Add files using upload-large-folder tool

opencompass/configs/models/deepseek/lmdeploy_deepseek_7b_chat.py

@@ -0,0 +1,15 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-7b-chat-turbomind',
+        path='deepseek-ai/deepseek-llm-7b-chat',
+        engine_config=dict(max_batch_size=16, tp=1),
+        gen_config=dict(top_k=1, temperature=1e-6, top_p=0.9),
+        max_seq_len=8192,
+        max_out_len=2048,
+        batch_size=16,
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/deepseek/lmdeploy_deepseek_r1_distill_llama_70b.py

@@ -0,0 +1,20 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+from opencompass.utils.text_postprocessors import extract_non_reasoning_content
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-r1-distill-llama-70b-turbomind',
+        path='deepseek-ai/DeepSeek-R1-Distill-Llama-70B',
+        engine_config=dict(session_len=32768, max_batch_size=8, tp=8),
+        gen_config=dict(top_k=1,
+                        temperature=1e-6,
+                        top_p=0.9,
+                        max_new_tokens=32768),
+        max_seq_len=32768,
+        max_out_len=32768,
+        batch_size=8,
+        run_cfg=dict(num_gpus=8),
+        pred_postprocessor=dict(type=extract_non_reasoning_content)
+    )
+]

opencompass/configs/models/deepseek/lmdeploy_deepseek_r1_distill_llama_8b.py

@@ -0,0 +1,20 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+from opencompass.utils.text_postprocessors import extract_non_reasoning_content
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-r1-distill-llama-8b-turbomind',
+        path='deepseek-ai/DeepSeek-R1-Distill-Llama-8B',
+        engine_config=dict(session_len=32768, max_batch_size=8, tp=1),
+        gen_config=dict(top_k=1,
+                        temperature=1e-6,
+                        top_p=0.9,
+                        max_new_tokens=32768),
+        max_seq_len=32768,
+        max_out_len=32768,
+        batch_size=8,
+        run_cfg=dict(num_gpus=1),
+        pred_postprocessor=dict(type=extract_non_reasoning_content)
+    )
+]

opencompass/configs/models/deepseek/lmdeploy_deepseek_r1_distill_qwen_14b.py

@@ -0,0 +1,20 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+from opencompass.utils.text_postprocessors import extract_non_reasoning_content
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-r1-distill-qwen-14b-turbomind',
+        path='deepseek-ai/DeepSeek-R1-Distill-Qwen-14B',
+        engine_config=dict(session_len=32768, max_batch_size=16, tp=2),
+        gen_config=dict(top_k=1,
+                        temperature=1e-6,
+                        top_p=0.9,
+                        max_new_tokens=32768),
+        max_seq_len=32768,
+        max_out_len=32768,
+        batch_size=16,
+        run_cfg=dict(num_gpus=2),
+        pred_postprocessor=dict(type=extract_non_reasoning_content)
+    )
+]

opencompass/configs/models/deepseek/lmdeploy_deepseek_r1_distill_qwen_1_5b.py

@@ -0,0 +1,20 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+from opencompass.utils.text_postprocessors import extract_non_reasoning_content
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-r1-distill-qwen-1_5b-turbomind',
+        path='deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B',
+        engine_config=dict(session_len=32768, max_batch_size=16, tp=1),
+        gen_config=dict(top_k=1,
+                        temperature=1e-6,
+                        top_p=0.9,
+                        max_new_tokens=32768),
+        max_seq_len=32768,
+        max_out_len=32768,
+        batch_size=16,
+        run_cfg=dict(num_gpus=1),
+        pred_postprocessor=dict(type=extract_non_reasoning_content)
+    )
+]

opencompass/configs/models/deepseek/lmdeploy_deepseek_r1_distill_qwen_32b.py

@@ -0,0 +1,20 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+from opencompass.utils.text_postprocessors import extract_non_reasoning_content
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-r1-distill-qwen-32b-turbomind',
+        path='deepseek-ai/DeepSeek-R1-Distill-Qwen-32B',
+        engine_config=dict(session_len=32768, max_batch_size=16, tp=4),
+        gen_config=dict(top_k=1,
+                        temperature=1e-6,
+                        top_p=0.9,
+                        max_new_tokens=32768),
+        max_seq_len=32768,
+        max_out_len=32768,
+        batch_size=16,
+        run_cfg=dict(num_gpus=4),
+        pred_postprocessor=dict(type=extract_non_reasoning_content)
+    )
+]

opencompass/configs/models/deepseek/lmdeploy_deepseek_r1_distill_qwen_7b.py

@@ -0,0 +1,20 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+from opencompass.utils.text_postprocessors import extract_non_reasoning_content
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-r1-distill-qwen-7b-turbomind',
+        path='deepseek-ai/DeepSeek-R1-Distill-Qwen-7B',
+        engine_config=dict(session_len=32768, max_batch_size=8, tp=1),
+        gen_config=dict(top_k=1,
+                        temperature=1e-6,
+                        top_p=0.9,
+                        max_new_tokens=32768),
+        max_seq_len=32768,
+        max_out_len=32768,
+        batch_size=8,
+        run_cfg=dict(num_gpus=1),
+        pred_postprocessor=dict(type=extract_non_reasoning_content)
+    )
+]

opencompass/configs/models/deepseek/lmdeploy_deepseek_series.py

@@ -0,0 +1,23 @@
+from opencompass.models import LmdeployPytorchModel
+
+settings = [
+    ('deepseek-7b-base-hf', 'deepseek-ai/deepseek-llm-7b-base', 1),
+    ('deepseek-67b-base-hf', 'deepseek-ai/deepseek-llm-67b-base', 4),
+]
+
+models = []
+for abbr, path, num_gpus in settings:
+    models.append(
+        dict(
+            type=LmdeployPytorchModel,
+            abbr=abbr,
+            path=path,
+            engine_config=dict(session_len=2048, max_batch_size=16, tp=num_gpus),
+            gen_config=dict(top_k=1, temperature=1, top_p=0.9, max_new_tokens=1024),
+            max_out_len=1024,
+            max_seq_len=2048,
+            batch_size=16,
+            concurrency=16,
+            run_cfg=dict(num_gpus=num_gpus),
+        )
+    )

opencompass/configs/models/deepseek/lmdeploy_deepseek_v2.py

@@ -0,0 +1,18 @@
+# flake8: noqa
+from mmengine.config import read_base
+from opencompass.models import (
+    TurboMindModel,
+)
+lmdeploy_deepseek_v2_model = [
+    dict(
+        type=TurboMindModel,
+        abbr='deepseek-v2-turbomind',
+        path='deepseek-ai/DeepSeek-V2',
+        engine_config=dict(session_len=7168, max_batch_size=4, tp=8, cache_max_entry_count=0.7),
+        gen_config=dict(top_k=1, temperature=1e-6, top_p=0.9),
+        max_seq_len=7168,
+        max_out_len=2048,
+        batch_size=4,
+        run_cfg=dict(num_gpus=8),
+    )
+]

opencompass/configs/models/deepseek/lmdeploy_deepseek_v2_5.py

@@ -0,0 +1,21 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-v2_5-turbomind',
+        path='deepseek-ai/DeepSeek-V2.5',
+        backend='pytorch',
+        engine_config=dict(
+            session_len=7168,
+            max_batch_size=4,
+            tp=8,
+            cache_max_entry_count=0.7,
+        ),
+        gen_config=dict(top_k=1, temperature=1e-6, top_p=0.9),
+        max_seq_len=7168,
+        max_out_len=2048,
+        batch_size=4,
+        run_cfg=dict(num_gpus=8),
+    )
+]

opencompass/configs/models/deepseek/lmdeploy_deepseek_v2_5_1210.py

@@ -0,0 +1,21 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-v2_5-1210-turbomind',
+        path='deepseek-ai/DeepSeek-V2.5-1210',
+        backend='pytorch',
+        engine_config=dict(
+            session_len=7168,
+            max_batch_size=4,
+            tp=8,
+            cache_max_entry_count=0.7,
+        ),
+        gen_config=dict(top_k=1, temperature=1e-6, top_p=0.9),
+        max_seq_len=7168,
+        max_out_len=2048,
+        batch_size=4,
+        run_cfg=dict(num_gpus=8),
+    )
+]

opencompass/configs/models/deepseek/lmdeploy_deepseek_v2_lite.py

@@ -0,0 +1,20 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-v2_lite-chat-turbomind',
+        path='deepseek-ai/DeepSeek-V2-Lite-Chat',
+        engine_config=dict(
+            session_len=7168,
+            max_batch_size=4,
+            tp=2,
+            cache_max_entry_count=0.7,
+        ),
+        gen_config=dict(top_k=1, temperature=1e-6, top_p=0.9),
+        max_seq_len=7168,
+        max_out_len=2048,
+        batch_size=4,
+        run_cfg=dict(num_gpus=2),
+    )
+]

opencompass/configs/models/deepseek/vllm_deepseek_67b_chat.py

@@ -0,0 +1,13 @@
+from opencompass.models import VLLMwithChatTemplate
+
+models = [
+    dict(
+        type=VLLMwithChatTemplate,
+        abbr='deepseek-67b-chat-vllm',
+        path='deepseek-ai/deepseek-llm-67b-chat',
+        max_out_len=1024,
+        batch_size=16,
+        model_kwargs=dict(tensor_parallel_size=4),
+        run_cfg=dict(num_gpus=4),
+    )
+]

opencompass/configs/models/deepseek/vllm_deepseek_7b_chat.py

@@ -0,0 +1,13 @@
+from opencompass.models import VLLMwithChatTemplate
+
+models = [
+    dict(
+        type=VLLMwithChatTemplate,
+        abbr='deepseek-7b-chat-vllm',
+        path='deepseek-ai/deepseek-llm-7b-chat',
+        max_out_len=1024,
+        batch_size=16,
+        model_kwargs=dict(tensor_parallel_size=1),
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/deepseek/vllm_deepseek_moe_16b_base.py

@@ -0,0 +1,15 @@
+from opencompass.models import VLLM
+
+models = [
+    dict(
+        type=VLLM,
+        abbr='deepseek-moe-16b-base-vllm',
+        path='deepseek-ai/deepseek-moe-16b-base',
+        model_kwargs=dict(tensor_parallel_size=1, gpu_memory_utilization=0.6),
+        max_out_len=1024,
+        max_seq_len=8192,
+        batch_size=16,
+        generation_kwargs=dict(temperature=0),
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/deepseek/vllm_deepseek_moe_16b_chat.py

@@ -0,0 +1,13 @@
+from opencompass.models import VLLMwithChatTemplate
+
+models = [
+    dict(
+        type=VLLMwithChatTemplate,
+        abbr='deepseek-moe-16b-chat-vllm',
+        path='deepseek-ai/deepseek-moe-16b-chat',
+        model_kwargs=dict(tensor_parallel_size=1, gpu_memory_utilization=0.6),
+        max_out_len=1024,
+        batch_size=16,
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/deltanet/delta_net-1.3B.py

@@ -0,0 +1,27 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        # HuggingFaceCasualLM
+        path='/mnt/jfzn/msj/delta_net-1.3B-100B',
+        tokenizer_path='/mnt/jfzn/msj/delta_net-1.3B-100B',
+        tokenizer_kwargs=dict(
+            padding_side='left', truncation_side='left',
+        ),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+            device_map='auto',
+            trust_remote_code=True,
+        ),
+        max_seq_len=2048,
+        abbr='delta_net',
+        batch_size=128,
+        # max_out_len=100,
+        run_cfg=dict(num_gpus=1),  # 模型部署在几个GPU上
+    )
+]
+
+
+
+

opencompass/configs/models/deltanet/gdn-1.3B.py

@@ -0,0 +1,24 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        # HuggingFaceCasualLM
+        path='/mnt/jfzn/msj/train_exp/gdn_1B_a800',
+        tokenizer_path='/mnt/jfzn/msj/train_exp/gdn_1B_a800',
+        tokenizer_kwargs=dict(
+            padding_side='left', truncation_side='left',
+        ),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+            device_map='auto',
+            trust_remote_code=True,
+        ),
+        max_seq_len=2048,
+        # opencompass
+        abbr='gated_deltanet',
+        batch_size=128,
+        # max_out_len=100,
+        run_cfg=dict(num_gpus=1),  # 模型部署在几个GPU上
+    )
+]

opencompass/configs/models/deltanet/gla-1.3B.py

@@ -0,0 +1,28 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        # HuggingFaceCasualLM
+        path='/mnt/jfzn/msj/gla-1.3B-100B',
+        tokenizer_path='/mnt/jfzn/msj/gla-1.3B-100B',
+        tokenizer_kwargs=dict(
+            padding_side='left', truncation_side='left',
+        ),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+            device_map='auto',
+            trust_remote_code=True,
+        ),
+        max_seq_len=2048,
+        # opencompass
+        abbr='gla',
+        batch_size=128,
+        # max_out_len=100,
+        run_cfg=dict(num_gpus=1),  # 模型部署在几个GPU上
+    )
+]
+
+
+
+

opencompass/configs/models/deltanet/hgrn2-1.3B.py

@@ -0,0 +1,24 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        # HuggingFaceCasualLM
+        path='/mnt/jfzn/msj/download_model/hgrn2-1.3B-100B',
+        tokenizer_path='/mnt/jfzn/msj/download_model/hgrn2-1.3B-100B',
+        tokenizer_kwargs=dict(
+            padding_side='left', truncation_side='left',
+        ),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+            device_map='auto',
+            trust_remote_code=True,
+        ),
+        max_seq_len=2048,
+        # opencompass
+        abbr='hgrn2',
+        batch_size=128,
+        # max_out_len=100,
+        run_cfg=dict(num_gpus=1),  # 模型部署在几个GPU上
+    )
+]

opencompass/configs/models/deltanet/mask_deltanet-1.3B.py

@@ -0,0 +1,24 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        # HuggingFaceCasualLM
+        path='/mnt/jfzn/msj/train_exp/mask_deltanet_1B_rank4',
+        tokenizer_path='/mnt/jfzn/msj/train_exp/mask_gdn_1B_hrr-rank4',
+        tokenizer_kwargs=dict(
+            padding_side='left', truncation_side='left',
+        ),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+            device_map='auto',
+            trust_remote_code=True,
+        ),
+        max_seq_len=2048,
+        # opencompass
+        abbr='mask_deltanet',
+        batch_size=128,
+        # max_out_len=100,
+        run_cfg=dict(num_gpus=1),  # 模型部署在几个GPU上
+    )
+]

opencompass/configs/models/deltanet/mask_gdn-1.3B.py

@@ -0,0 +1,24 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        # HuggingFaceCasualLM
+        path='/mnt/jfzn/msj/train_exp/mask_gdn_1B_hrr-rank4',
+        tokenizer_path='/mnt/jfzn/msj/train_exp/mask_gdn_1B_hrr-rank4',
+        tokenizer_kwargs=dict(
+            padding_side='left', truncation_side='left',
+        ),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+            device_map='auto',
+            trust_remote_code=True,
+        ),
+        max_seq_len=2048,
+        # opencompass
+        abbr='mask_gdn',
+        batch_size=128,
+        # max_out_len=100,
+        run_cfg=dict(num_gpus=1),  # 模型部署在几个GPU上
+    )
+]

opencompass/configs/models/deltanet/mask_gdn_t-1.3B.py

@@ -0,0 +1,24 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        # HuggingFaceCasualLM
+        path='/mnt/jfzn/msj/train_exp/mask_gdn_1B_hrr4_byt',
+        tokenizer_path='/mnt/jfzn/msj/train_exp/mask_gdn_1B_hrr4_byt',
+        tokenizer_kwargs=dict(
+            padding_side='left', truncation_side='left',
+        ),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+            device_map='auto',
+            trust_remote_code=True,
+        ),
+        max_seq_len=2048,
+        # opencompass
+        abbr='mask_gdn',
+        batch_size=128,
+        # max_out_len=100,
+        run_cfg=dict(num_gpus=1),  # 模型部署在几个GPU上
+    )
+]

opencompass/configs/models/deltanet/retnet-1.3B.py

@@ -0,0 +1,24 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        # HuggingFaceCasualLM
+        path='/mnt/jfzn/msj/download_model/retnet-1.3B-100B',
+        tokenizer_path='/mnt/jfzn/msj/download_model/retnet-1.3B-100B',
+        tokenizer_kwargs=dict(
+            padding_side='left', truncation_side='left',
+        ),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+            device_map='auto',
+            trust_remote_code=True,
+        ),
+        max_seq_len=2048,
+        # opencompass
+        abbr='retnet',
+        batch_size=128,
+        # max_out_len=48,
+        run_cfg=dict(num_gpus=1),  # 模型部署在几个GPU上
+    )
+]

opencompass/configs/models/deltanet/transformer++-1.3B.py

@@ -0,0 +1,28 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        # HuggingFaceCasualLM
+        path='/mnt/jfzn/msj/transformer-1.3B-100B',
+        tokenizer_path='/mnt/jfzn/msj/transformer-1.3B-100B',
+        tokenizer_kwargs=dict(
+            padding_side='left', truncation_side='left',
+        ),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+            device_map='auto',
+            trust_remote_code=True,
+        ),
+        max_seq_len=2048,
+        # opencompass
+        abbr='transformer',
+        batch_size=128,
+        # max_out_len=100,
+        run_cfg=dict(num_gpus=1),  # 模型部署在几个GPU上
+    )
+]
+
+
+
+

opencompass/configs/models/falcon/hf_falcon_40b.py

@@ -0,0 +1,12 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        abbr='falcon-40b-hf',
+        path='tiiuae/falcon-40b',
+        max_out_len=1024,
+        batch_size=8,
+        run_cfg=dict(num_gpus=4),
+    )
+]

opencompass/configs/models/falcon/hf_falcon_7b.py

@@ -0,0 +1,12 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        abbr='falcon-7b-hf',
+        path='tiiuae/falcon-7b',
+        max_out_len=1024,
+        batch_size=8,
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/gemini/gemini_1_5_flash.py

@@ -0,0 +1,22 @@
+from opencompass.models import Gemini
+
+api_meta_template = dict(round=[
+    dict(role='HUMAN', api_role='HUMAN'),
+    dict(role='BOT', api_role='BOT', generate=True),
+], )
+
+models = [
+    dict(
+        abbr='gemini-1.5-flash',
+        type=Gemini,
+        path='gemini-1.5-flash',
+        key=
+        'ENV',  # The key will be obtained from $GEMINI_API_KEY, but you can write down your key here as well
+        meta_template=api_meta_template,
+        query_per_second=15,
+        max_out_len=100,
+        max_seq_len=2048,
+        batch_size=1,
+        temperature=1,
+    )
+]

opencompass/configs/models/gemini/gemini_1_5_pro.py

@@ -0,0 +1,22 @@
+from opencompass.models import Gemini
+
+api_meta_template = dict(round=[
+    dict(role='HUMAN', api_role='HUMAN'),
+    dict(role='BOT', api_role='BOT', generate=True),
+], )
+
+models = [
+    dict(
+        abbr='gemini-1.5-pro',
+        type=Gemini,
+        path='gemini-1.5-pro',
+        key=
+        'ENV',  # The key will be obtained from $GEMINI_API_KEY, but you can write down your key here as well
+        meta_template=api_meta_template,
+        query_per_second=2,
+        max_out_len=100,
+        max_seq_len=2048,
+        batch_size=1,
+        temperature=1,
+    )
+]

opencompass/configs/models/gemini/gemini_pro.py

@@ -0,0 +1,22 @@
+from opencompass.models import Gemini
+
+
+api_meta_template = dict(
+    round=[
+            dict(role='HUMAN', api_role='HUMAN'),
+            dict(role='BOT', api_role='BOT', generate=True),
+    ],
+)
+
+models = [
+    dict(abbr='gemini',
+    type=Gemini,
+    path='gemini-pro',
+    key='ENV',  # The key will be obtained from $GEMINI_API_KEY, but you can write down your key here as well
+    meta_template=api_meta_template,
+    query_per_second=16,
+    max_out_len=100,
+    max_seq_len=2048,
+    batch_size=1,
+    temperature=1,)
+]

opencompass/configs/models/gemma/hf_gemma2_27b.py

@@ -0,0 +1,15 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        abbr='gemma2-27b-hf',
+        path='google/gemma-2-27b',
+        max_out_len=1024,
+        batch_size=4,
+        run_cfg=dict(num_gpus=2),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+        ),
+    )
+]

opencompass/configs/models/gemma/hf_gemma2_27b_it.py

@@ -0,0 +1,16 @@
+from opencompass.models import HuggingFacewithChatTemplate
+
+models = [
+    dict(
+        type=HuggingFacewithChatTemplate,
+        abbr='gemma2-27b-it-hf',
+        path='google/gemma-2-27b-it',
+        max_out_len=2048,
+        batch_size=1,
+        run_cfg=dict(num_gpus=2),
+        stop_words=['<end_of_turn>'],
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+        )
+    )
+]

opencompass/configs/models/gemma/hf_gemma2_2b.py

@@ -0,0 +1,15 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        abbr='gemma2-2b-hf',
+        path='google/gemma-2-2b',
+        max_out_len=1024,
+        batch_size=4,
+        run_cfg=dict(num_gpus=1),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+        ),
+    )
+]

opencompass/configs/models/gemma/hf_gemma2_2b_it.py

@@ -0,0 +1,16 @@
+from opencompass.models import HuggingFacewithChatTemplate
+
+models = [
+    dict(
+        type=HuggingFacewithChatTemplate,
+        abbr='gemma2-2b-it-hf',
+        path='google/gemma-2-2b-it',
+        max_out_len=2048,
+        batch_size=1,
+        run_cfg=dict(num_gpus=1),
+        stop_words=['<end_of_turn>'],
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+        )
+    )
+]

opencompass/configs/models/gemma/hf_gemma2_9b.py

@@ -0,0 +1,15 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        abbr='gemma2-9b-hf',
+        path='google/gemma-2-9b',
+        max_out_len=1024,
+        batch_size=4,
+        run_cfg=dict(num_gpus=1),
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+        ),
+    )
+]

opencompass/configs/models/gemma/hf_gemma2_9b_it.py

@@ -0,0 +1,16 @@
+from opencompass.models import HuggingFacewithChatTemplate
+
+models = [
+    dict(
+        type=HuggingFacewithChatTemplate,
+        abbr='gemma2-9b-it-hf',
+        path='google/gemma-2-9b-it',
+        max_out_len=2048,
+        batch_size=1,
+        run_cfg=dict(num_gpus=1),
+        stop_words=['<end_of_turn>'],
+        model_kwargs=dict(
+            torch_dtype='torch.bfloat16',
+        )
+    )
+]

opencompass/configs/models/gemma/hf_gemma_2b.py

@@ -0,0 +1,12 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        abbr='gemma-2b-hf',
+        path='google/gemma-2b',
+        max_out_len=1024,
+        batch_size=8,
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/gemma/hf_gemma_2b_it.py

@@ -0,0 +1,12 @@
+from opencompass.models import HuggingFacewithChatTemplate
+
+models = [
+    dict(
+        type=HuggingFacewithChatTemplate,
+        abbr='gemma-2b-it-hf',
+        path='google/gemma-2b-it',
+        max_out_len=1024,
+        batch_size=1,
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/gemma/hf_gemma_7b.py

@@ -0,0 +1,12 @@
+from opencompass.models import HuggingFaceBaseModel
+
+models = [
+    dict(
+        type=HuggingFaceBaseModel,
+        abbr='gemma-7b-hf',
+        path='google/gemma-7b',
+        max_out_len=1024,
+        batch_size=8,
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/gemma/hf_gemma_7b_it.py

@@ -0,0 +1,12 @@
+from opencompass.models import HuggingFacewithChatTemplate
+
+models = [
+    dict(
+        type=HuggingFacewithChatTemplate,
+        abbr='gemma-7b-it-hf',
+        path='google/gemma-7b-it',
+        max_out_len=1024,
+        batch_size=1,
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/gemma/lmdeploy_gemma_27b.py

@@ -0,0 +1,17 @@
+from opencompass.models import TurboMindModel
+
+models = [
+    dict(
+        type=TurboMindModel,
+        abbr='gemma-2-27b-turbomind',
+        path='google/gemma-2-27b',
+        engine_config=dict(session_len=16384, max_batch_size=16, tp=2),
+        gen_config=dict(
+            top_k=1, temperature=1e-6, top_p=0.9, max_new_tokens=4096
+        ),
+        max_seq_len=16384,
+        max_out_len=4096,
+        batch_size=16,
+        run_cfg=dict(num_gpus=2),
+    )
+]

opencompass/configs/models/gemma/lmdeploy_gemma_27b_it.py

@@ -0,0 +1,17 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='gemma-2-27b-it-turbomind',
+        path='google/gemma-2-27b-it',
+        engine_config=dict(session_len=16384, max_batch_size=16, tp=1),
+        gen_config=dict(
+            top_k=1, temperature=1e-6, top_p=0.9, max_new_tokens=4096
+        ),
+        max_seq_len=16384,
+        max_out_len=4096,
+        batch_size=16,
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/gemma/lmdeploy_gemma_9b.py

@@ -0,0 +1,17 @@
+from opencompass.models import TurboMindModel
+
+models = [
+    dict(
+        type=TurboMindModel,
+        abbr='gemma-2-9b-turbomind',
+        path='google/gemma-2-9b',
+        engine_config=dict(session_len=16384, max_batch_size=16, tp=1),
+        gen_config=dict(
+            top_k=1, temperature=1e-6, top_p=0.9, max_new_tokens=4096
+        ),
+        max_seq_len=16384,
+        max_out_len=4096,
+        batch_size=16,
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/gemma/lmdeploy_gemma_9b_it.py

@@ -0,0 +1,17 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='gemma-2-9b-it-turbomind',
+        path='google/gemma-2-9b-it',
+        engine_config=dict(session_len=16384, max_batch_size=16, tp=1),
+        gen_config=dict(
+            top_k=1, temperature=1e-6, top_p=0.9, max_new_tokens=4096
+        ),
+        max_seq_len=16384,
+        max_out_len=4096,
+        batch_size=16,
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/gemma/vllm_gemma_2b.py

@@ -0,0 +1,15 @@
+from opencompass.models import VLLM
+
+models = [
+    dict(
+        type=VLLM,
+        abbr='gemma-2b-vllm',
+        path='google/gemma-2b',
+        model_kwargs=dict(tensor_parallel_size=1, gpu_memory_utilization=0.5),
+        max_out_len=1024,
+        max_seq_len=8192,
+        batch_size=16,
+        generation_kwargs=dict(temperature=0),
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/gemma/vllm_gemma_2b_it.py

@@ -0,0 +1,14 @@
+from opencompass.models import VLLMwithChatTemplate
+
+models = [
+    dict(
+        type=VLLMwithChatTemplate,
+        abbr='gemma-2b-it-vllm',
+        path='google/gemma-2b-it',
+        model_kwargs=dict(tensor_parallel_size=1, gpu_memory_utilization=0.5),
+        max_out_len=1024,
+        batch_size=16,
+        generation_kwargs=dict(temperature=0),
+        run_cfg=dict(num_gpus=1),
+    )
+]

opencompass/configs/models/gemma/vllm_gemma_3_12b_it.py

@@ -0,0 +1,16 @@
+from opencompass.models import VLLMwithChatTemplate
+
+models = [
+    dict(
+        type=VLLMwithChatTemplate,
+        abbr='gemma-3-12b-it-vllm',
+        path='google/gemma-3-12b-it',
+        model_kwargs=dict(tensor_parallel_size=4,
+                          # for long context
+                          rope_scaling={'factor': 8.0, 'rope_type': 'linear'}),
+        max_out_len=4096,
+        batch_size=1,
+        generation_kwargs=dict(temperature=0),
+        run_cfg=dict(num_gpus=4),
+    )
+]

opencompass/configs/models/gemma/vllm_gemma_3_27b_it.py

@@ -0,0 +1,16 @@
+from opencompass.models import VLLMwithChatTemplate
+
+models = [
+    dict(
+        type=VLLMwithChatTemplate,
+        abbr='gemma-3-27b-it-vllm',
+        path='google/gemma-3-27b-it',
+        model_kwargs=dict(tensor_parallel_size=4,
+                          # for long context
+                          rope_scaling={'factor': 8.0, 'rope_type': 'linear'}),
+        max_out_len=4096,
+        batch_size=1,
+        generation_kwargs=dict(temperature=0),
+        run_cfg=dict(num_gpus=4),
+    )   
+]

opencompass/configs/models/gemma/vllm_gemma_3_4b_it.py

@@ -0,0 +1,17 @@
+from opencompass.models import VLLMwithChatTemplate
+
+models = [
+    dict(
+        type=VLLMwithChatTemplate,
+        abbr='gemma-3-4b-it-vllm',
+        path='google/gemma-3-4b-it',
+        model_kwargs=dict(tensor_parallel_size=2, 
+                          # for long context
+                          rope_scaling={'factor': 8.0, 'rope_type': 'linear'}),
+        max_seq_len=140000,
+        max_out_len=4096,
+        batch_size=1,
+        generation_kwargs=dict(temperature=0),
+        run_cfg=dict(num_gpus=2),
+    )
+]

opencompass/configs/models/gemma/vllm_gemma_7b.py

@@ -0,0 +1,15 @@
+from opencompass.models import VLLM
+
+models = [
+    dict(
+        type=VLLM,
+        abbr='gemma-7b-vllm',
+        path='google/gemma-7b',
+        model_kwargs=dict(tensor_parallel_size=1, gpu_memory_utilization=0.5),
+        max_out_len=1024,
+        max_seq_len=8192,
+        batch_size=16,
+        generation_kwargs=dict(temperature=0),
+        run_cfg=dict(num_gpus=1),
+    )
+]