chore: remove prints and breakpoints

zama-ai · Jun 4, 2024 · 364dc5a · 364dc5a
1 parent 7291684
commit 364dc5a
Show file tree

Hide file tree

Showing 3 changed files with 37 additions and 33 deletions.
diff --git a/README.md b/README.md
@@ -254,18 +254,22 @@ This software is distributed under the **BSD-3-Clause-Clear** license. Read [thi
 #### FAQ
 
 **Is Zama’s technology free to use?**
->Zama’s libraries are free to use under the BSD 3-Clause Clear license only for development, research, prototyping, and experimentation purposes. However, for any commercial use of Zama's open source code, companies must purchase Zama’s commercial patent license.
+
+> Zama’s libraries are free to use under the BSD 3-Clause Clear license only for development, research, prototyping, and experimentation purposes. However, for any commercial use of Zama's open source code, companies must purchase Zama’s commercial patent license.
 >
->All our work is open source and we strive for full transparency about Zama's IP strategy. To know more about what this means for Zama product users, read about how we monetize our open source products in [this blog post](https://www.zama.ai/post/open-source).
+> All our work is open source and we strive for full transparency about Zama's IP strategy. To know more about what this means for Zama product users, read about how we monetize our open source products in [this blog post](https://www.zama.ai/post/open-source).
 
 **What do I need to do if I want to use Zama’s technology for commercial purposes?**
->To commercially use Zama’s technology you need to be granted Zama’s patent license. Please contact us at hello@zama.ai for more information.
+
+> To commercially use Zama’s technology you need to be granted Zama’s patent license. Please contact us at hello@zama.ai for more information.
 
 **Do you file IP on your technology?**
->Yes, all of Zama’s technologies are patented.
+
+> Yes, all of Zama’s technologies are patented.
 
 **Can you customize a solution for my specific use case?**
->We are open to collaborating and advancing the FHE space with our partners. If you have specific needs, please email us at hello@zama.ai.
+
+> We are open to collaborating and advancing the FHE space with our partners. If you have specific needs, please email us at hello@zama.ai.
 
 <p align="right">
   <a href="#about" > ↑ Back to top </a> 

diff --git a/src/concrete/ml/common/preprocessors.py b/src/concrete/ml/common/preprocessors.py
@@ -170,9 +170,9 @@ def vectorized_graph_eval(
 
         res = node.evaluator(*pred_results)
         node_results[node] = res
-        print(
-            f"{node.properties['name'] if 'name' in node.properties else node}, {node.operation=}({[pred.shape for pred in pred_results]})={res.shape=},"
-        )
+        # print(
+        #     f"{node.properties['name'] if 'name' in node.properties else node}, {node.operation=}({[pred.shape for pred in pred_results]})={res.shape=},"
+        # )
 
     result = tuple(node_results[node] for node in graph.ordered_outputs())
     assert len(result) > 0, "Empty results"
@@ -773,7 +773,7 @@ def bias_closed_form(
     # hackish -> breaks the target-bit-width assumption
     if bit_width(input_range - bias) != bit_width(input_range):
         msbs_to_keep += 1
-        print(f"debug: adding 1 to {msbs_to_keep=}")
+        # print(f"debug: adding 1 to {msbs_to_keep=}")
 
     return bias, msbs_to_keep
 
@@ -912,7 +912,7 @@ def decompose_1_bit_tlu(
 
             # todo: get the proper value here: must be the result of f(x-1) - f(x) or smth like that
             res = tlu_coefs[threshold_index + 1] - tlu_coefs[threshold_index]
-            print(res)
+            # print(res)
             coefficients[best_indexes + (threshold_index,)] = res
 
             acc_size = bit_width(scale_up(input_range, scaling_factor=1, bias=offset))
@@ -1078,7 +1078,7 @@ def delta_optimize(
 
         # todo: re-activate this assert as this shouldn't happen
         if acc_size > target_bit_width:
-            print(f"{acc_size=} > {target_bit_width=}")
+            # print(f"{acc_size=} > {target_bit_width=}")
             triggered = True
 
         # todo: remove this after check that the accuracy is fine
@@ -1475,15 +1475,15 @@ def apply(self, graph: Graph):
 
             # Only one input should be non-constant per LUT
             if len(variable_input_indices) != 1:
-                print("SKIPPING BECAUSE MORE THAN 1 INPUT")
+                # print("SKIPPING BECAUSE MORE THAN 1 INPUT")
                 continue
 
             if pred_nodes[0].properties["name"] in {"round_bit_pattern", "truncate_bit_pattern"}:
-                print("SKIPPING BECAUSE ROUNDING ALREADY PRESENT")
+                # print("SKIPPING BECAUSE ROUNDING ALREADY PRESENT")
                 continue
 
             if tlu_node.properties["name"] in {"round_bit_pattern", "truncate_bit_pattern"}:
-                print("SKIPPING BECAUSE NODE IS ROUNDING NODE")
+                # print("SKIPPING BECAUSE NODE IS ROUNDING NODE")
                 continue
 
             # Get variable input
@@ -1498,13 +1498,13 @@ def apply(self, graph: Graph):
             if variable_input_bit_width <= self.rounding_threshold:
                 # No need to do anything if the bit-width is actually lower or equal
                 # to the rounding threshold value
-                print("SKIPPING BECAUSE ROUNDING ABOVE ACCUMULATOR SIZE")
+                # print("SKIPPING BECAUSE ROUNDING ABOVE ACCUMULATOR SIZE")
                 continue
 
             # Compute lsbs to remove
             lsbs_to_remove = variable_input_bit_width - self.rounding_threshold
 
-            print(f"ADDING ROUNDING NODE to {tlu_node.properties['name']}")
+            # print(f"ADDING ROUNDING NODE to {tlu_node.properties['name']}")
             # Rounding node
             rounding_node = Node.generic(
                 "round_bit_pattern",
@@ -1556,9 +1556,9 @@ def update_constant_node(node, new_value):
 def add_extra_shape_to_subgraph(tlu_node, extra_dim: Tuple[int, ...]):
     # TODO: maybe collect all shapes and just broadcast everything to the correct shape
     # and reshape if constant
-    print()
-    print()
-    print("UPDATING SUBGRAPH SHAPES")
+    # print()
+    # print()
+    # print("UPDATING SUBGRAPH SHAPES")
     tlu_node.inputs[0] = deepcopy(tlu_node.inputs[0])
     tlu_node.inputs[0].shape = tlu_node.inputs[0].shape + extra_dim
 
@@ -1576,17 +1576,17 @@ def add_extra_shape_to_subgraph(tlu_node, extra_dim: Tuple[int, ...]):
         node.inputs[0].shape = deepcopy(node.inputs[0].shape)
         node.inputs[0].shape = node.inputs[0].shape + extra_dim
 
-        print(f"{node.output.shape=}")
+        # print(f"{node.output.shape=}")
         node.output.shape = deepcopy(node.output.shape)
 
         if node.output.shape[-1:] == extra_dim:  # BIG HACK
             continue
 
         node.output.shape = node.output.shape + extra_dim
-        print(f"{node.output.shape=}")
-        print(f"{node}")
-    print()
-    print()
+    #     print(f"{node.output.shape=}")
+    #     print(f"{node}")
+    # print()
+    # print()
 
 
 class Debug(GraphProcessor):
@@ -1597,7 +1597,6 @@ def __init__(
 
     def apply(self, graph: Graph) -> None:
         print(graph.format())
-        breakpoint()
 
 
 # todo: fix insert rounding to make sure that it only adds rounding if there is
@@ -1698,7 +1697,7 @@ def apply(self, graph: Graph) -> None:
             print(f"{number_of_tlus=}, {lsbs_to_remove=} {coefficients.shape=}, {offsets.shape=}")
             if number_of_tlus == -1:
                 # Don't do here for now
-                print("SKIPPING NODE")
+                # print("SKIPPING NODE")
                 continue
 
             # ####################### This is the part where we modify the graph ###############
@@ -1746,7 +1745,7 @@ def apply(self, graph: Graph) -> None:
             # >= 0
             def compute_lut(args, **kwargs):
                 res = (args >= 0).astype(np.int64)
-                print(f"{args=}, {res=}, {np.unique(args)=}")
+                # print(f"{args=}, {res=}, {np.unique(args)=}")
                 return res
 
                 # This works but this isn't what we want
@@ -1797,10 +1796,10 @@ def compute_lut(args, **kwargs):
                 current_node, base.astype(np.int64), add, graph=graph.graph
             )
 
-            print(f"Done modifying TLU: {tlu_node}")
-        print("DONE")
+        #     print(f"Done modifying TLU: {tlu_node}")
+        # print("DONE")
         # recursion error when printing :thinking-face:
-        print(graph.format())
+        # print(graph.format())
 
     def modify_subgraph_for_rounded_inputs(
         self,
@@ -1911,14 +1910,16 @@ def modify_graph_to_round_subgraph_inputs(
         if np.any(a_int != 1):
             previous_node = add_leveled_op_with_cst(previous_node, a_int, multiply, graph.graph)
         else:
-            print("not inserting multiplicative factor")
+            pass
+            # print("not inserting multiplicative factor")
 
         # Subtract offset that matches step threshold to rounding step treshold
         b_int = b.astype(np.int64)
         if np.any(b_int != 0):
             previous_node = add_leveled_op_with_cst(previous_node, b_int, subtract, graph.graph)
         else:
-            print("not inserting bias")
+            pass
+            # print("not inserting bias")
 
         # # Compute the offset needed to cancel out the rounding offset
         # # Then broadcast it to match the rounded tensor shape

diff --git a/src/concrete/ml/pytest/utils.py b/src/concrete/ml/pytest/utils.py
@@ -19,7 +19,6 @@
 from torch.utils.data import DataLoader, TensorDataset
 
 from concrete.ml.quantization.qat_quantizers import Int8ActPerTensorPoT, Int8WeightPerTensorPoT
-
 from concrete.ml.sklearn.linear_model import SGDClassifier
 
 from ..common.serialization.dumpers import dump, dumps