Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

runtimeerror when nbit = 4 and group_size =64 #32

Open
Eutenacity opened this issue Apr 13, 2024 · 0 comments
Open

runtimeerror when nbit = 4 and group_size =64 #32

Eutenacity opened this issue Apr 13, 2024 · 0 comments

Comments

@Eutenacity
Copy link

changes to the demo

ffn_config = BaseQuantizeConfig(
nbits=4,
group_size=64,
quant_zero=True,
quant_scale=True,
)
errors are below
Traceback (most recent call last):
File "/workspace/accelerate_files/demo.py", line 121, in
result = model.generate(
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/transformers/generation/utils.py", line 1718, in generate
return self.greedy_search(
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/transformers/generation/utils.py", line 2579, in greedy_search
outputs = self(
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/transformers/models/mixtral/modeling_mixtral.py", line 1228, in forward
outputs = self.model(
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/transformers/models/mixtral/modeling_mixtral.py", line 1093, in forward
layer_outputs = decoder_layer(
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/transformers/models/mixtral/modeling_mixtral.py", line 818, in forward
hidden_states, router_logits = self.block_sparse_moe(hidden_states,prefetch_uids,next_layer_moe)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/workspace/accelerate_files/mixtraloffloading/src/custom_layers.py", line 320, in forward
current_hidden_states = expert_layer(current_state) * routing_weights[top_x_list, idx_list, None]
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/workspace/accelerate_files/mixtraloffloading/src/expert_wrapper.py", line 33, in forward
return self.expert_module(*args, **kwargs)
File "/workspace/accelerate_files/mixtraloffloading/src/expert_wrapper.py", line 18, in
self.expert_module = lambda *args, **kwargs: expert_module(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/workspace/accelerate_files/mixtraloffloading/src/custom_layers.py", line 255, in forward
current_hidden_states = self.act_fn(self.w1(hidden_states)) * self.w3(hidden_states)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
return forward_call(*args, **kwargs)
File "/workspace/accelerate_files/mixtraloffloading/src/custom_layers.py", line 50, in forward
return self.forward_triton(x)
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/workspace/accelerate_files/mixtraloffloading/src/custom_layers.py", line 65, in forward_triton
meta['scale'] = Quantizer.dequantize(meta['scale_q'], meta['meta_scale']); del_keys.append('scale')
File "/root/anaconda3/envs/dsmii/lib/python3.10/site-packages/hqq/core/quantize.py", line 86, in dequantize
W_r = ((W_q_p - meta['zero'])*meta['scale']).reshape(meta['shape'])
RuntimeError: shape '[1, 917504]' is invalid for input of size 3670016

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant