first commit

tests/models/ernie4_5_moe/test_modeling_ernie4_5_moe.py

@@ -0,0 +1,226 @@
+# Copyright 2025 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Testing suite for the PyTorch Ernie4.5 MoE model."""
+
+import tempfile
+import unittest
+
+import pytest
+
+from transformers import BitsAndBytesConfig, is_torch_available
+from transformers.models.ernie4_5_moe.modeling_ernie4_5_moe import load_balancing_loss_func
+from transformers.testing_utils import (
+    cleanup,
+    is_flaky,
+    require_bitsandbytes,
+    require_flash_attn,
+    require_torch,
+    require_torch_accelerator,
+    require_torch_large_accelerator,
+    slow,
+    torch_device,
+)
+from transformers.trainer_utils import set_seed
+
+
+if is_torch_available():
+    import torch
+
+    from transformers import (
+        AutoTokenizer,
+        Ernie4_5_MoeForCausalLM,
+        Ernie4_5_MoeModel,
+    )
+
+from ...causal_lm_tester import CausalLMModelTest, CausalLMModelTester
+
+
+class Ernie4_5_MoeModelTester(CausalLMModelTester):
+    if is_torch_available():
+        base_model_class = Ernie4_5_MoeModel
+
+
+@require_torch
+class Ernie4_5_MoeModelTest(CausalLMModelTest, unittest.TestCase):
+    test_all_params_have_gradient = False
+    model_tester_class = Ernie4_5_MoeModelTester
+
+    @require_flash_attn
+    @require_torch_accelerator
+    @pytest.mark.flash_attn_test
+    @is_flaky()
+    @slow
+    def test_flash_attn_2_equivalence(self):
+        for model_class in self.all_model_classes:
+            if not model_class._supports_flash_attn:
+                self.skipTest(reason="Model does not support Flash Attention 2")
+
+            config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
+            model = model_class(config)
+
+            with tempfile.TemporaryDirectory() as tmpdirname:
+                model.save_pretrained(tmpdirname)
+                model_fa = model_class.from_pretrained(
+                    tmpdirname, dtype=torch.bfloat16, attn_implementation="flash_attention_2"
+                )
+                model_fa.to(torch_device)
+
+                model = model_class.from_pretrained(tmpdirname, dtype=torch.bfloat16, attn_implementation="eager")
+                model.to(torch_device)
+
+                dummy_input = inputs_dict[model_class.main_input_name]
+                dummy_input = dummy_input.to(torch_device)
+                outputs = model(dummy_input, output_hidden_states=True)
+                outputs_fa = model_fa(dummy_input, output_hidden_states=True)
+
+                logits = outputs.hidden_states[-1]
+                logits_fa = outputs_fa.hidden_states[-1]
+
+                # higher tolerance, not sure where it stems from
+                assert torch.allclose(logits_fa, logits, atol=1e-2, rtol=1e-2)
+
+    @is_flaky(max_attempts=2)
+    def test_load_balancing_loss(self):
+        r"""
+        Let's make sure we can actually compute the loss and do a backward on it.
+        """
+        set_seed(42)
+        config, input_dict = self.model_tester.prepare_config_and_inputs_for_common()
+        config.num_labels = 3
+        config.num_experts = 3
+        config.output_router_logits = True
+        input_ids = input_dict["input_ids"]
+        attention_mask = input_ids.ne(config.pad_token_id).to(torch_device)
+        model = Ernie4_5_MoeForCausalLM(config)
+        model.to(torch_device)
+        model.eval()
+        result = model(input_ids, attention_mask=attention_mask)
+        bs, seqlen = input_ids.shape
+        self.assertEqual(result.router_logits[0].shape, (bs * seqlen, config.num_experts))
+        torch.testing.assert_close(result.aux_loss.cpu(), torch.tensor(2, dtype=torch.float32), rtol=1e-2, atol=1e-2)
+
+        # First, we make sure that adding padding tokens doesn't change the loss
+        # loss(input_ids, attention_mask=None) == loss(input_ids + padding, attention_mask=attention_mask_with_padding)
+        # (This length is selected from experiments)
+        pad_length = input_ids.shape[1] * 4
+        # Add extra tokens to input_ids and mask them out to simulate left padding
+        padding_block = torch.randint(
+            low=0,
+            high=config.vocab_size,
+            size=(input_ids.shape[0], pad_length),
+            dtype=torch.int32,
+            device=torch_device,
+        )
+        padding_block[padding_block == config.pad_token_id] = (config.pad_token_id + 1) % config.vocab_size
+        padded_input_ids = torch.cat((padding_block, input_ids), dim=1)
+        padded_attention_mask = torch.zeros_like(padded_input_ids, dtype=torch.long)
+        padded_attention_mask[:, pad_length:] = 1
+
+        padded_result = model(padded_input_ids, attention_mask=padded_attention_mask)
+        torch.testing.assert_close(result.aux_loss.cpu(), padded_result.aux_loss.cpu(), rtol=1e-4, atol=1e-4)
+
+        # We make sure that masking can change the loss using a deterministic synthetic example.
+        # This avoids flakiness when the model routes tokens uniformly.
+        num_experts = 3
+        top_k = 1
+        synthetic_logits = torch.tensor(
+            [
+                [10.0, 0.0, 0.0],  # unmasked token -> expert 0
+                [10.0, 0.0, 0.0],  # unmasked token -> expert 0
+                [0.0, 10.0, 0.0],  # masked token -> expert 1
+                [0.0, 10.0, 0.0],  # masked token -> expert 1
+            ],
+            device=torch_device,
+        )
+        synthetic_mask = torch.tensor([[1, 1, 0, 0]], device=torch_device)
+        masked_loss = load_balancing_loss_func((synthetic_logits,), num_experts, top_k, synthetic_mask)
+        unmasked_loss = load_balancing_loss_func((synthetic_logits,), num_experts, top_k, attention_mask=None)
+        self.assertNotAlmostEqual(masked_loss.item(), unmasked_loss.item(), places=6)
+
+
+@slow
+@require_torch
+class Ernie4_5_MoeIntegrationTest(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = None
+
+    @classmethod
+    def tearDownClass(cls):
+        del cls.model
+        cleanup(torch_device, gc_collect=True)
+
+    def setup(self):
+        cleanup(torch_device, gc_collect=True)
+
+    def tearDown(self):
+        cleanup(torch_device, gc_collect=True)
+
+    @classmethod
+    def get_large_model(cls):
+        cls.model = Ernie4_5_MoeForCausalLM.from_pretrained(
+            "baidu/ERNIE-4.5-21B-A3B-PT",
+            device_map="auto",
+            experts_implementation="eager",
+            quantization_config=BitsAndBytesConfig(load_in_4bit=True),
+        )
+
+        return cls.model
+
+    @classmethod
+    def get_small_model(cls):
+        cls.model = Ernie4_5_MoeForCausalLM.from_pretrained(
+            "hf-internal-testing/ERNIE-4.5-Small-Moe", device_map="auto", dtype="auto", experts_implementation="eager"
+        )
+
+        return cls.model
+
+    @require_torch_large_accelerator(memory=48)  # Tested on A100 but requires around 48GiB
+    @require_bitsandbytes
+    def test_model_21b_a3b_generation(self):
+        EXPECTED_TEXT_COMPLETION = "User: Hey, are you conscious? Can you talk to me?\nAssistant: \nI don't have consciousness in the way humans do. I don't feel emotions, have thoughts, or experience awareness. However, I'm"  # fmt: skip
+
+        model = self.get_large_model()
+        tokenizer = AutoTokenizer.from_pretrained("baidu/ERNIE-4.5-21B-A3B-PT")
+        prompt = "Hey, are you conscious? Can you talk to me?"
+        messages = [{"role": "user", "content": prompt}]
+        text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+        model_inputs = tokenizer([text], add_special_tokens=False, return_tensors="pt").to(model.device)
+
+        generated_ids = model.generate(
+            model_inputs.input_ids,
+            max_new_tokens=32,
+            do_sample=False,
+        )
+        text = tokenizer.decode(generated_ids[0], skip_special_tokens=True).strip("\n")
+        self.assertEqual(EXPECTED_TEXT_COMPLETION, text)
+
+    def test_shortened_model_generation(self):
+        # This is gibberish which is expected as the model are the first x layers of the original 28B model
+        EXPECTED_TEXT_COMPLETION = 'User: Hey, are you conscious? Can you talk to me?\nAssistant: 不了的 tongues说话 dagat绵席裹着头phones<mask:11>odikèkèk<mask:11><mask:11>bun褶席席地说起来这么说的话的话retti upside upsideolate疡疡疡'  # fmt: skip
+
+        model = self.get_small_model()
+        tokenizer = AutoTokenizer.from_pretrained("baidu/ERNIE-4.5-21B-A3B-PT")
+        prompt = "Hey, are you conscious? Can you talk to me?"
+        messages = [{"role": "user", "content": prompt}]
+        text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+        model_inputs = tokenizer([text], add_special_tokens=False, return_tensors="pt").to(model.device)
+
+        generated_ids = model.generate(
+            model_inputs.input_ids,
+            max_new_tokens=32,
+            do_sample=False,
+        )
+        text = tokenizer.decode(generated_ids[0], skip_special_tokens=True).strip("\n")
+        self.assertEqual(EXPECTED_TEXT_COMPLETION, text)