[Feaure] Add new models: baichuan2, tigerbot, vicuna v1.5 (#373)

* add bag of new models: baichuan2, tigerbot, vicuna v1.5

* update

* re-organize models

* update readme

* update

configs/models/hf_tigerbot_7b_sft.py → configs/models/tigerbot/hf_tigerbot_7b_sft.py

@@ -23,7 +23,7 @@ models = [
         max_seq_len=2048,
         batch_size=8,
         meta_template=_meta_template,
-        model_kwargs=dict(trust_remote_code=True, device_map='auto', revision='0ba4d6fc479bdedd6a3f8d4d3425025c5f501800'),
+        model_kwargs=dict(trust_remote_code=True, device_map='auto'),
         run_cfg=dict(num_gpus=1, num_procs=1),
     )
 ]

configs/models/hf_vicuna_13b.py → configs/models/vicuna/hf_vicuna_13b_v13.py

@@ -4,7 +4,7 @@ from opencompass.models import HuggingFaceCausalLM
 models = [
     dict(
         type=HuggingFaceCausalLM,
-        abbr='vicuna-13b-hf',
+        abbr='vicuna-13b-v1.3-hf',
         path="lmsys/vicuna-13b-v1.3",
         tokenizer_path='lmsys/vicuna-13b-v1.3',
         tokenizer_kwargs=dict(

configs/models/vicuna/hf_vicuna_13b_v15.py

+from opencompass.models import HuggingFaceCausalLM
+
+
+models = [
+    dict(
+        type=HuggingFaceCausalLM,
+        abbr='vicuna-13b-v1.5-hf',
+        path="lmsys/vicuna-13b-v1.5",
+        tokenizer_path='lmsys/vicuna-13b-v1.5',
+        tokenizer_kwargs=dict(
+            padding_side='left',
+            truncation_side='left',
+            use_fast=False,
+        ),
+        max_out_len=100,
+        max_seq_len=2048,
+        batch_size=8,
+        model_kwargs=dict(device_map='auto'),
+        batch_padding=False, # if false, inference with for-loop without batch padding
+        run_cfg=dict(num_gpus=2, num_procs=1)
+    )
+]

configs/models/vicuna/hf_vicuna_13b_v15_16k.py

+from opencompass.models import HuggingFaceCausalLM
+
+
+models = [
+    dict(
+        type=HuggingFaceCausalLM,
+        abbr='vicuna-13b-v1.5-16k-hf',
+        path="lmsys/vicuna-13b-v1.5-16k",
+        tokenizer_path='lmsys/vicuna-13b-v1.5-16k',
+        tokenizer_kwargs=dict(
+            padding_side='left',
+            truncation_side='left',
+            use_fast=False,
+        ),
+        max_out_len=100,
+        max_seq_len=8192,
+        batch_size=8,
+        model_kwargs=dict(device_map='auto'),
+        batch_padding=False, # if false, inference with for-loop without batch padding
+        run_cfg=dict(num_gpus=2, num_procs=1)
+    )
+]

configs/models/hf_vicuna_33b.py → configs/models/vicuna/hf_vicuna_33b_v13.py

@@ -4,7 +4,7 @@ from opencompass.models import HuggingFaceCausalLM
 models = [
     dict(
         type=HuggingFaceCausalLM,
-        abbr='vicuna-33b-hf',
+        abbr='vicuna-33b-v1.3-hf',
         path="lmsys/vicuna-33b-v1.3",
         tokenizer_path='lmsys/vicuna-33b-v1.3',
         tokenizer_kwargs=dict(

configs/models/hf_vicuna_7b.py → configs/models/vicuna/hf_vicuna_7b_v13.py

@@ -4,7 +4,7 @@ from opencompass.models import HuggingFaceCausalLM
 models = [
     dict(
         type=HuggingFaceCausalLM,
-        abbr='vicuna-7b-hf',
+        abbr='vicuna-7b-v1.3-hf',
         path="lmsys/vicuna-7b-v1.3",
         tokenizer_path='lmsys/vicuna-7b-v1.3',
         tokenizer_kwargs=dict(

configs/models/vicuna/hf_vicuna_7b_v15.py

+from opencompass.models import HuggingFaceCausalLM
+
+
+models = [
+    dict(
+        type=HuggingFaceCausalLM,
+        abbr='vicuna-7b-v1.5-hf',
+        path="lmsys/vicuna-7b-v1.5",
+        tokenizer_path='lmsys/vicuna-7b-v1.5',
+        tokenizer_kwargs=dict(
+            padding_side='left',
+            truncation_side='left',
+            use_fast=False,
+        ),
+        max_out_len=100,
+        max_seq_len=2048,
+        batch_size=8,
+        model_kwargs=dict(device_map='auto'),
+        batch_padding=False, # if false, inference with for-loop without batch padding
+        run_cfg=dict(num_gpus=1, num_procs=1)
+    )
+]

configs/models/vicuna/hf_vicuna_7b_v15_16k.py

+from opencompass.models import HuggingFaceCausalLM
+
+
+models = [
+    dict(
+        type=HuggingFaceCausalLM,
+        abbr='vicuna-7b-v1.5-16k-hf',
+        path="lmsys/vicuna-7b-v1.5-16k",
+        tokenizer_path='lmsys/vicuna-7b-v1.5-16k',
+        tokenizer_kwargs=dict(
+            padding_side='left',
+            truncation_side='left',
+            use_fast=False,
+        ),
+        max_out_len=100,
+        max_seq_len=8192,
+        batch_size=8,
+        model_kwargs=dict(device_map='auto'),
+        batch_padding=False, # if false, inference with for-loop without batch padding
+        run_cfg=dict(num_gpus=1, num_procs=1)
+    )
+]

docs/en/get_started.md

@@ -155,8 +155,8 @@ from mmengine.config import read_base
 with read_base():
     from .datasets.siqa.siqa_gen import siqa_datasets
     from .datasets.winograd.winograd_ppl import winograd_datasets
-    from .models.hf_opt_125m import opt125m
-    from .models.hf_opt_350m import opt350m
+    from .models.opt.hf_opt_125m import opt125m
+    from .models.opt.hf_opt_350m import opt350m
 
 datasets = [*siqa_datasets, *winograd_datasets]
 models = [opt125m, opt350m]

docs/zh_cn/get_started.md

@@ -154,8 +154,8 @@ from mmengine.config import read_base
 with read_base():
     from .datasets.siqa.siqa_gen import siqa_datasets
     from .datasets.winograd.winograd_ppl import winograd_datasets
-    from .models.hf_opt_125m import opt125m
-    from .models.hf_opt_350m import opt350m
+    from .models.opt.hf_opt_125m import opt125m
+    from .models.opt.hf_opt_350m import opt350m
 
 datasets = [*siqa_datasets, *winograd_datasets]
 models = [opt125m, opt350m]