qqqq

minigpt4/configs/models/minigpt_v2.yaml

@@ -11,7 +11,7 @@ model:
   # generation configs
   prompt: ""
 
-  llama_model: /home/czb/project/MiniGPT-4-main/checkpoints/Llama-2-7b-chat-hf
+  llama_model: "ZebangCheng/Emotion-LLaMA"
   # llama_model: "/home/user/project/Emotion-LLaMA/checkpoints/Llama-2-7b-chat-hf"
   lora_r: 64
   lora_alpha: 16

minigpt4/conversation/conversation.py

@@ -12,6 +12,7 @@ import torch
 from transformers import AutoTokenizer, AutoModelForCausalLM, LlamaTokenizer
 from transformers import StoppingCriteria, StoppingCriteriaList, TextIteratorStreamer
 from transformers import Wav2Vec2FeatureExtractor
+from transformers import AutoProcessor, AutoModel
 
 import dataclasses
 from enum import auto, Enum
@@ -263,11 +264,13 @@ class Chat:
             # model_file = "checkpoints/transformer/chinese-hubert-large"
             model_file = "ZebangCheng/chinese-hubert-large"
             feature_extractor = Wav2Vec2FeatureExtractor.from_pretrained(model_file)
+
             input_values = feature_extractor(samples, sampling_rate=sr, return_tensors="pt").input_values
             # print("input_values:", input_values)
 
             from transformers import HubertModel
-            hubert_model = HubertModel.from_pretrained(model_file)
+            # hubert_model = HubertModel.from_pretrained(model_file)
+            hubert_model = AutoModel.from_pretrained("ZebangCheng/chinese-hubert-large")
             hubert_model.eval()
             with torch.no_grad():
                 hidden_states = hubert_model(input_values, output_hidden_states=True).hidden_states # tuple of (B, T, D)

minigpt4/models/base_model.py

@@ -13,7 +13,9 @@ from omegaconf import OmegaConf
 import numpy as np
 import torch
 import torch.nn as nn
-from transformers import LlamaTokenizer
+# from transformers import LlamaTokenizer
+from transformers import AutoTokenizer
+
 from peft import (
     LoraConfig,
     get_peft_model,
@@ -23,7 +25,8 @@ from peft import (
 from minigpt4.common.dist_utils import download_cached_file
 from minigpt4.common.utils import get_abs_path, is_url
 from minigpt4.models.eva_vit import create_eva_vit_g
-from minigpt4.models.modeling_llama import LlamaForCausalLM
+# from minigpt4.models.modeling_llama import LlamaForCausalLM
+from transformers.models.llama.modeling_llama import LlamaForCausalLM
 
 
 
@@ -172,7 +175,9 @@ class BaseModel(nn.Module):
     def init_llm(cls, llama_model_path, low_resource=False, low_res_device=0, lora_r=0,
                  lora_target_modules=["q_proj","k_proj"], **lora_kargs):
         logging.info('Loading LLAMA')
-        llama_tokenizer = LlamaTokenizer.from_pretrained(llama_model_path, use_fast=False)
+        llama_model_path
+        # llama_tokenizer = LlamaTokenizer.from_pretrained(llama_model_path, use_fast=False)
+        llama_tokenizer = AutoTokenizer.from_pretrained(llama_model_path)
         llama_tokenizer.pad_token = "$$"
 
         if low_resource: