Upload folder using huggingface_hub

config.json

@@ -1,13 +1,9 @@
 {
   "_commit_hash": null,
-  "_name_or_path": "OpenGVLab/Mini-InternVL-Chat-V1-5",
+  "_name_or_path": "OpenGVLab/Mini-InternVL-Chat-2B-V1-5",
   "architectures": [
     "InternVLChatModel"
   ],
-  "auto_map": {
-    "AutoConfig": "configuration_internvl_chat.InternVLChatConfig",
-    "AutoModel": "modeling_internvl_chat.InternVLChatModel"
-  },
   "downsample_ratio": 0.5,
   "dynamic_image_size": true,
   "force_image_size": 448,
@@ -77,7 +73,10 @@
     "return_dict": true,
     "return_dict_in_generate": false,
     "rms_norm_eps": 1e-05,
-    "rope_scaling": null,
+    "rope_scaling": {
+      "factor": 3.0,
+      "type": "dynamic"
+    },
     "rope_theta": 1000000,
     "sep_token_id": null,
     "suppress_tokens": null,
@@ -91,7 +90,7 @@
     "top_p": 1.0,
     "torch_dtype": "bfloat16",
     "torchscript": false,
-    "transformers_version": "4.37.2",
+    "transformers_version": "4.36.2",
     "typical_p": 1.0,
     "use_bfloat16": false,
     "use_cache": false,
@@ -188,7 +187,7 @@
     "top_p": 1.0,
     "torch_dtype": "float32",
     "torchscript": false,
-    "transformers_version": "4.37.2",
+    "transformers_version": "4.36.2",
     "typical_p": 1.0,
     "use_bfloat16": false,
     "use_flash_attn": true

generation_config.json

@@ -1,4 +1,4 @@
 {
   "_from_model_config": true,
-  "transformers_version": "4.37.2"
+  "transformers_version": "4.36.2"
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:508428116bf0dd1e782f9738ed8dd1742e91bd61ab46eac47c5a77c8c2154f1f
+oid sha256:4d5f814d2759a5de0e628ef0003c45a68fa4b6183ff905cc905c3d7ca3831805
 size 4411571040

modeling_internlm2.py

@@ -49,6 +49,21 @@ _CONFIG_FOR_DOC = 'InternLM2Config'
 flash_attn_func, flash_attn_varlen_func = None, None
 pad_input, index_first_axis, unpad_input = None, None, None
 
+try:
+    from flash_attn import flash_attn_func as _flash_attn_func
+    from flash_attn import \
+        flash_attn_varlen_func as _flash_attn_varlen_func
+    from flash_attn.bert_padding import \
+        index_first_axis as _index_first_axis
+    from flash_attn.bert_padding import pad_input as _pad_input
+    from flash_attn.bert_padding import unpad_input as _unpad_input
+    
+    flash_attn_func, flash_attn_varlen_func = _flash_attn_func, _flash_attn_varlen_func
+    pad_input, index_first_axis, unpad_input = _pad_input, _index_first_axis, _unpad_input
+    has_flash_attn = True
+except:
+    has_flash_attn = False
+
 
 def _import_flash_attn():
     global flash_attn_func, flash_attn_varlen_func
@@ -855,7 +870,7 @@ class InternLM2Model(InternLM2PreTrainedModel):
 
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
-        if self.config.attn_implementation == 'flash_attention_2':
+        if self.config.attn_implementation == 'flash_attention_2' and has_flash_attn:
             _import_flash_attn()
 
         # retrieve input_ids and inputs_embeds

modeling_internvl_chat.py

@@ -246,7 +246,7 @@ class InternVLChatModel(PreTrainedModel):
     
         queries = []
         image_bs = pixel_values.shape[0]
-        print(f'dynamic ViT batch size: {image_bs}, image_counts: {image_counts}')
+        # print(f'dynamic ViT batch size: {image_bs}, image_counts: {image_counts}')
         for idx, image_count in enumerate(image_counts):
             image_token = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * image_count + IMG_END_TOKEN
             question = image_token + '\n' + questions[idx]

tokenizer_config.json

@@ -171,8 +171,8 @@
   "bos_token": "<s>",
   "chat_template": "{{ bos_token }}{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}",
   "clean_up_tokenization_spaces": false,
-  "eos_token": "</s>",
-  "model_max_length": 5120,
+  "eos_token": "<|im_end|>",
+  "model_max_length": 8192,
   "pad_token": "</s>",
   "tokenizer_class": "InternLM2Tokenizer",
   "unk_token": "<unk>"