_not_convert=["vpm", "resampler"],
+                                          trust_remote_code=True, use_cache=True, cpu_embedding=cpu_embedding,
+                                          torch_dtype=torch.float16).eval()
+        tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+        model = model.to('xpu')
+        model = model.llm
     else:
         model = AutoModelForCausalLM.from_pretrained(model_path, optimize_model=True, load_in_low_bit=low_bit,
                                                      trust_remote_code=True, use_cache=True, cpu_embedding=cpu_embedding,

python/llm/test/benchmark/igpu-perf/1024-128_437.yaml

@@ -5,6 +5,7 @@ repo_id:
   - 'microsoft/Phi-3-mini-4k-instruct'
   - 'microsoft/Phi-3-mini-128k-instruct'
   - 'microsoft/phi-3-vision-128k-instruct'
+  - 'openbmb/MiniCPM-V-2_6'
 local_model_hub: 'path to your local model hub'
 warm_up: 1
 num_trials: 3

python/llm/test/benchmark/igpu-perf/1024-128_int4_fp16_437.yaml

@@ -5,6 +5,7 @@ repo_id:
   - 'microsoft/Phi-3-mini-4k-instruct'
   - 'microsoft/Phi-3-mini-128k-instruct'
   - 'microsoft/phi-3-vision-128k-instruct'
+  - 'openbmb/MiniCPM-V-2_6'
 local_model_hub: 'path to your local model hub'
 warm_up: 1
 num_trials: 3

python/llm/test/benchmark/igpu-perf/2048-256_int4_fp16_437.yaml

@@ -5,6 +5,7 @@ repo_id:
   - 'microsoft/Phi-3-mini-4k-instruct'
   - 'microsoft/Phi-3-mini-128k-instruct'
   - 'microsoft/phi-3-vision-128k-instruct'
+  - 'openbmb/MiniCPM-V-2_6'
 local_model_hub: 'path to your local model hub'
 warm_up: 1
 num_trials: 3

python/llm/test/benchmark/igpu-perf/3072-384_int4_fp16_437.yaml

@@ -5,6 +5,7 @@ repo_id:
   - 'microsoft/Phi-3-mini-4k-instruct'
   - 'microsoft/Phi-3-mini-128k-instruct'
   - 'microsoft/phi-3-vision-128k-instruct'
+  - 'openbmb/MiniCPM-V-2_6'
 local_model_hub: 'path to your local model hub'
 warm_up: 1
 num_trials: 3

python/llm/test/benchmark/igpu-perf/32-32_int4_fp16_437.yaml

@@ -5,6 +5,7 @@ repo_id:
   - 'microsoft/Phi-3-mini-4k-instruct'
   - 'microsoft/Phi-3-mini-128k-instruct'
   - 'microsoft/phi-3-vision-128k-instruct'
+  - 'openbmb/MiniCPM-V-2_6'
 local_model_hub: 'path to your local model hub'
 warm_up: 3
 num_trials: 5

python/llm/test/benchmark/igpu-perf/4096-512_int4_fp16_437.yaml

@@ -5,6 +5,7 @@ repo_id:
   - 'microsoft/Phi-3-mini-4k-instruct'
   - 'microsoft/Phi-3-mini-128k-instruct'
   - 'microsoft/phi-3-vision-128k-instruct'
+  - 'openbmb/MiniCPM-V-2_6'
 local_model_hub: 'path to your local model hub'
 warm_up: 1
 num_trials: 3