test_compressed_tensors_runtime¶

Functions¶

`test_create_weights_registers_checkpoint_parameter_names`()
`test_process_and_apply_use_gptq_repack_and_uint4b8`(...)
`test_w8a8_create_weights_registers_weight_and_scale`()
`test_w8a8_process_weights_transposes_and_flattens_scales`()
`test_w8a8_apply_matches_reference_for_large_m`()
`test_w8a8_apply_supports_small_m_by_padding`()
`test_w8a8_apply_uses_triton_quant_and_torch_int_mm`(...)	Verify the W8A8 forward path uses Triton activation quant + torch._int_mm.

test_compressed_tensors_runtime.test_create_weights_registers_checkpoint_parameter_names()¶

test_compressed_tensors_runtime.test_process_and_apply_use_gptq_repack_and_uint4b8(monkeypatch)¶

test_compressed_tensors_runtime.test_w8a8_create_weights_registers_weight_and_scale()¶

test_compressed_tensors_runtime.test_w8a8_process_weights_transposes_and_flattens_scales()¶

test_compressed_tensors_runtime.test_w8a8_apply_matches_reference_for_large_m()¶

test_compressed_tensors_runtime.test_w8a8_apply_supports_small_m_by_padding()¶

test_compressed_tensors_runtime.test_w8a8_apply_uses_triton_quant_and_torch_int_mm(monkeypatch)¶

Verify the W8A8 forward path uses Triton activation quant + torch._int_mm.