Disable support for autobatching for TFLITE

Author: lantiga

src/redisai.c

@@ -667,13 +667,19 @@ int RedisAI_ModelSet_RedisCommand(RedisModuleCtx *ctx, RedisModuleString **argv,
 
   unsigned long long batchsize = 0;
   if (AC_AdvanceIfMatch(&ac, "BATCHSIZE")) {
+    if (backend == RAI_BACKEND_TFLITE) {
+      return RedisModule_ReplyWithError(ctx, "Auto-batching not supported by the TFLITE backend.");
+    }
     if (AC_GetUnsignedLongLong(&ac, &batchsize, 0) != AC_OK) {
       return RedisModule_ReplyWithError(ctx, "Invalid argument for BATCHSIZE.");
     }
   }
 
   unsigned long long minbatchsize = 0;
   if (AC_AdvanceIfMatch(&ac, "MINBATCHSIZE")) {
+    if (batchsize == 0) {
+      return RedisModule_ReplyWithError(ctx, "MINBATCHSIZE specified without BATCHSIZE.");
+    }
     if (AC_GetUnsignedLongLong(&ac, &minbatchsize, 0) != AC_OK) {
       return RedisModule_ReplyWithError(ctx, "Invalid argument for MINBATCHSIZE");
     }

test/basic_tests.py

@@ -801,47 +801,52 @@ def test_run_tflite_model(env):
 # TODO: Autobatch is tricky with TFLITE because TFLITE expects a fixed batch
 #       size. At least we should constrain MINBATCHSIZE according to the
 #       hard-coded dims in the tflite model.
-# def test_run_tflite_model_autobatch(env):
-#     if not TEST_PT:
-#         return
-# 
-#     con = env.getConnection()
-# 
-#     test_data_path = os.path.join(os.path.dirname(__file__), 'test_data')
-#     model_filename = os.path.join(test_data_path, 'mnist_model_quant.tflite')
-#     sample_filename = os.path.join(test_data_path, 'one.raw')
-# 
-#     with open(model_filename, 'rb') as f:
-#         model_pb = f.read()
-# 
-#     with open(sample_filename, 'rb') as f:
-#         sample_raw = f.read()
-# 
-#     ret = con.execute_command('AI.MODELSET', 'm', 'TFLITE', 'CPU',
-#                               'BATCHSIZE', 2, 'MINBATCHSIZE', 2, model_pb)
-#     env.assertEqual(ret, b'OK')
-# 
-#     con.execute_command('AI.TENSORSET', 'a', 'FLOAT', 1, 1, 28, 28, 'BLOB', sample_raw)
-#     con.execute_command('AI.TENSORSET', 'c', 'FLOAT', 1, 1, 28, 28, 'BLOB', sample_raw)
-# 
-#     def run():
-#         con = env.getConnection()
-#         con.execute_command('AI.MODELRUN', 'm', 'INPUTS', 'c', 'OUTPUTS', 'd', 'd2')
-# 
-#     t = threading.Thread(target=run)
-#     t.start()
-# 
-#     con.execute_command('AI.MODELRUN', 'm', 'INPUTS', 'a', 'OUTPUTS', 'b', 'b2')
-# 
-#     tensor = con.execute_command('AI.TENSORGET', 'b', 'VALUES')
-#     value = tensor[-1][0]
-# 
-#     env.assertEqual(value, 1)
-# 
-#     tensor = con.execute_command('AI.TENSORGET', 'd', 'VALUES')
-#     value = tensor[-1][0]
-# 
-#     env.assertEqual(value, 1)
+def test_run_tflite_model_autobatch(env):
+    if not TEST_PT:
+        return
+
+    con = env.getConnection()
+
+    test_data_path = os.path.join(os.path.dirname(__file__), 'test_data')
+    model_filename = os.path.join(test_data_path, 'mnist_model_quant.tflite')
+    sample_filename = os.path.join(test_data_path, 'one.raw')
+
+    with open(model_filename, 'rb') as f:
+        model_pb = f.read()
+
+    with open(sample_filename, 'rb') as f:
+        sample_raw = f.read()
+
+    try:
+        ret = con.execute_command('AI.MODELSET', 'm', 'TFLITE', 'CPU',
+                                  'BATCHSIZE', 2, 'MINBATCHSIZE', 2, model_pb)
+    except Exception as e:
+        exception = e
+    env.assertEqual(type(exception), redis.exceptions.ResponseError)
+
+    # env.assertEqual(ret, b'OK')
+
+    # con.execute_command('AI.TENSORSET', 'a', 'FLOAT', 1, 1, 28, 28, 'BLOB', sample_raw)
+    # con.execute_command('AI.TENSORSET', 'c', 'FLOAT', 1, 1, 28, 28, 'BLOB', sample_raw)
+
+    # def run():
+    #     con = env.getConnection()
+    #     con.execute_command('AI.MODELRUN', 'm', 'INPUTS', 'c', 'OUTPUTS', 'd', 'd2')
+
+    # t = threading.Thread(target=run)
+    # t.start()
+
+    # con.execute_command('AI.MODELRUN', 'm', 'INPUTS', 'a', 'OUTPUTS', 'b', 'b2')
+
+    # tensor = con.execute_command('AI.TENSORGET', 'b', 'VALUES')
+    # value = tensor[-1][0]
+
+    # env.assertEqual(value, 1)
+
+    # tensor = con.execute_command('AI.TENSORGET', 'd', 'VALUES')
+    # value = tensor[-1][0]
+
+    # env.assertEqual(value, 1)
 
 
 def test_set_tensor_multiproc(env):