Skip to content

Commit

Permalink
update metric
Browse files Browse the repository at this point in the history
  • Loading branch information
yujiepan-work committed Jul 8, 2024
1 parent 334a9ca commit 17792bd
Show file tree
Hide file tree
Showing 2 changed files with 9 additions and 9 deletions.
Original file line number Diff line number Diff line change
Expand Up @@ -296,7 +296,7 @@ def _dump_model_fp32(self):
def _compress(self):
super()._compress()
if self.backend in PT_BACKENDS:
# This helps reproducibility but is not needed in actual usage.
# This helps reproducibility but is not needed in actual use.
for sparsifier in PTSparsifyActivationsAlgoBackend.get_sparsifiers(self.compressed_model):
original_dtype = sparsifier.running_threshold.dtype
sparsifier.running_threshold = sparsifier.running_threshold.half().to(original_dtype)
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -4,26 +4,26 @@ tinyllama_backend_FP32:
num_int8: 0
num_sparse_activations: 0
tinyllama_ffn_sparse20_backend_CUDA_TORCH:
metric_value: 0.7970
atol: 0.005
metric_value: 0.7858
atol: 0.02
num_int4: 0
num_int8: 0
num_sparse_activations: 44
tinyllama_ffn_sparse20_backend_TORCH:
metric_value: 0.7697
atol: 0.005
metric_value: 0.7882
atol: 0.02
num_int4: 0
num_int8: 0
num_sparse_activations: 44
tinyllama_int8_asym_data_free_ffn_sparse20_backend_CUDA_TORCH:
metric_value: 0.7952
atol: 0.005
metric_value: 0.8044
atol: 0.02
num_int4: 0
num_int8: 312
num_sparse_activations: 44
tinyllama_int8_asym_data_free_ffn_sparse20_backend_TORCH:
metric_value: 0.8013
atol: 0.005
metric_value: 0.7977
atol: 0.02
num_int4: 0
num_int8: 312
num_sparse_activations: 44
Expand Down

0 comments on commit 17792bd

Please sign in to comment.