Truncate the hash to first 6 hexdigits. 6 hexdigits is equivalent to …

…6*4=24 bits. The number of unique hashes that we can generate before the probability of hash collision between any 2 model hashes reaches 50% is sqrt(2*0.5*2^(24))=4096. We are unlikely to run training over hyperparameter space that large on a single iteration.
eic · Jan 22, 2024 · 2cbb87e · 2cbb87e
1 parent 08fa4c5
commit 2cbb87e
Showing 1 changed file with 3 additions and 2 deletions.
diff --git a/benchmarks/roman_pots/Snakefile b/benchmarks/roman_pots/Snakefile
@@ -31,8 +31,9 @@ N_LAYERS_PX = [10]
 SIZE_FIRST_HIDDEN_LAYER_PX = [10]
 MULTIPLIER_PX = [10]
 LEAK_RATE_PX = [0.025]
+MAX_HASH = 6
 MODEL_VERSION = [
-    hashlib.sha512("_".join(map(str,x)).encode()).hexdigest()
+    hashlib.sha512("_".join(map(str,x)).encode()).hexdigest()[:MAX_HASH]
     for x in product(
         NUM_TRAINING_INPUTS,
         NUM_EPOCHS_PZ, LEARNING_RATE_PZ, SIZE_INPUT_PZ, SIZE_OUTPUT_PZ, N_LAYERS_PZ, SIZE_FIRST_HIDDEN_LAYER_PZ, MULTIPLIER_PZ, LEAK_RATE_PZ,
@@ -70,7 +71,7 @@ rule roman_pots_generate_neural_network_configs:
                     str(num_epochs_pz)+"_"+str(learning_rate_pz)+"_"+str(size_input_pz)+"_"+str(size_output_pz)+"_"+str(n_layers_pz)+"_"+str(size_first_hidden_layer_pz)+"_"+str(multiplier_pz)+"_"+str(leak_rate_pz)+"_"+\
                     str(num_epochs_py)+"_"+str(learning_rate_py)+"_"+str(size_input_py)+"_"+str(size_output_py)+"_"+str(n_layers_py)+"_"+str(size_first_hidden_layer_py)+"_"+str(multiplier_py)+"_"+str(leak_rate_py)+"_"+\
                     str(num_epochs_px)+"_"+str(learning_rate_px)+"_"+str(size_input_px)+"_"+str(size_output_px)+"_"+str(n_layers_px)+"_"+str(size_first_hidden_layer_px)+"_"+str(multiplier_px)+"_"+str(leak_rate_px)
-      output_file_location = open(str(output_dir)+"/"+str(detector_version)+"_"+str(detector_config)+"_"+str(SUBSYSTEM)+"_"+str(BENCHMARK)+"_"+str(hashlib.sha512(output_file.encode()).hexdigest())+".txt","w")
+      output_file_location = open(str(output_dir)+"/"+str(detector_version)+"_"+str(detector_config)+"_"+str(SUBSYSTEM)+"_"+str(BENCHMARK)+"_"+str(hashlib.sha512(output_file.encode()).hexdigest()[:MAX_HASH])+".txt","w")
       output_file_location.write(output_file)
       print(output_file_location)
       output_file_location.close()