From edd345b52e54003bb5b0baf85f4f8f4e98089365 Mon Sep 17 00:00:00 2001 From: cyyever Date: Fri, 4 Apr 2025 19:31:11 +0800 Subject: [PATCH] Fix deprecated PT functions (#37237) * Fix deprecated PT functions Signed-off-by: cyy * Revert some changes Signed-off-by: cyy --------- Signed-off-by: cyy --- tests/generation/test_utils.py | 8 ++++---- tests/models/falcon_mamba/test_modeling_falcon_mamba.py | 2 +- tests/models/mamba/test_modeling_mamba.py | 2 +- 3 files changed, 6 insertions(+), 6 deletions(-) diff --git a/tests/generation/test_utils.py b/tests/generation/test_utils.py index e726ecbd50..a01b33703e 100644 --- a/tests/generation/test_utils.py +++ b/tests/generation/test_utils.py @@ -2724,7 +2724,7 @@ class UtilsFunctionsTest(unittest.TestCase): # Case 1 input_ids = torch.randint(0, 16, (2, 8), dtype=torch.int64)[:, :0] inputs_embeds = torch.rand((2, 8), dtype=torch.float32) - cache_position = torch.range(0, 7, dtype=torch.int64) + cache_position = torch.arange(0, 8, dtype=torch.int64) eager1, eager2 = GenerationMixin()._cache_dependant_input_preparation(input_ids, inputs_embeds, cache_position) export1, export2 = GenerationMixin()._cache_dependant_input_preparation_exporting( input_ids, inputs_embeds, cache_position @@ -2735,7 +2735,7 @@ class UtilsFunctionsTest(unittest.TestCase): # Case 2 input_ids = torch.randint(0, 16, (2, 8), dtype=torch.int64) inputs_embeds = torch.rand((2, 8), dtype=torch.float32) - cache_position = torch.range(0, 7, dtype=torch.int64) + cache_position = torch.arange(0, 8, dtype=torch.int64) eager1, eager2 = GenerationMixin()._cache_dependant_input_preparation(input_ids, inputs_embeds, cache_position) export1, export2 = GenerationMixin()._cache_dependant_input_preparation_exporting( input_ids, inputs_embeds, cache_position @@ -2746,7 +2746,7 @@ class UtilsFunctionsTest(unittest.TestCase): # Case 3 input_ids = torch.randint(0, 16, (2, 12), dtype=torch.int64) inputs_embeds = None - cache_position = torch.range(0, 7, dtype=torch.int64) + cache_position = torch.arange(0, 8, dtype=torch.int64) eager1, eager2 = GenerationMixin()._cache_dependant_input_preparation(input_ids, inputs_embeds, cache_position) export1, export2 = GenerationMixin()._cache_dependant_input_preparation_exporting( input_ids, inputs_embeds, cache_position @@ -2757,7 +2757,7 @@ class UtilsFunctionsTest(unittest.TestCase): # Case 4 input_ids = torch.randint(0, 16, (2, 8), dtype=torch.int64) inputs_embeds = None - cache_position = torch.range(0, 7, dtype=torch.int64) + cache_position = torch.arange(0, 8, dtype=torch.int64) eager1, eager2 = GenerationMixin()._cache_dependant_input_preparation(input_ids, inputs_embeds, cache_position) export1, export2 = GenerationMixin()._cache_dependant_input_preparation_exporting( input_ids, inputs_embeds, cache_position diff --git a/tests/models/falcon_mamba/test_modeling_falcon_mamba.py b/tests/models/falcon_mamba/test_modeling_falcon_mamba.py index a75238265c..4e372a5fd9 100644 --- a/tests/models/falcon_mamba/test_modeling_falcon_mamba.py +++ b/tests/models/falcon_mamba/test_modeling_falcon_mamba.py @@ -231,7 +231,7 @@ class FalconMambaModelTester: token_emb, cache, cache_position=torch.arange(0, config.conv_kernel, device=input_ids.device) ) - loss = torch.log(1 + torch.abs(outputs.sum())) + loss = torch.log1p(torch.abs(outputs.sum())) self.parent.assertEqual(loss.shape, ()) self.parent.assertEqual(outputs.shape, (self.batch_size, self.seq_length, self.hidden_size)) loss.backward() diff --git a/tests/models/mamba/test_modeling_mamba.py b/tests/models/mamba/test_modeling_mamba.py index bd69446e3b..02ecbbfd3c 100644 --- a/tests/models/mamba/test_modeling_mamba.py +++ b/tests/models/mamba/test_modeling_mamba.py @@ -205,7 +205,7 @@ class MambaModelTester: token_emb, cache, cache_position=torch.arange(0, config.conv_kernel, device=input_ids.device) ) - loss = torch.log(1 + torch.abs(outputs.sum())) + loss = torch.log1p(torch.abs(outputs.sum())) self.parent.assertEqual(loss.shape, ()) self.parent.assertEqual(outputs.shape, (self.batch_size, self.seq_length, self.hidden_size)) loss.backward()