Skip to content

Fix Failing GPTQ tests #36666

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
MekkCyber merged 2 commits into from
Mar 12, 2025
Merged

Fix Failing GPTQ tests #36666

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
15b715c
fix tests
MekkCyber Mar 12, 2025
4141926
Merge branch 'main' into fixing_tests
MekkCyber Mar 12, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 16 additions & 5 deletions tests/quantization/gptq/test_gptq.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,6 +94,7 @@ class GPTQTest(unittest.TestCase):
EXPECTED_OUTPUTS.add("Hello my name is Aiden, I am a student at the University")
EXPECTED_OUTPUTS.add("Hello my name is Nate and I am a member of the N")
EXPECTED_OUTPUTS.add("Hello my name is Nellie and I am a student at the")
EXPECTED_OUTPUTS.add("Hello my name is Nate and I am a new member of the")

# this seems a little small considering that we are doing 4bit quant but we have a small model and ww don't quantize the embeddings
EXPECTED_RELATIVE_DIFFERENCE = 1.664253062
Expand Down Expand Up @@ -260,7 +261,9 @@ def test_serialization(self):
if self.device_map == "cpu":
quant_type = "ipex" if is_ipex_available() else "torch"
else:
quant_type = "exllama"
# We expecte tritonv2 to be used here, because exllama backend doesn't support packing https://github.com/ModelCloud/GPTQModel/issues/1354
# TODO: Remove this once GPTQModel exllama kernels supports packing
quant_type = "tritonv2"
quantized_model_from_saved = AutoModelForCausalLM.from_pretrained(
tmpdirname, device_map=self.device_map
)
Expand Down Expand Up @@ -424,10 +427,18 @@ def setUpClass(cls):
cls.tokenizer = AutoTokenizer.from_pretrained(cls.model_name, use_fast=True)

def test_quantized_layers_type(self):
self.assertEqual(
self.quantized_model.model.layers[0].self_attn.k_proj.QUANT_TYPE,
"exllama" if is_gptqmodel_available() else "exllamav2",
)
if is_auto_gptq_available() and not is_gptqmodel_available():
self.assertEqual(
self.quantized_model.model.layers[0].self_attn.k_proj.QUANT_TYPE,
"exllamav2",
)
else:
# We expecte tritonv2 to be used here, because exllama backend doesn't support packing https://github.com/ModelCloud/GPTQModel/issues/1354
# TODO: Remove this once GPTQModel exllama kernels supports packing
self.assertEqual(
self.quantized_model.model.layers[0].self_attn.k_proj.QUANT_TYPE,
"tritonv2",
)

def check_inference_correctness(self, model):
"""
Expand Down