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.

Sign up for GitHub

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

Jump to bottom

[BUG] Model compression fails for some descriptor neuron structures #1370

Closed
ssrokyz opened this issue Dec 22, 2021 · 4 comments · Fixed by #2933
Closed

[BUG] Model compression fails for some descriptor neuron structures #1370

ssrokyz opened this issue Dec 22, 2021 · 4 comments · Fixed by #2933
Labels
enhancement

Comments

@ssrokyz
Copy link

ssrokyz commented Dec 22, 2021 •
edited
Loading

Summary

When I compress the model with some neuron structures of the descriptor, it fails.
When I use the neuron of [25, 50, 100], it succeeds.
But when I use [50, 50, 50] or [100, 50, 25], it fails.

Deepmd-kit version, installation way, input file, running commands, error log, etc.
DP-kit version: v2.0.2
installation: source install
running commands: dp compress -i frozen_model.pb -t input.json
frozen_model.pb files will be attached.
frozen_model.zip

input.json:
`

{
"_comment": " model parameters",
"model": {
"type_map": ["Ge", "Sb", "Te"],
"descriptor" :{
"type": "se_e2_a",
"sel": [30, 30, 30],
"rcut_smth": 0.10,
"rcut": 6.00,
"neuron": [100, 50, 25],
"resnet_dt": false,
"axis_neuron": 16,
"activation_function": "tanh",
"seed": 1,
"_comment": " that's all"
},
"fitting_net" : {
"neuron": [200, 200, 200],
"activation_function": "tanh",
"resnet_dt": true,
"seed": 1,
"_comment": " that's all"
},
"_comment": " that's all"
},

"learning_rate" :{
"type":		"exp",
"start_lr":	1e-3,	
"stop_lr":	1e-3,
"_comment":	"that's all"
},

"loss" :{
"type":		"ener",
"start_pref_e": 1e-2,
"limit_pref_e": 1e-2,
"start_pref_f": 1e-0,
"limit_pref_f": 1e-0,
"start_pref_v": 1e-2,
"limit_pref_v": 1e-2,
"_comment":	" that's all"
},

"training" : {
    "training_data": {
        "systems": [
            "../raw-sffld_gst-hex-gather.traj.d/train",
            "../raw-sffld_gst-cubic-gather.traj.d/train",
            "../raw-sffld_gs-cubic-liquid-gather.traj.d",
            "../raw-sffld_sb-hex-liquid-8th.traj.d",
            "../raw-sffld_ge-cubic-liquid-gather.traj.d",
            "../raw-sffld_te-cubic-liquid-gather.traj.d"
        ],
        "batch_size":	2,
        "_comment":		"that's all"
    },
    "validation_data":{
        "systems": [
            "../raw-sffld_gst-hex-gather.traj.d/test",
            "../raw-sffld_gst-cubic-gather.traj.d/test"
        ],
        "batch_size":	50,
        "numb_btch":	1,
        "_comment":		"that's all"
    },
    "numb_steps":	2000000,
    "seed":		10,
    "disp_file":	"lcurve.out",
    "disp_freq":	1000,
    "save_freq":	1000,
    "save_ckpt": "bu/model.ckpt",
    "_comment":	"that's all"
    },    

"_comment":		"that's all"

}
`

Error log:
`
Traceback (most recent call last):
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/framework/ops.py", line 1853, in _create_c_op
c_op = pywrap_tf_session.TF_FinishOperation(op_desc)
tensorflow.python.framework.errors_impl.InvalidArgumentError: Dimensions must be equal, but are 50 and 200 for '{{node add_4/add}} = AddV2[T=DT_DOUBLE](Tanh_1, concat)' with input shapes: [1501,50], [1501,200].

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/ops/math_ops.py", line 1175, in binary_op_wrapper
out = r_op(x)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/ops/math_ops.py", line 1195, in r_binary_op_wrapper
return func(x, y, name=name)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/util/dispatch.py", line 201, in wrapper
return target(*args, **kwargs)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/ops/math_ops.py", line 1486, in _add_dispatch
return gen_math_ops.add_v2(x, y, name=name)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/ops/gen_math_ops.py", line 481, in add_v2
_, _, _op, _outputs = _op_def_library._apply_op_helper(
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/framework/op_def_library.py", line 748, in _apply_op_helper
op = g._create_op_internal(op_type_name, inputs, dtypes=None,
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/framework/ops.py", line 3528, in _create_op_internal
ret = Operation(
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/framework/ops.py", line 2015, in init
self._c_op = _create_c_op(self._graph, node_def, inputs,
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/framework/ops.py", line 1856, in _create_c_op
raise ValueError(str(e))
ValueError: Dimensions must be equal, but are 50 and 200 for '{{node add_4/add}} = AddV2[T=DT_DOUBLE](Tanh_1, concat)' with input shapes: [1501,50], [1501,200].

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/home/grad/ssrokyz/.conda/envs/tf24/bin/dp", line 33, in
sys.exit(load_entry_point('deepmd-kit==2.0.2', 'console_scripts', 'dp')())
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/deepmd_kit-2.0.2-py3.8-linux-x86_64.egg/deepmd/entrypoints/main.py", line 447, in main
compress(**dict_args)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/deepmd_kit-2.0.2-py3.8-linux-x86_64.egg/deepmd/entrypoints/compress.py", line 119, in compress
train(
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/deepmd_kit-2.0.2-py3.8-linux-x86_64.egg/deepmd/entrypoints/train.py", line 103, in train
_do_work(jdata, run_opt, is_compress)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/deepmd_kit-2.0.2-py3.8-linux-x86_64.egg/deepmd/entrypoints/train.py", line 159, in _do_work
model.build(train_data, stop_batch)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/deepmd_kit-2.0.2-py3.8-linux-x86_64.egg/deepmd/train/trainer.py", line 285, in build
self.descrpt.enable_compression(self.model_param['compress']["min_nbor_dist"], self.model_param['compress']['model_file'], self.model_param['compress']['table_config'][0], self.model_param['compress']['table_config'][1], self.model_param['compress']['table_config'][2], self.model_param['compress']['table_config'][3])
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/deepmd_kit-2.0.2-py3.8-linux-x86_64.egg/deepmd/descriptor/se_a.py", line 335, in enable_compression
= self.table.build(min_nbor_dist,
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/deepmd_kit-2.0.2-py3.8-linux-x86_64.egg/deepmd/utils/tabulate.py", line 148, in build
vv, dd, d2 = self._make_data(xx, ii)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/deepmd_kit-2.0.2-py3.8-linux-x86_64.egg/deepmd/utils/tabulate.py", line 236, in _make_data
tt, zz = self._layer_1(
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/deepmd_kit-2.0.2-py3.8-linux-x86_64.egg/deepmd/utils/tabulate.py", line 255, in _layer_1
return t, self.activation_fn(tf.matmul(x, w) + b) + t
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/ops/math_ops.py", line 1180, in binary_op_wrapper
raise e
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/ops/math_ops.py", line 1164, in binary_op_wrapper
return func(x, y, name=name)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/util/dispatch.py", line 201, in wrapper
return target(*args, **kwargs)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/ops/math_ops.py", line 1486, in _add_dispatch
return gen_math_ops.add_v2(x, y, name=name)
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/ops/gen_math_ops.py", line 481, in add_v2
_, _, _op, _outputs = _op_def_library._apply_op_helper(
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/framework/op_def_library.py", line 748, in _apply_op_helper
op = g._create_op_internal(op_type_name, inputs, dtypes=None,
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/framework/ops.py", line 3528, in _create_op_internal
ret = Operation(
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/framework/ops.py", line 2015, in init
self._c_op = _create_c_op(self._graph, node_def, inputs,
File "/home/grad/ssrokyz/.conda/envs/tf24/lib/python3.8/site-packages/tensorflow/python/framework/ops.py", line 1856, in _create_c_op
raise ValueError(str(e))
ValueError: Dimensions must be equal, but are 50 and 200 for '{{node add_4}} = AddV2[T=DT_DOUBLE](Tanh_1, concat)' with input shapes: [1501,50], [1501,200].
`
@ssrokyz ssrokyz added the bug label Dec 22, 2021
@njzjz
Copy link
Member

njzjz commented Dec 23, 2021

Currently only [25, 50, 100] is supported.

@wanghan-iapcm
Copy link
Collaborator

We only support the case neuron[i+1] == neuron[i] * 2, neuron[i] can be any number

@ssrokyz
Copy link
Author

ssrokyz commented Dec 24, 2021 •
edited
Loading

Thanks for the immediate and clear answer!

@ssrokyz ssrokyz closed this as completed Dec 24, 2021
@njzjz njzjz added duplicate and removed bug labels Jan 2, 2022
@njzjz njzjz reopened this Jan 16, 2022
@njzjz
Copy link
Member

njzjz commented Jan 16, 2022

Feature request in #1422.

njzjz added a commit to njzjz/deepmd-kit that referenced this issue Oct 18, 2023
@njzjz
support compressing any neuron structure
d9a68c0 
Fix deepmodeling#1370.

Signed-off-by: Jinzhe Zeng <[email protected]>
@njzjz njzjz mentioned this issue Oct 18, 2023
Merged
@njzjz njzjz linked a pull request Oct 18, 2023 that will close this issue
support compressing any neuron structure #2933
Merged
wanghan-iapcm pushed a commit that referenced this issue Oct 19, 2023
@njzjz
support compressing any neuron structure (#2933)
299184c 
Fix #1370.

Support compressing any neuron structure; fix compression error when
`neuron[0]` is 1 (although no one will set `neuron[0]` to 1).

Signed-off-by: Jinzhe Zeng <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
enhancement
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

support compressing any neuron structure njzjz/deepmd-kit
3 participants
@njzjz @ssrokyz @wanghan-iapcm