Merge pull request NVIDIA#246 from swethmandava/master

nvpstr · web-flow · commit 4c385522d1d8 · 2019-10-09T14:10:56.000+02:00
BERT TF syntactical fixes
diff --git a/TensorFlow/LanguageModeling/BERT/README.md b/TensorFlow/LanguageModeling/BERT/README.md
diff --git a/TensorFlow/LanguageModeling/BERT/configurations.yml b/TensorFlow/LanguageModeling/BERT/configurations.yml
@@ -108,18 +108,6 @@ bert--DGX2_n32:
     NUM_ACCUMULATION_STEPS: "8"
     PHASE: "1"
 
-#64 DGX2 phase1
-bert--DGX2_n64:
-  <<: *BERT_ON_CLUSTER
-  <<: *DGX2
-  variables:
-    <<: *DGX2_VARS
-    NNODES: "32"
-    BATCHSIZE: "32"
-    LEARNING_RATE: "1.171875e-5"
-    NUM_ACCUMULATION_STEPS: "4"
-    PHASE: "1"
-
 #1 DGX1 phase2
 bert--DGX1_n1p2:
   <<: *BERT_ON_CLUSTER
diff --git a/TensorFlow/LanguageModeling/BERT/data/bertPrep.py b/TensorFlow/LanguageModeling/BERT/data/bertPrep.py
@@ -288,14 +288,14 @@ def create_record_worker(filename_prefix, shard_id, output_format='hdf5'):
         '--n_training_shards',
         type=int,
         help='Specify the number of training shards to generate',
-        default=256
+        default=1472
     )
 
     parser.add_argument(
         '--n_test_shards',
         type=int,
         help='Specify the number of test shards to generate',
-        default=256
+        default=1472
     )
 
     parser.add_argument(
diff --git a/TensorFlow/LanguageModeling/BERT/notebooks/bert_squad_tf_finetuning.ipynb b/TensorFlow/LanguageModeling/BERT/notebooks/bert_squad_tf_finetuning.ipynb
@@ -26,7 +26,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "<img src=http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png style=\"width: 90px; float: right;\">\n",
+    "<img src=\"http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png\" style=\"width: 90px; float: right;\">\n",
     "\n",
     "# BERT Question Answering Fine-Tuning with Mixed Precision"
    ]
diff --git a/TensorFlow/LanguageModeling/BERT/notebooks/bert_squad_tf_inference.ipynb b/TensorFlow/LanguageModeling/BERT/notebooks/bert_squad_tf_inference.ipynb
@@ -26,7 +26,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "<img src=http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png style=\"width: 90px; float: right;\">\n",
+    "<img src=\"http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png\" style=\"width: 90px; float: right;\">\n",
     "\n",
     "# BERT Question Answering Inference with Mixed Precision\n"
    ]
diff --git a/TensorFlow/LanguageModeling/BERT/notebooks/bert_squad_tf_inference_colab.ipynb b/TensorFlow/LanguageModeling/BERT/notebooks/bert_squad_tf_inference_colab.ipynb
diff --git a/TensorFlow/LanguageModeling/BERT/run_pretraining.py b/TensorFlow/LanguageModeling/BERT/run_pretraining.py
@@ -469,7 +469,7 @@ def input_fn():
             lambda record: _decode_record(record, name_to_features),
             batch_size=batch_size,
             num_parallel_batches=num_cpu_threads,
-            drop_remainder=True))
+            drop_remainder=True if is_training else False))
     return d
 
   return input_fn

Original file line number	Diff line number	Diff line change
`@@ -26,7 +26,7 @@`
`26`	`26`	`"cell_type": "markdown",`
`27`	`27`	`"metadata": {},`
`28`	`28`	`"source": [`
`29`		`- "<img src=http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png style=\"width: 90px; float: right;\">\n",`
	`29`	`+ "<img src=\"http://developer.download.nvidia.com/compute/machine-learning/frameworks/nvidia_logo.png\" style=\"width: 90px; float: right;\">\n",`
`30`	`30`	`"\n",`
`31`	`31`	`"# BERT Question Answering Fine-Tuning with Mixed Precision"`
`32`	`32`	`]`