search

madaan / minimal-text-diffusion

A minimal implementation of diffusion models for text generation
MIT License
310 stars 29 forks source link

Getting Error while Training #18

Open heychhavi opened 10 months ago

heychhavi commented 10 months ago

When I am trying to run this

bash scripts/run_train.sh greetings 1 False False False 1

I am getting the below error: Training: 0% 0/1 [00:00<?, ?it/s]INFO:root:Reading data from data/greetings-train.txt INFO:root:Tokenizing 298 sentences Traceback (most recent call last): File "/content/gdrive/MyDrive/minimal-text-diffusion/src/train_infer/train.py", line 117, in main() File "/content/gdrive/MyDrive/minimal-text-diffusion/src/train_infer/train.py", line 106, in main ).run_loop() ^^^^^^^^^^ File "/content/gdrive/MyDrive/minimal-text-diffusion/src/train_infer/train_loop.py", line 186, in run_loop self.run_step(batch, cond) File "/content/gdrive/MyDrive/minimal-text-diffusion/src/train_infer/train_loop.py", line 206, in run_step self.forward_backward(batch, cond) File "/content/gdrive/MyDrive/minimal-text-diffusion/src/train_infer/train_loop.py", line 268, in forward_backward losses = compute_losses() ^^^^^^^^^^^^^^^^ File "/content/gdrive/MyDrive/minimal-text-diffusion/src/modeling/diffusion/respace.py", line 99, in training_losses return super().training_losses(self._wrap_model(model), *args, **kwargs) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/content/gdrive/MyDrive/minimal-text-diffusion/src/modeling/diffusion/gaussian_diffusion.py", line 246, in training_losses x_start_mean = model.model.module.get_embeds(input_ids) ^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1695, in getattr raise AttributeError(f"'{type(self).name}' object has no attribute '{name}'") AttributeError: 'TransformerNetModel' object has no attribute 'module'

AI7Xiao commented 7 months ago

I meet the same problem. Do you resolve this?

Henabo commented 5 months ago

I meet the same problem

Henabo commented 5 months ago

I solved the problem, because you are running the model with the cpu, so you need to remove the "module" to solve the problem in the gaussian_diffusion.py. such as "x_start_mean = model.model.get_embeds(input_ids)". You won't have this problem if you're running on a gpu.