Merge pull request #77 from deepmodeling/devel

merge from devel

.github/workflows/build_wheel.yml

@@ -25,7 +25,7 @@ jobs:
 
       - name: Build wheels
         env:
-          CIBW_BUILD: "cp36-* cp37-* cp38-* cp39-*"
+          CIBW_BUILD: "cp36-* cp37-* cp38-* cp39-* cp310-*"
           CIBW_MANYLINUX_X86_64_IMAGE: ghcr.io/deepmodeling/manylinux2010_x86_64_tensorflow
           CIBW_BEFORE_BUILD: pip install tensorflow
           CIBW_SKIP: "*-win32 *-manylinux_i686 *-musllinux*"

.github/workflows/test_python.yml

@@ -42,10 +42,10 @@ jobs:
           - python: 3.7
             gcc: 8
             tf:
-          - python: 3.8
+          - python: "3.10"
             gcc: 5
             tf:
-          - python: 3.8
+          - python: "3.10"
             gcc: 8
             tf:
 

.gitignore

@@ -29,6 +29,8 @@ _templates
 API_CC
 doc/api_py/
 dp/
+dp_test/
+dp_test_cc/
 build_lammps/
 .idea/
 build_tests/

.readthedocs.yml

@@ -1,4 +1,8 @@
 version: 2
+build:
+  os: ubuntu-20.04
+  tools:
+    python: mambaforge-4.10
 conda:
   environment: doc/environment.yml
 formats: all

README.md

@@ -97,6 +97,7 @@ A full [document](doc/train/train-input-auto.rst) on options in the training inp
     - [Fit energy](doc/model/train-energy.md)
     - [Fit `tensor` like `Dipole` and `Polarizability`](doc/model/train-fitting-tensor.md)
     - [Train a Deep Potential model using `type embedding` approach](doc/model/train-se-e2-a-tebd.md)
+    - [Deep potential long-range](doc/model/dplr.md)
 - [Training](doc/train/index.md)
     - [Training a model](doc/train/training.md)
     - [Advanced options](doc/train/training-advanced.md)

deepmd/common.py

@@ -538,11 +538,6 @@ def cast_precision(func: Callable) -> Callable:
     If it does not match (e.g. it is an integer), the decorator
     will do nothing on it.
 
-    Parameters
-    ----------
-    precision : tf.DType
-        Tensor data type that casts to
-
     Returns
     -------
     Callable
@@ -560,6 +555,7 @@ def cast_precision(func: Callable) -> Callable:
     ...   def f(x: tf.Tensor, y: tf.Tensor) -> tf.Tensor:
     ...     return x ** 2 + y
     """
+    @wraps(func)
     def wrapper(self, *args, **kwargs):
         # only convert tensors
         returned_tensor = func(

deepmd/entrypoints/main.py

@@ -361,7 +361,7 @@ def parse_args(args: Optional[List[str]] = None):
         "--system",
         default=".",
         type=str,
-        help="The system directory, not support recursive detection.",
+        help="The system directory. Recursively detect systems in this directory.",
     )
     parser_model_devi.add_argument(
         "-S", "--set-prefix", default="set", type=str, help="The set prefix"

deepmd/fit/ener.py

@@ -332,7 +332,7 @@ def _build_lower(
     def build (self, 
                inputs : tf.Tensor,
                natoms : tf.Tensor,
-               input_dict : dict = {},
+               input_dict : dict = None,
                reuse : bool = None,
                suffix : str = '', 
     ) -> tf.Tensor:
@@ -362,6 +362,8 @@ def build (self,
         ener
                 The system energy
         """
+        if input_dict is None:
+            input_dict = {}
         bias_atom_e = self.bias_atom_e
         if self.numb_fparam > 0 and ( self.fparam_avg is None or self.fparam_inv_std is None ):
             raise RuntimeError('No data stat result. one should do data statisitic, before build')
@@ -401,7 +403,12 @@ def build (self,
         inputs = tf.reshape(inputs, [-1, self.dim_descrpt * natoms[0]])
         if len(self.atom_ener):
             # only for atom_ener
-            inputs_zero = tf.zeros_like(inputs, dtype=self.fitting_precision)
+            nframes = input_dict.get('nframes')
+            if nframes is not None:
+                # like inputs, but we don't want to add a dependency on inputs
+                inputs_zero = tf.zeros((nframes, self.dim_descrpt * natoms[0]), dtype=self.fitting_precision)
+            else:
+                inputs_zero = tf.zeros_like(inputs, dtype=self.fitting_precision)
 
 
         if bias_atom_e is not None :
@@ -419,10 +426,7 @@ def build (self,
             aparam = (aparam - t_aparam_avg) * t_aparam_istd
             aparam = tf.reshape(aparam, [-1, self.numb_aparam * natoms[0]])
 
-        if input_dict is not None:
-            type_embedding = input_dict.get('type_embedding', None)
-        else:
-            type_embedding = None
+        type_embedding = input_dict.get('type_embedding', None)
         if type_embedding is not None:
             atype_embed = embed_atom_type(self.ntypes, natoms, type_embedding)
             atype_embed = tf.tile(atype_embed,[tf.shape(inputs)[0],1])

deepmd/infer/model_devi.py

@@ -2,6 +2,7 @@
 from .deep_pot import DeepPot
 from ..utils.data import DeepmdData
 from ..utils.batch_size import AutoBatchSize
+from deepmd.common import expand_sys_str
 
 
 def calc_model_devi_f(fs: np.ndarray):
@@ -56,11 +57,12 @@ def write_model_devi_out(devi: np.ndarray, fname: str):
     header = "%10s" % "step"
     for item in 'vf':
         header += "%19s%19s%19s" % (f"max_devi_{item}", f"min_devi_{item}", f"avg_devi_{item}")
-    np.savetxt(fname,
-               devi,
-               fmt=['%12d'] + ['%19.6e' for _ in range(6)],
-               delimiter='',
-               header=header)
+    with open(fname, "ab") as fp:
+        np.savetxt(fp,
+                   devi,
+                   fmt=['%12d'] + ['%19.6e' for _ in range(6)],
+                   delimiter='',
+                   header=header)
     return devi
 
 def _check_tmaps(tmaps, ref_tmap=None):
@@ -185,25 +187,31 @@ def make_model_devi(
         tmap = tmaps[0]
     else:
         raise RuntimeError("The models does not have the same type map.")
-
-    # create data-system
-    dp_data = DeepmdData(system, set_prefix, shuffle_test=False, type_map=tmap)
-    if dp_data.pbc:
-        nopbc = False
-    else:
-        nopbc = True
 
-    data_sets = [dp_data._load_set(set_name) for set_name in dp_data.dirs]
-    nframes_tot = 0
-    devis = []
-    for data in data_sets:
-        coord = data["coord"]
-        box = data["box"]
-        atype = data["type"][0] 
-        devi = calc_model_devi(coord, box, atype, dp_models, nopbc=nopbc)
-        nframes_tot += coord.shape[0]
-        devis.append(devi)
-    devis = np.vstack(devis)
-    devis[:, 0] = np.arange(nframes_tot) * frequency
-    write_model_devi_out(devis, output)
-    return devis
+    all_sys = expand_sys_str(system)
+    if len(all_sys) == 0:
+        raise RuntimeError("Did not find valid system")
+    devis_coll = []
+    for system in all_sys:
+        # create data-system
+        dp_data = DeepmdData(system, set_prefix, shuffle_test=False, type_map=tmap)
+        if dp_data.pbc:
+            nopbc = False
+        else:
+            nopbc = True
+
+        data_sets = [dp_data._load_set(set_name) for set_name in dp_data.dirs]
+        nframes_tot = 0
+        devis = []
+        for data in data_sets:
+            coord = data["coord"]
+            box = data["box"]
+            atype = data["type"][0] 
+            devi = calc_model_devi(coord, box, atype, dp_models, nopbc=nopbc)
+            nframes_tot += coord.shape[0]
+            devis.append(devi)
+        devis = np.vstack(devis)
+        devis[:, 0] = np.arange(nframes_tot) * frequency
+        write_model_devi_out(devis, output)
+        devis_coll.append(devis)
+    return devis_coll

deepmd/loss/ener.py

@@ -129,7 +129,7 @@ def build (self,
         return l2_loss, more_loss
 
     def eval(self, sess, feed_dict, natoms):
-        placeholder = tf.no_op()
+        placeholder = self.l2_l
         run_data = [
             self.l2_l,
             self.l2_more['l2_ener_loss'] if self.has_e else placeholder,

deepmd/model/ener.py

@@ -117,7 +117,9 @@ def build (self,
                frz_model = None,
                suffix = '', 
                reuse = None):
-
+
+        if input_dict is None:
+            input_dict = {}
         with tf.variable_scope('model_attr' + suffix, reuse = reuse) :
             t_tmap = tf.constant(' '.join(self.type_map), 
                                  name = 'tmap', 
@@ -144,6 +146,7 @@ def build (self,
 
         coord = tf.reshape (coord_, [-1, natoms[1] * 3])
         atype = tf.reshape (atype_, [-1, natoms[1]])
+        input_dict['nframes'] = tf.shape(coord)[0]
 
         # type embedding if any
         if self.typeebd is not None:
@@ -270,4 +273,4 @@ def build (self,
 
     def _import_graph_def_from_frz_model(self, frz_model, feed_dict, return_elements):
         graph, graph_def = load_graph_def(frz_model)
-        return tf.import_graph_def(graph_def, input_map = feed_dict, return_elements = return_elements, name = "")
+        return tf.import_graph_def(graph_def, input_map = feed_dict, return_elements = return_elements, name = "")

deepmd/train/trainer.py

@@ -282,7 +282,12 @@ def build (self,
                 ))
             self.type_map = data.get_type_map()
             self.batch_size = data.get_batch_size()
-            self.model.data_stat(data)
+            if self.run_opt.init_mode not in ('init_from_model', 'restart'):
+                # self.saver.restore (in self._init_session) will restore avg and std variables, so data_stat is useless
+                # currently init_from_frz_model does not restore data_stat variables
+                # TODO: restore avg and std in the init_from_frz_model mode
+                log.info("data stating... (this step may take long time)")
+                self.model.data_stat(data)
 
             # config the init_frz_model command
             if self.run_opt.init_mode == 'init_from_frz_model':

deepmd/utils/network.py

@@ -56,7 +56,7 @@ def one_layer(inputs,
             w = tf.cast(w, get_precision(mixed_prec['compute_prec']))
             b = tf.cast(b, get_precision(mixed_prec['compute_prec']))
 
-        hidden = tf.matmul(inputs, w) + b
+        hidden = tf.nn.bias_add(tf.matmul(inputs, w), b)
         if activation_fn != None and use_timestep :
             idt_initializer = tf.random_normal_initializer(
                                     stddev=0.001,
@@ -196,7 +196,7 @@ def embedding_net(xx,
         variable_summaries(w, 'matrix_'+str(ii)+name_suffix)
 
         b = tf.get_variable('bias_'+str(ii)+name_suffix, 
-                            [1, outputs_size[ii]], 
+                            [outputs_size[ii]], 
                             precision,
                             b_initializer, 
                             trainable = trainable)
@@ -206,7 +206,7 @@ def embedding_net(xx,
             xx = tf.cast(xx, get_precision(mixed_prec['compute_prec']))
             w  = tf.cast(w,  get_precision(mixed_prec['compute_prec']))
             b  = tf.cast(b,  get_precision(mixed_prec['compute_prec']))
-        hidden = tf.reshape(activation_fn(tf.matmul(xx, w) + b), [-1, outputs_size[ii]])
+        hidden = tf.reshape(activation_fn(tf.nn.bias_add(tf.matmul(xx, w), b)), [-1, outputs_size[ii]])
         if resnet_dt :
             idt_initializer = tf.random_normal_initializer(
                                   stddev=0.001, 

doc/data/data-conv.md

@@ -51,5 +51,5 @@ box.raw  coord.raw  energy.raw  force.raw  set.000  set.001  set.002  type.raw
 ```
 It generates three sets `set.000`, `set.001` and `set.002`, with each set contains 2000 frames. One do not need to take care of the binary data files in each of the `set.*` directories. The path containing `set.*` and `type.raw` is called a *system*. 
 
-If one needs to train a non-periodic system, an empty `nopbc` file should be put under the system directory. `box.raw` is not necessary is a non-periodic system.
+If one needs to train a non-periodic system, an empty `nopbc` file should be put under the system directory. `box.raw` is not necessary in a non-periodic system.
 

doc/install/easy-install.md

@@ -53,3 +53,8 @@ To pull the GPU version:
 ```bash
 docker pull ghcr.io/deepmodeling/deepmd-kit:2.0.0_cuda10.1_gpu
 ```
+
+To pull the ROCm version:
+```bash
+docker pull deepmodeling/dpmdkit-rocm:dp2.0.3-rocm4.5.2-tf2.6-lmp29Sep2021
+```