kyy/llm_trainer
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

use pre-commit

Browse Source
This commit is contained in:
hiyouga
2024-10-29 09:07:46 +00:00
parent 77666bd227
commit 21db8ed2f4
86 changed files with 1048 additions and 1064 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
src/llamafactory/model/model_utils/quantization.py
View File

@@ -130,7 +130,7 @@ def configure_quantization(
quantization_config["bits"] = 2
quant_bits = quantization_config.get("bits", "?")
logger.info("Loading {}-bit {}-quantized model.".format(quant_bits, quant_method.upper()))
logger.info(f"Loading {quant_bits}-bit {quant_method.upper()}-quantized model.")
elif model_args.export_quantization_bit is not None: # auto-gptq
if model_args.export_quantization_bit not in [8, 4, 3, 2]:
@@ -149,7 +149,7 @@ def configure_quantization(
)
init_kwargs["device_map"] = "auto"
init_kwargs["max_memory"] = get_max_memory()
logger.info("Quantizing model to {} bit with AutoGPTQ.".format(model_args.export_quantization_bit))
logger.info(f"Quantizing model to {model_args.export_quantization_bit} bit with AutoGPTQ.")
elif model_args.quantization_bit is not None: # on-the-fly
if model_args.quantization_method == QuantizationMethod.BITS_AND_BYTES.value:
@@ -179,7 +179,7 @@ def configure_quantization(
else:
init_kwargs["device_map"] = {"": get_current_device()} # change auto device map for inference
logger.info("Quantizing model to {} bit with bitsandbytes.".format(model_args.quantization_bit))
logger.info(f"Quantizing model to {model_args.quantization_bit} bit with bitsandbytes.")
elif model_args.quantization_method == QuantizationMethod.HQQ.value:
if model_args.quantization_bit not in [8, 6, 5, 4, 3, 2, 1]:
raise ValueError("HQQ only accepts 1/2/3/4/5/6/8-bit quantization.")
@@ -191,7 +191,7 @@ def configure_quantization(
init_kwargs["quantization_config"] = HqqConfig(
nbits=model_args.quantization_bit, quant_zero=False, quant_scale=False, axis=0
) # use ATEN kernel (axis=0) for performance
logger.info("Quantizing model to {} bit with HQQ.".format(model_args.quantization_bit))
logger.info(f"Quantizing model to {model_args.quantization_bit} bit with HQQ.")
elif model_args.quantization_method == QuantizationMethod.EETQ.value:
if model_args.quantization_bit != 8:
raise ValueError("EETQ only accepts 8-bit quantization.")
@@ -201,4 +201,4 @@ def configure_quantization(
require_version("eetq", "To fix: pip install eetq")
init_kwargs["quantization_config"] = EetqConfig()
logger.info("Quantizing model to {} bit with EETQ.".format(model_args.quantization_bit))
logger.info(f"Quantizing model to {model_args.quantization_bit} bit with EETQ.")