Skip to content

fix: FP8 quantized lm_head (NvBug 5214229) #3567

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
merged 1 commit into from
Apr 17, 2025
Merged

fix: FP8 quantized lm_head (NvBug 5214229) #3567

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
25 changes: 15 additions & 10 deletions tensorrt_llm/quantization/quantize.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,14 +28,16 @@ def quantize_layers(
quant_map,
preprocess_init_params=None,
):
exclude_modules = quant_config.exclude_modules or [
'*lm_head',
'*router',
'*vocab_embedding',
'*position_embedding',
'*block_embedding',
'*shared_expert_gate',
]
exclude_modules = quant_config.exclude_modules
if exclude_modules is None:
exclude_modules = [
'*lm_head',
'*router',
'*vocab_embedding',
'*position_embedding',
'*block_embedding',
'*shared_expert_gate',
]

for name, module, parent in model.named_modules_with_parent():
module_name = name.rsplit('.', 1)[-1]
Expand Down Expand Up @@ -244,9 +246,12 @@ def fp8_rowwise_quantize(model, quant_config: QuantConfig):
Attention: Fp8RowwiseAttention,
}

exclude_modules = quant_config.exclude_modules
if exclude_modules is None:
exclude_modules = []
# Always exclude these modules for FP8 rowwise
exclude_modules = list(
set((quant_config.exclude_modules or []) +
['*ln_f', '*ln_embed', '*lm_head']))
set(exclude_modules + ['*ln_f', '*ln_embed', '*lm_head']))

def extract_layer_idx(name):
ss = name.split('.')
Expand Down
1 change: 0 additions & 1 deletion tests/integration/test_lists/waives.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -419,7 +419,6 @@ examples/test_multimodal.py::test_llm_multimodal_general[neva-22b-pp:1-tp:1-bflo
examples/test_recurrentgemma.py::test_llm_recurrentgemma_1gpu[use_py_session-recurrentgemma-2b-no_paged_cache-disable_quant-float16-disable_attn_plugin-enable_gemm_plugin] SKIP (https://nvbugs/5214221)
examples/test_recurrentgemma.py::test_llm_recurrentgemma_1gpu[use_py_session-recurrentgemma-2b-no_paged_cache-disable_quant-float16-enable_attn_plugin-enable_gemm_plugin] SKIP (https://nvbugs/5214221)
examples/test_recurrentgemma.py::test_llm_recurrentgemma_1gpu[use_py_session-recurrentgemma-2b-use_paged_cache-disable_quant-float16-enable_attn_plugin-enable_gemm_plugin] SKIP (https://nvbugs/5214221)
accuracy/test_cli_flow.py::TestGpt2Medium::test_fp8_lm_head SKIP (https://nvbugs/5214229)
examples/test_multimodal.py::test_llm_multimodal_general[VILA1.5-3b-pp:1-tp:1-float16-bs:1-cpp_e2e:False-nb:1] SKIP (https://nvbugs/5214239)
examples/test_multimodal.py::test_llm_fp8_multimodal_general[fp8-fp8-scienceqa-Llama-3.2-11B-Vision-Instruct-pp:1-tp:1-bfloat16-bs:1-cpp_e2e:False] SKIP (https://nvbugs/5222697)
examples/test_gpt.py::test_llm_gpt2_santacoder_1node_4gpus[parallel_build-enable_fmha-enable_gemm_plugin-enable_attention_plugin] SKIP (https://nvbugs/5219531)
Expand Down