working on automatic parameter for layers

1 files changed, 23 insertions, 19 deletions

diff --git a/nerv/examples/lmptb/tnn_ptb_main.lua b/nerv/examples/lmptb/tnn_ptb_main.lua
index 059d52a..6afecbf 100644
--- a/nerv/examples/lmptb/tnn_ptb_main.lua
+++ b/nerv/examples/lmptb/tnn_ptb_main.lua
@@ -17,6 +17,9 @@ local LMTrainer = nerv.LMTrainer
 function prepare_parameters(global_conf, iter)
     printf("%s preparing parameters...\n", global_conf.sche_log_pre) 
     
+    global_conf.paramRepo = nerv.ParamRepo()
+    local paramRepo = global_conf.paramRepo
+
     if iter == -1 then --first time
         printf("%s first time, generating parameters...\n", global_conf.sche_log_pre) 
         ltp_ih = nerv.LinearTransParam("ltp_ih", global_conf)  
@@ -27,43 +30,44 @@ function prepare_parameters(global_conf, iter)
         ltp_hh.trans = global_conf.cumat_type(global_conf.hidden_size, global_conf.hidden_size)
         ltp_hh.trans:generate(global_conf.param_random) 
 
-        ltp_ho = nerv.LinearTransParam("ltp_ho", global_conf)
-        ltp_ho.trans = global_conf.cumat_type(global_conf.hidden_size, global_conf.vocab:size())
-        ltp_ho.trans:generate(global_conf.param_random)
+        --ltp_ho = nerv.LinearTransParam("ltp_ho", global_conf)
+        --ltp_ho.trans = global_conf.cumat_type(global_conf.hidden_size, global_conf.vocab:size())
+        --ltp_ho.trans:generate(global_conf.param_random)
 
         bp_h = nerv.BiasParam("bp_h", global_conf)
         bp_h.trans = global_conf.cumat_type(1, global_conf.hidden_size)
         bp_h.trans:generate(global_conf.param_random)
 
-        bp_o = nerv.BiasParam("bp_o", global_conf)
-        bp_o.trans = global_conf.cumat_type(1, global_conf.vocab:size())
-        bp_o.trans:generate(global_conf.param_random)
+        --bp_o = nerv.BiasParam("bp_o", global_conf)
+        --bp_o.trans = global_conf.cumat_type(1, global_conf.vocab:size())
+        --bp_o.trans:generate(global_conf.param_random)
 
         local f = nerv.ChunkFile(global_conf.param_fn .. '.0', 'w')
         f:write_chunk(ltp_ih)
         f:write_chunk(ltp_hh)
-        f:write_chunk(ltp_ho)
+        --f:write_chunk(ltp_ho)
         f:write_chunk(bp_h)
-        f:write_chunk(bp_o)
+        --f:write_chunk(bp_o)
         f:close()
 
         return nil
     end
     
     printf("%s loading parameter from file %s...\n", global_conf.sche_log_pre, global_conf.param_fn .. '.' .. tostring(iter)) 
-    local paramRepo = nerv.ParamRepo()
     paramRepo:import({global_conf.param_fn .. '.' .. tostring(iter)}, nil, global_conf)
 
     printf("%s preparing parameters end.\n", global_conf.sche_log_pre)
 
-    return paramRepo
+    return nil
 end
 
 --global_conf: table
 --Returns: nerv.LayerRepo
-function prepare_layers(global_conf, paramRepo)
+function prepare_layers(global_conf)
     printf("%s preparing layers...\n", global_conf.sche_log_pre)
     
+    local paramRepo = global_conf.paramRepo
+
     local du = false
 
     --local recurrentLconfig = {{["bp"] = "bp_h", ["ltp_hh"] = "ltp_hh"}, {["dim_in"] = {global_conf.hidden_size, global_conf.hidden_size}, ["dim_out"] = {global_conf.hidden_size}, ["break_id"] = global_conf.vocab:get_sen_entry().id, ["independent"] = global_conf.independent, ["clip"] = 10}}
@@ -146,10 +150,10 @@ function prepare_tnn(global_conf, layerRepo)
 end
 
 function load_net(global_conf, next_iter)
-    local paramRepo = prepare_parameters(global_conf, next_iter)
-    local layerRepo = prepare_layers(global_conf, paramRepo)
+    prepare_parameters(global_conf, next_iter)
+    local layerRepo = prepare_layers(global_conf)
     local tnn = prepare_tnn(global_conf, layerRepo)
-    return tnn, paramRepo
+    return tnn
 end
 
 local train_fn, valid_fn, test_fn
@@ -233,7 +237,7 @@ global_conf = {
 
     hidden_size = 20,
     chunk_size = 2,
-    batch_size = 3, 
+    batch_size = 10, 
     max_iter = 3,
     param_random = function() return (math.random() / 5 - 0.1) end,
 
@@ -297,7 +301,7 @@ end
 
 if start_iter == -1 or start_iter == 0 then
     print("===INITIAL VALIDATION===") 
-    local tnn, paramRepo = load_net(global_conf, 0)
+    local tnn = load_net(global_conf, 0)
     local result = LMTrainer.lm_process_file(global_conf, global_conf.valid_fn, tnn, false) --false update!
     nerv.LMUtil.wait(1)
     ppl_rec[0] = {} 
@@ -316,7 +320,7 @@ local final_iter
 for iter = start_iter, global_conf.max_iter, 1 do
     final_iter = iter --for final testing
     global_conf.sche_log_pre = "[SCHEDULER ITER"..iter.." LR"..global_conf.lrate.."]:" 
-    tnn, paramRepo = load_net(global_conf, iter - 1) 
+    tnn = load_net(global_conf, iter - 1) 
     printf("===ITERATION %d LR %f===\n", iter, global_conf.lrate) 
     result = LMTrainer.lm_process_file(global_conf, global_conf.train_fn_shuf, tnn, true) --true update!
     ppl_rec[iter] = {}
@@ -337,7 +341,7 @@ for iter = start_iter, global_conf.max_iter, 1 do
     end 
     if ppl_rec[iter].valid < ppl_last then 
         printf("%s PPL improves, saving net to file %s.%d...\n", global_conf.sche_log_pre, global_conf.param_fn, iter) 
-        paramRepo:export(global_conf.param_fn .. '.' .. tostring(iter), nil) 
+        global_conf.paramRepo:export(global_conf.param_fn .. '.' .. tostring(iter), nil) 
     else 
         printf("%s PPL did not improve, rejected, copying param file of last iter...\n", global_conf.sche_log_pre) 
         os.execute('cp ' .. global_conf.param_fn..'.'..tostring(iter - 1) .. ' ' .. global_conf.param_fn..'.'..tostring(iter))
@@ -358,6 +362,6 @@ end
 printf("\n") 
 printf("===FINAL TEST===\n") 
 global_conf.sche_log_pre = "[SCHEDULER FINAL_TEST]:" 
-tnn, paramRepo = load_net(global_conf, final_iter) 
+tnn = load_net(global_conf, final_iter) 
 LMTrainer.lm_process_file(global_conf, global_conf.test_fn, tnn, false) --false update!