riaz.somc/transformers
1
0
Fork 0
mirror of https://github.com/huggingface/transformers.git synced 2025-08-01 02:31:11 +06:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

Fix new BNB test failures (#35345)

Browse Source
This commit is contained in:
Matthew Douglas 2025-01-02 05:24:52 -05:00 committed by GitHub
parent 5b516b06c8
commit 6b1e86fd4d
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 7 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tests/quantization/bnb/test_4bit.py
View File

@ -172,7 +172,7 @@ class Bnb4BitTest(Base4bitTest):
mem_fp16 = self.model_fp16.get_memory_footprint() mem_fp16 = self.model_fp16.get_memory_footprint()
mem_4bit = self.model_4bit.get_memory_footprint() mem_4bit = self.model_4bit.get_memory_footprint()
self.assertAlmostEqual(mem_fp16 / mem_4bit, self.EXPECTED_RELATIVE_DIFFERENCE) self.assertAlmostEqual(mem_fp16 / mem_4bit, self.EXPECTED_RELATIVE_DIFFERENCE, delta=1e-5)
linear = get_some_linear_layer(self.model_4bit) linear = get_some_linear_layer(self.model_4bit)
self.assertTrue(linear.weight.__class__ == Params4bit) self.assertTrue(linear.weight.__class__ == Params4bit)

7
tests/quantization/bnb/test_mixed_int8.py
View File

@ -229,7 +229,7 @@ class MixedInt8Test(BaseMixedInt8Test):
mem_fp16 = self.model_fp16.get_memory_footprint() mem_fp16 = self.model_fp16.get_memory_footprint()
mem_8bit = self.model_8bit.get_memory_footprint() mem_8bit = self.model_8bit.get_memory_footprint()
self.assertAlmostEqual(mem_fp16 / mem_8bit, self.EXPECTED_RELATIVE_DIFFERENCE) self.assertAlmostEqual(mem_fp16 / mem_8bit, self.EXPECTED_RELATIVE_DIFFERENCE, delta=1e-5)
self.assertTrue(get_some_linear_layer(self.model_8bit).weight.__class__ == Int8Params) self.assertTrue(get_some_linear_layer(self.model_8bit).weight.__class__ == Int8Params)
def test_linear_are_8bit(self): def test_linear_are_8bit(self):
@ -938,8 +938,13 @@ class MixedInt8LlamaTest(MixedInt8Test):
model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0" model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
EXPECTED_RELATIVE_DIFFERENCE = 1.7869331026479096 EXPECTED_RELATIVE_DIFFERENCE = 1.7869331026479096
EXPECTED_OUTPUTS = set() EXPECTED_OUTPUTS = set()
# Expected on Intel XPU
EXPECTED_OUTPUTS.add("Hello my name is John Smith and I am a software engineer. I") EXPECTED_OUTPUTS.add("Hello my name is John Smith and I am a software engineer. I")
# Expected on NVIDIA T4
EXPECTED_OUTPUTS.add("Hello my name is John and I am a software engineer. I have")
def test_int8_from_pretrained(self): def test_int8_from_pretrained(self):
r""" r"""
Test whether loading a 8bit model from the Hub works as expected Test whether loading a 8bit model from the Hub works as expected