[BUG]Still not working, there is an error (TypeError: unsupported opera type (s) for//: 'NoneType' and 'int') when running Python playground. py #35
Comments
|
this is my pip list: accelerate 0.24.1 |
|
this is my Weight files, The following figure：
|
|
this is my /root/MiniGPT-5/config/minigpt4.yaml: datasets: run: optimizerlr_sched: "linear_warmup_cosine_lr" weight_decay: 0.05 seed: 42 amp: True evaluate: False device: "cuda" |
|
this is my /root/MiniGPT-5/minigpt4/configs/models/minigpt4.yaml: model: vit encoderimage_size: 224 Q-Formernum_query_token: 32 Vicunallama_model: "/root/vicuna-7b-v1.1" generation configsprompt: "" preprocess: |
|
this is my /root/vicuna-7b-v1.1, The following figure： |
|
Including "Vicuna-7b-v1.1" is all good. The path and configuration file are fine, but why is there still such an error. My pip "torch=2.0.1, lighting=2.0.9.post0" |
|
I'm trying to help. According to your error record: File "/root/MiniGPT-5/minigpt4/models/eva_vit.py", line 190, in init
num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0])
TypeError: unsupported operand type(s) for //: 'NoneType' and 'int'where But according to your config file, you have set |
|
May I ask which file has line 67？/root/MiniGPT-5/model.py 67 line? |
|
Yes |
Please wait a moment, I need to set up a debugging environment |
|
File "/root/MiniGPT-5/model.py", line 68, in init Seems unable to: print('minigpt4_config.model_cfg.image_size' + str(minigpt4_config.model_cfg.image_size))
|
|
Then, can you check the |
|
File "/root/MiniGPT-5/minigpt4/models/eva_vit.py", line 190, in init Because these errors were executed first, an error was reported before reaching "minipt4/common/config. py". Here are the modifications I made, but the program has not yet been executed here。 |
|
We are currently creating the environment and will use 'pycharm' to debug once it is ready。Then I will take a screenshot to show you the situation |
|
This is weird. According to your error: Traceback (most recent call last):
File "/root/MiniGPT-5/examples/playground.py", line 40, in
minigpt5 = MiniGPT5_Model.load_from_checkpoint(stage1_ckpt, strict=False, map_location="cpu", encoder_model_config=model_args, **vars(training_args))
File "/root/anaconda3/envs/minigpt5/lib/python3.9/site-packages/lightning/pytorch/core/module.py", line 1552, in load_from_checkpoint
loaded = _load_from_checkpoint(
File "/root/anaconda3/envs/minigpt5/lib/python3.9/site-packages/lightning/pytorch/core/saving.py", line 89, in _load_from_checkpoint
model = _load_state(cls, checkpoint, strict=strict, kwargs)
File "/root/anaconda3/envs/minigpt5/lib/python3.9/site-packages/lightning/pytorch/core/saving.py", line 156, in _load_state
obj = cls(_cls_kwargs)
File "/root/MiniGPT-5/model.py", line 68, in init
self.model = MiniGPT5.from_config(minigpt4_config.model_cfg)
File "/root/MiniGPT-5/minigpt4/models/mini_gpt4.py", line 247, in from_config
model = cls(
File "/root/MiniGPT-5/minigpt4/models/mini_gpt5.py", line 46, in init
super().init(*args, **kwargs)
File "/root/MiniGPT-5/minigpt4/models/mini_gpt4.py", line 53, in init
self.visual_encoder, self.ln_vision = self.init_vision_encoder(
File "/root/MiniGPT-5/minigpt4/models/blip2.py", line 65, in init_vision_encoder
visual_encoder = create_eva_vit_g(
File "/root/MiniGPT-5/minigpt4/models/eva_vit.py", line 416, in create_eva_vit_g
model = VisionTransformer(
File "/root/MiniGPT-5/minigpt4/models/eva_vit.py", line 259, in init
self.patch_embed = PatchEmbed(
File "/root/MiniGPT-5/minigpt4/models/eva_vit.py", line 190, in init
num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0])
TypeError: unsupported operand type(s) for //: 'NoneType' and 'int'Your error starts from |
yes,Start the next call with "self.model = MiniGPT5.from_config(minigpt4_config.model_cfg)". Before this, there were no errors reported。 About errors：“(img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0])” |
|
I found that the values of ‘img_size[1]’ and ‘img_size[0]’ are none, and I think this is the reason for the error |
|
Because img_size[1]=None and img_size[0]=None, an error of "TypeError: unsupported operand type(s) for //: 'NoneType' and 'int'" occurred |
Same like I said here. |
|
Now it can be DEBUG img_size in minigpt4_config.datasets_cfg.cc_sbu_align.vis_processor.train.img_size。 is not minigpt4_config.model_cfg img_size==224. is right. NO ERROR
|
|
But it should be also inside both
|
Is there a problem with my 'minigpt4.yaml' file or did I read the wrong file? Could you please take a look at the address and content of the 'minigpt4.yaml' file in my reply above |
I didn't see the wrong config in your file. To check whether you read the correct file. You should check the |
|
An error occurred before running to 'self.tokenizer = self.model.llama_tokenizer'.
|
|
I made a breakpoint here, but I didn't come in。
|
|
I think that's maybe the reason. You have multiple |
|
restart。 /root/minigpt555/minigpt4/common/config.py 27 line. minigpt4.yaml is OK. /root/minigpt555/minigpt4/common/config.py 71 line. minigpt4.yaml is OK. /root/minigpt555/model.py 67 line. minigpt4.yaml is OK. They all point to the path: /root/MiniGPT-5/config/minigpt4.yaml /root/minigpt555 and /root/MiniGPT-5 is same files |
|
Please check the function default_config_path in Line 79 of |
**Thank you, the problem has been resolved. But a new error has occurred:** Traceback (most recent call last): During handling of the above exception, another exception occurred: Traceback (most recent call last): During handling of the above exception, another exception occurred: Traceback (most recent call last): During handling of the above exception, another exception occurred: Traceback (most recent call last): The above exception was the direct cause of the following exception: Traceback (most recent call last): During handling of the above exception, another exception occurred: Traceback (most recent call last): **I wonder if I want to download 'https://huggingface.co/julien-c/EsperBERTo-small/resolve/main/pytorch_model.bin'. The network environment here is not good, so I want to download 'pytorch_model.bin' first and put it in a local folder. But I don't know which folder is better to put it in, please let me know, thank you. If it's not this file, please tell me the other file names so that I can download it and place it locally.** |
|
I have now cloned 'stablityai/table diffusion 2-1-base', should I also put it in the/root/MiniGPT-5 directory? Where exactly is it placed? |
You can place it anywhere you want. Just change |
Now the models are all installed. Result error: CUDA out of memory. File "/root/anaconda3/envs/minigpt555/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl I remember the 'python3 playground.py --stage1_weight WEIGHT_FOLDER/stage1_cc3m.ckpt ' command, it doesn't take up much memory. My server has a single card with 24GB of graphics memory. 2 graphics cards
|
|
Regarding the error: torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 128.00 MiB (GPU 0; 21.99 GiB total capacity; 21.42 GiB already allocated; 107.00 MiB free; 21.57 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF I once tried to reduce the video memory fragment to 32MB, but it sfailed, and it reported the same error. I estimate we need you to find another way to help solve it: export PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:32 |
Hello, there is an error (TypeError: unsupported operator type (s) for//: 'NoneType' and 'int') running Python playground.py
Operating system: ubuntu 20.04
Python 3.9.18
Other parameters: Same as MiniGPT-5/requirements. txt
All three ckpt files are located in MiniGPT-5/config. The configuration files have all been changed. The weight used is Vicuna-7b-v1.1. However, the following error still occurred.
Run "Python playground. py --stage1_weight /root/MiniGPT-5/config/stage1_cc3m.ckpt --test_weight /root/MiniGPT-5/config/stage2_vist.ckpt" The following error occurred during command execution:
Seed set to 42
Loading VIT
Traceback (most recent call last):
File "/root/MiniGPT-5/examples/playground.py", line 40, in
minigpt5 = MiniGPT5_Model.load_from_checkpoint(stage1_ckpt, strict=False, map_location="cpu", encoder_model_config=model_args, **vars(training_args))
File "/root/anaconda3/envs/minigpt5/lib/python3.9/site-packages/lightning/pytorch/core/module.py", line 1552, in load_from_checkpoint
loaded = _load_from_checkpoint(
File "/root/anaconda3/envs/minigpt5/lib/python3.9/site-packages/lightning/pytorch/core/saving.py", line 89, in _load_from_checkpoint
model = _load_state(cls, checkpoint, strict=strict, kwargs)
File "/root/anaconda3/envs/minigpt5/lib/python3.9/site-packages/lightning/pytorch/core/saving.py", line 156, in _load_state
obj = cls(_cls_kwargs)
File "/root/MiniGPT-5/model.py", line 68, in init
self.model = MiniGPT5.from_config(minigpt4_config.model_cfg)
File "/root/MiniGPT-5/minigpt4/models/mini_gpt4.py", line 247, in from_config
model = cls(
File "/root/MiniGPT-5/minigpt4/models/mini_gpt5.py", line 46, in init
super().init(*args, **kwargs)
File "/root/MiniGPT-5/minigpt4/models/mini_gpt4.py", line 53, in init
self.visual_encoder, self.ln_vision = self.init_vision_encoder(
File "/root/MiniGPT-5/minigpt4/models/blip2.py", line 65, in init_vision_encoder
visual_encoder = create_eva_vit_g(
File "/root/MiniGPT-5/minigpt4/models/eva_vit.py", line 416, in create_eva_vit_g
model = VisionTransformer(
File "/root/MiniGPT-5/minigpt4/models/eva_vit.py", line 259, in init
self.patch_embed = PatchEmbed(
File "/root/MiniGPT-5/minigpt4/models/eva_vit.py", line 190, in init
num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0])
TypeError: unsupported operand type(s) for //: 'NoneType' and 'int'
The text was updated successfully, but these errors were encountered: