KlayMa527
commented
10 months ago I solve this problem by this link https://github.com/Dao-AILab/flash-attention/issues/742
Closed KlayMa527 closed 10 months ago
KlayMa527
commented
10 months ago I solve this problem by this link https://github.com/Dao-AILab/flash-attention/issues/742
Traceback (most recent call last): File "vtimellm/train/train_mem.py", line 20, in
train()
File "/root/autodl-tmp/VTimeLLM/vtimellm/train/train.py", line 353, in train
trainer.train()
File "/root/miniconda3/lib/python3.8/site-packages/transformers/trainer.py", line 1537, in train
return inner_training_loop(
File "/root/miniconda3/lib/python3.8/site-packages/transformers/trainer.py", line 1854, in _inner_training_loop
tr_loss_step = self.training_step(model, inputs)
File "/root/miniconda3/lib/python3.8/site-packages/transformers/trainer.py", line 2735, in training_step
loss = self.compute_loss(model, inputs)
File "/root/miniconda3/lib/python3.8/site-packages/transformers/trainer.py", line 2758, in compute_loss
outputs = model(inputs)
File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
return forward_call(*args, *kwargs)
File "/root/miniconda3/lib/python3.8/site-packages/deepspeed/utils/nvtx.py", line 15, in wrapped_fn
ret_val = func(args, kwargs)
File "/root/miniconda3/lib/python3.8/site-packages/deepspeed/runtime/engine.py", line 1833, in forward
loss = self.module(*inputs, kwargs)
File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1538, in _call_impl
result = forward_call(*args, *kwargs)
File "/root/autodl-tmp/VTimeLLM/vtimellm/train/../../vtimellm/model/vtimellm.py", line 171, in forward
outputs = self.model(
File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1538, in _call_impl
result = forward_call(args, kwargs)
File "/root/miniconda3/lib/python3.8/site-packages/transformers/models/llama/modeling_llama.py", line 1058, in forward
layer_outputs = self._gradient_checkpointing_func(
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/checkpoint.py", line 249, in checkpoint
return CheckpointFunction.apply(function, preserve, args)
File "/root/miniconda3/lib/python3.8/site-packages/torch/autograd/function.py", line 506, in apply
return super().apply(args, kwargs) # type: ignore[misc]
File "/root/miniconda3/lib/python3.8/site-packages/torch/utils/checkpoint.py", line 107, in forward
outputs = run_function(args)
File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1538, in _call_impl
result = forward_call(args, kwargs)
File "/root/miniconda3/lib/python3.8/site-packages/transformers/models/llama/modeling_llama.py", line 796, in forward
hidden_states, self_attn_weights, present_key_value = self.self_attn(
File "/root/miniconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1538, in _call_impl
result = forward_call(*args, *kwargs)
File "/root/autodl-tmp/VTimeLLM/vtimellm/train/llama_flash_attn_monkey_patch.py", line 92, in forward
output_unpad = flash_attn_unpadded_qkvpacked_func(
File "/root/miniconda3/lib/python3.8/site-packages/flash_attn/flash_attn_interface.py", line 887, in flash_attn_varlen_qkvpacked_func
return FlashAttnVarlenQKVPackedFunc.apply(
File "/root/miniconda3/lib/python3.8/site-packages/torch/autograd/function.py", line 506, in apply
return super().apply(args, **kwargs) # type: ignore[misc]
File "/root/miniconda3/lib/python3.8/site-packages/flash_attn/flash_attn_interface.py", line 288, in forward
out, q, k, v, out_padded, softmax_lse, S_dmask, rng_state = _flash_attn_varlen_forward(
File "/root/miniconda3/lib/python3.8/site-packages/flash_attn/flash_attn_interface.py", line 85, in _flash_attn_varlen_forward
out, q, k, v, out_padded, softmax_lse, S_dmask, rng_state = flash_attn_cuda.varlen_fwd(
RuntimeError: cu_seqlens_q must have shape (batch_size + 1)
I followed the instructions to set data and feat, but encountered the following error during stage 1 training. I would appreciate it if you could help me answer this questions.