add more model case into daily test

tests/test_hf_model.py

@@ -1,6 +1,8 @@
 import pytest
 import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer
+from auto_gptq.modeling import BaseGPTQForCausalLM
+from PIL import Image
+from transformers import AutoModel, AutoModelForCausalLM, AutoTokenizer
 
 prompts = ['你好', "what's your name"]
 
@@ -16,8 +18,8 @@ class TestChat:
     @pytest.mark.parametrize(
         'model_name',
         [
-            'internlm/internlm2-chat-7b',
-            'internlm/internlm2-chat-7b-sft',
+            'internlm/internlm2-chat-7b', 'internlm/internlm2-chat-7b-sft',
+            'internlm/internlm2-chat-1_8b', 'internlm/internlm2-chat-1_8b-sft'
         ],
     )
     def test_demo_default(self, model_name):
@@ -29,11 +31,13 @@ class TestChat:
             model_name, torch_dtype=torch.float16,
             trust_remote_code=True).cuda()
         model = model.eval()
+        history = []
         for prompt in prompts:
-            response, history = model.chat(tokenizer, prompt, history=[])
+            response, history = model.chat(tokenizer, prompt, history=history)
             print(response)
             assert_model(response)
 
+        history = []
         for prompt in prompts:
             length = 0
             for response, history in model.stream_chat(tokenizer,
@@ -50,8 +54,8 @@ class TestBase:
     @pytest.mark.parametrize(
         'model_name',
         [
-            'internlm/internlm2-7b',
-            'internlm/internlm2-base-7b',
+            'internlm/internlm2-7b', 'internlm/internlm2-base-7b',
+            'internlm/internlm2-1_8b'
         ],
     )
     def test_demo_default(self, model_name):
@@ -78,3 +82,149 @@ class TestBase:
                                       skip_special_tokens=True)
             print(output)
             assert_model(output)
+
+
+class TestMath:
+    """Test cases for base model."""
+
+    @pytest.mark.parametrize(
+        'model_name',
+        ['internlm/internlm2-math-7b', 'internlm/internlm2-math-base-7b'],
+    )
+    def test_demo_default(self, model_name):
+        tokenizer = AutoTokenizer.from_pretrained(model_name,
+                                                  trust_remote_code=True)
+        # Set `torch_dtype=torch.float16` to load model in float16, otherwise
+        # it will be loaded as float32 and might cause OOM Error.
+        model = AutoModelForCausalLM.from_pretrained(
+            model_name, trust_remote_code=True,
+            torch_dtype=torch.float16).cuda()
+        model = model.eval()
+        model = model.eval()
+        response, history = model.chat(tokenizer,
+                                       '1+1=',
+                                       history=[],
+                                       meta_instruction='')
+        print(response)
+        assert_model(response)
+
+
+class TestMMModel:
+    """Test cases for base model."""
+
+    @pytest.mark.parametrize(
+        'model_name',
+        [
+            'internlm/internlm-xcomposer2-7b',
+        ],
+    )
+    def test_demo_default(self, model_name):
+        tokenizer = AutoTokenizer.from_pretrained(model_name,
+                                                  trust_remote_code=True)
+        # Set `torch_dtype=torch.float16` to load model in float16, otherwise
+        # it will be loaded as float32 and might cause OOM Error.
+        if '4bit' in model_name:
+            model = InternLMXComposer2QForCausalLM.from_quantized(
+                model_name, trust_remote_code=True, device='cuda:0').eval()
+            tokenizer = AutoTokenizer.from_pretrained(model_name,
+                                                      trust_remote_code=True)
+        else:
+            model = AutoModelForCausalLM.from_pretrained(
+                model_name, torch_dtype=torch.float16,
+                trust_remote_code=True).cuda()
+            tokenizer = AutoTokenizer.from_pretrained(model_name,
+                                                      trust_remote_code=True)
+
+        model = model.eval()
+        img_path_list = [
+            'tests/panda.jpg',
+            'tests/bamboo.jpeg',
+        ]
+        images = []
+        for img_path in img_path_list:
+            image = Image.open(img_path).convert('RGB')
+            image = model.vis_processor(image)
+            images.append(image)
+        image = torch.stack(images)
+        query = '<ImageHere> <ImageHere>please write an article ' \
+            + 'based on the images. Title: my favorite animal.'
+        with torch.cuda.amp.autocast():
+            response, history = model.chat(tokenizer,
+                                           query=query,
+                                           image=image,
+                                           history=[],
+                                           do_sample=False)
+        print(response)
+        assert len(response) != 0
+        assert 'panda' in response
+
+        query = '<ImageHere> <ImageHere>请根据图片写一篇作文：我最喜欢的小动物。' \
+            + '要求：选准角度，确定立意，明确文体，自拟标题。'
+        with torch.cuda.amp.autocast():
+            response, history = model.chat(tokenizer,
+                                           query=query,
+                                           image=image,
+                                           history=[],
+                                           do_sample=False)
+        print(response)
+        assert len(response) != 0
+        assert '熊猫' in response
+
+
+class TestMMVlModel:
+    """Test cases for base model."""
+
+    @pytest.mark.parametrize(
+        'model_name',
+        [
+            'internlm/internlm-xcomposer2-vl-7b',
+            'internlm/internlm-xcomposer2-vl-7b-4bit'
+        ],
+    )
+    def test_demo_default(self, model_name):
+        tokenizer = AutoTokenizer.from_pretrained(model_name,
+                                                  trust_remote_code=True)
+
+        torch.set_grad_enabled(False)
+
+        # init model and tokenizer
+        if '4bit' in model_name:
+            model = InternLMXComposer2QForCausalLM.from_quantized(
+                model_name, trust_remote_code=True, device='cuda:0').eval()
+            tokenizer = AutoTokenizer.from_pretrained(model_name,
+                                                      trust_remote_code=True)
+        else:
+            model = AutoModel.from_pretrained(
+                model_name, trust_remote_code=True).cuda().eval()
+            tokenizer = AutoTokenizer.from_pretrained(model_name,
+                                                      trust_remote_code=True)
+
+        query = '<ImageHere>Please describe this image in detail.'
+        image = 'tests/image.webp'
+        with torch.cuda.amp.autocast():
+            response, _ = model.chat(tokenizer,
+                                     query=query,
+                                     image=image,
+                                     history=[],
+                                     do_sample=False)
+        print(response)
+        assert len(response) != 0
+        assert 'Oscar Wilde' in response
+        assert 'Live life with no excuses, travel with no regret' in response
+
+
+class InternLMXComposer2QForCausalLM(BaseGPTQForCausalLM):
+    layers_block_name = 'model.layers'
+    outside_layer_modules = [
+        'vit',
+        'vision_proj',
+        'model.tok_embeddings',
+        'model.norm',
+        'output',
+    ]
+    inside_layer_modules = [
+        ['attention.wqkv.linear'],
+        ['attention.wo.linear'],
+        ['feed_forward.w1.linear', 'feed_forward.w3.linear'],
+        ['feed_forward.w2.linear'],
+    ]