|
@@ -59,6 +59,7 @@ def load_pretrained_block(
|
|
|
block_prefix = f"{config.block_prefix}.{block_index}."
|
|
|
|
|
|
if is_gptq_quant(config):
|
|
|
+ print("Now loading GPTQ")
|
|
|
hf_quantizer = AutoHfQuantizer.from_config(config.quantization_config, pre_quantized=True)
|
|
|
hf_quantizer.optimum_quantizer.block_name_to_quantize = str(block_index)
|
|
|
tmp_block_list = torch.nn.ModuleList([block])
|
|
@@ -77,7 +78,7 @@ def load_pretrained_block(
|
|
|
cache_dir=cache_dir,
|
|
|
max_disk_space=max_disk_space,
|
|
|
)
|
|
|
-
|
|
|
+ print("now printing", block)
|
|
|
for param_name, _ in block.named_parameters():
|
|
|
assert param_name in state_dict, f"{param_name} not in state dict"
|
|
|
param = state_dict[param_name]
|