Upload 15 files

config.json

@@ -1,6 +1,6 @@
 {
   "_commit_hash": null,
-  "_name_or_path": "internvl2-tiny",
+  "_name_or_path": "/home/ea/work/my_optimum_intel/optimum-intel/internvl2-tiny",
   "architectures": [
     "InternVLChatModel"
   ],
@@ -14,6 +14,7 @@
   "force_image_size": 28,
   "img_context_token_id": 151648,
   "llm_config": {
+    "_attn_implementation_autoset": true,
     "_name_or_path": "Qwen/Qwen2-0.5B-Instruct",
     "add_cross_attention": false,
     "architectures": [
@@ -89,7 +90,7 @@
     "top_p": 1.0,
     "torch_dtype": "bfloat16",
     "torchscript": false,
-    "transformers_version": "4.45.2",
+    "transformers_version": "4.46.2",
     "typical_p": 1.0,
     "use_bfloat16": true,
     "use_cache": true,
@@ -108,6 +109,7 @@
   "use_llm_lora": 0,
   "use_thumbnail": true,
   "vision_config": {
+    "_attn_implementation_autoset": true,
     "_name_or_path": "",
     "add_cross_attention": false,
     "architectures": [
@@ -186,7 +188,7 @@
     "top_p": 1.0,
     "torch_dtype": "bfloat16",
     "torchscript": false,
-    "transformers_version": "4.45.2",
+    "transformers_version": "4.46.2",
     "typical_p": 1.0,
     "use_bfloat16": true,
     "use_flash_attn": false

generation_config.json

@@ -1,4 +1,4 @@
 {
   "_from_model_config": true,
-  "transformers_version": "4.45.2"
+  "transformers_version": "4.46.2"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1c91dc9c94bbae4e6d8d16aab28765983e8b12cd0565b9e8a98f80073b0b5e14
-size 34321392
+oid sha256:e9f0331124d4efa1050990abf1755529dc4c161abcd9efa40dab5df83495c290
+size 68292216

modeling_internvl_chat.py

@@ -20,7 +20,7 @@ from .conversation import get_conv_template
 from .modeling_intern_vit import InternVisionModel, has_flash_attn
 
 
-logger = logging.get_logger(__name__)
+#logger = logging.get_logger(__name__)
 
 
 def version_cmp(v1, v2, op="eq"):
@@ -39,7 +39,7 @@ class InternVLChatModel(PreTrainedModel):
     _supports_flash_attn_2 = True
     _no_split_modules = ["InternVisionModel", "LlamaDecoderLayer", "Qwen2DecoderLayer"]
 
-    def __init__(self, config: InternVLChatConfig, vision_model=None, language_model=None, use_flash_attn=True):
+    def __init__(self, config: InternVLChatConfig, vision_model=None, language_model=None, use_flash_attn=False):
         super().__init__(config)
 
         assert version_cmp(transformers.__version__, "4.37.0", "ge")
@@ -55,8 +55,8 @@ class InternVLChatModel(PreTrainedModel):
         config.vision_config.use_flash_attn = True if use_flash_attn else False
         config.llm_config._attn_implementation = "flash_attention_2" if use_flash_attn else "eager"
 
-        logger.info(f"num_image_token: {self.num_image_token}")
-        logger.info(f"ps_version: {self.ps_version}")
+        #logger.info(f"num_image_token: {self.num_image_token}")
+        #logger.info(f"ps_version: {self.ps_version}")
         if vision_model is not None:
             self.vision_model = vision_model
         else:
@@ -103,7 +103,7 @@ class InternVLChatModel(PreTrainedModel):
         input_embeds = self.language_model.get_input_embeddings()(input_ids).clone()
 
         vit_embeds = self.extract_feature(pixel_values)
-        vit_batch_size = pixel_values.shape[0]
+        pixel_values.shape[0]
 
         B, N, C = input_embeds.shape
         input_embeds = input_embeds.reshape(B * N, C)
@@ -289,9 +289,6 @@ class InternVLChatModel(PreTrainedModel):
         template.append_message(template.roles[1], None)
         query = template.get_prompt()
 
-        print(self.num_image_token)
-        print(num_patches_list)
-
         if verbose and pixel_values is not None:
             image_bs = pixel_values.shape[0]
             print(f"dynamic ViT batch size: {image_bs}")
@@ -300,7 +297,7 @@ class InternVLChatModel(PreTrainedModel):
             image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * num_patches + IMG_END_TOKEN
             query = query.replace("<image>", image_tokens, 1)
 
-        print(query)
+ 
         model_inputs = tokenizer(query, return_tensors="pt")
         input_ids = model_inputs["input_ids"].to(self.device)
         attention_mask = model_inputs["attention_mask"].to(self.device)
@@ -356,6 +353,7 @@ class InternVLChatModel(PreTrainedModel):
             attention_mask=attention_mask,
             generation_config=generation_config,
             output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
             use_cache=True,
             **generate_kwargs,
         )