fixed bugs in guided_choices and utils

gaochangw · May 13, 2019 · 9c40dbe · 9c40dbe
1 parent 3bff7fd
commit 9c40dbe
Show file tree

Hide file tree

Showing 5 changed files with 89 additions and 97 deletions.
diff --git a/.idea/workspace.xml b/.idea/workspace.xml
diff --git a/cfg/TIMIT_CGS/TIMIT_LSTM_fmllr_ghcgs.cfg b/cfg/TIMIT_CGS/TIMIT_LSTM_fmllr_ghcgs.cfg
@@ -5,13 +5,13 @@ cfg_proto_chunk = proto/global_chunk.proto
 [exp]
 cmd =
 run_nn_script = run_nn
-out_folder = exp/TIMIT_LSTM_fmllr_test2_ghcgs_onlyFinal_25d32b_75d4b
+out_folder = exp/TIMIT_LSTM_fmllr_test2_gcgs_onlyFirst_81p25d32b
 seed = 22341
 use_cuda = True
 multi_gpu = False
 save_gpumem = False
 n_epochs_tr = 8
-apply_guided_ep = 8
+apply_guided_ep = 2
 
 [dataset1]
 data_name = TIMIT_tr
@@ -160,10 +160,10 @@ out_folder =
 lstm_hcgs = False
 guided_hcgs = True
 apply_guided_hcgs = False
-hcgsx_block = 32,4
-hcgsx_sparse = 25,75
-hcgsh_block = 32,4
-hcgsh_sparse = 25,75
+hcgsx_block = 32
+hcgsx_sparse = 81.25
+hcgsh_block = 32
+hcgsh_sparse = 81.25
 lstm_quant = False
 param_quant = 6,6
 lstm_quant_inp = False
@@ -200,8 +200,8 @@ out_folder =
 mlp_hcgs = False
 guided_hcgs = True
 apply_guided_hcgs = False
-hcgs_block = 32,4
-hcgs_sparse = 25,75
+hcgs_block = 32
+hcgs_sparse = 81.25
 mlp_quant = False
 param_quant = 5
 mlp_quant_inp = False

diff --git a/cfg/TIMIT_CGS/TIMIT_LSTM_fmllr_ghcgs_L1.cfg b/cfg/TIMIT_CGS/TIMIT_LSTM_fmllr_ghcgs_L1.cfg
@@ -5,13 +5,13 @@ cfg_proto_chunk = proto/global_chunk.proto
 [exp]
 cmd =
 run_nn_script = run_nn
-out_folder = exp/TIMIT_LSTM_fmllr_test_ghcgs_l1_onlyFinal_25d32b_75d4b
-seed = 2234
+out_folder = exp/TIMIT_LSTM_fmllr_test2_ghcgs_l1_onlyFirst_25d32b_75d4b
+seed = 22341
 use_cuda = True
 multi_gpu = False
 save_gpumem = False
 n_epochs_tr = 8
-apply_guided_ep = 8
+apply_guided_ep = 2
 
 [dataset1]
 data_name = TIMIT_tr

diff --git a/guided_choices.py b/guided_choices.py
@@ -18,7 +18,7 @@ def guided_array_rows (w_mat, n_blk, n_blk_sels, blk_size):
         w_mat2 = w_mat[: ,x:x + c1]
         temp = torch.randn(1, 1, r, c1)
         temp[0, 0, :, :] = w_mat2
-        if r == blk_size:
+        if r == blk_size and c1 >= blk_size:
             avg = torch.nn.AvgPool2d(blk_size, c1)
         else:
             avg = torch.nn.AvgPool2d((r, c1), c1)

diff --git a/utils.py b/utils.py
@@ -1898,7 +1898,7 @@ def forward_model(fea_dict, lab_dict, arch_dict, model, nns, costs, inp, inp_out
     for line in model:
         [out_name, operation, inp1, inp2] = list(re.findall(pattern, line)[0])
 
-        if out_name == 'loss_gl':
+        if out_name[0:7] == 'loss_gl':
             pattern2 = '(.*)=(.*)\((.*),(.*),(.*)\)'
             [out_name, operation, inp1, inp2, inp3] = list(re.findall(pattern2, line)[0])