multiple gpu evaluation/calibration refine

README.md

@@ -57,12 +57,12 @@ pip install auto-round
 
 ### Basic Usage (Gaudi2/CPU/GPU)
 
-A user guide detailing the full list of supported arguments is provided by calling ```auto-round -h``` on the terminal.
+[//]: # (A user guide detailing the full list of supported arguments is provided by calling ```auto-round -h``` on the terminal.)
 Alternatively, you can use ```auto_round``` instead of ```auto-round```. Set the format you want in `format` and
-multiple formats exporting has been supported.
+multiple formats exporting has been supported. Please check out [step-by-step-instruction](./docs/step_by_step.md) for more details about calibration dataset or evaluation.
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 auto-round \
+auto-round \
     --model facebook/opt-125m \
     --bits 4 \
     --group_size 128 \
@@ -77,7 +77,7 @@ We provide two recipes for best accuracy and fast running speed with low memory.
 
   ```bash
 ## best accuracy, 3X slower, low_gpu_mem_usage could save ~20G but ~30% slower
-CUDA_VISIBLE_DEVICES=0 auto-round \
+auto-round \
     --model facebook/opt-125m \
     --bits 4 \
     --group_size 128 \
@@ -89,7 +89,7 @@ CUDA_VISIBLE_DEVICES=0 auto-round \
 
   ```bash
 ## fast and low memory, 2-3X speedup, slight accuracy drop at W4G128
-CUDA_VISIBLE_DEVICES=0 auto-round \
+auto-round \
     --model facebook/opt-125m \
     --bits 4 \
     --group_size 128 \

auto_round/autoround.py

@@ -264,7 +264,7 @@ def quantize(self):
         Returns:
         The quantized model and layer configurations.
         """
-        # logger.info("cache block input")
+
         if bool(self.quant_block_list):
             all_blocks = self.quant_block_list
         else:
@@ -280,10 +280,12 @@ def quantize(self):
         layer_names = self.get_quantized_layer_names_outside_blocks()
         self.start_time = time.time()
         all_first_block_names = [block[0] for block in all_blocks]
+        logger.info("start calibration")
         all_inputs = self.try_cache_inter_data_gpucpu(all_first_block_names, self.nsamples, layer_names=layer_names)
         if hasattr(self.model, "hf_device_map") and len(self.model.hf_device_map) > 1:
             accelerate.hooks.remove_hook_from_submodules(self.model)  ##self.model.hf_device_map has not been changed
         self.model = mv_module_from_gpu(self.model, self.low_cpu_mem_usage)
+        logger.info("calibration done")
         for block_names in all_blocks:
             inputs = all_inputs[block_names[0]]
             all_inputs.pop(block_names[0])
@@ -611,10 +613,12 @@ def try_cache_inter_data_gpucpu(self, block_names, nsamples, layer_names=None, l
         except RuntimeError as e:
             if "CUDA out of memory" in str(e):
                 logger.info("switch to cpu to cache inputs")
-                if "lm_head" in self.layer_config and self.layer_config["lm_head"]["bits"] < 8:
+                if (("lm_head" in self.layer_config and self.layer_config["lm_head"]["bits"] <= 16) or
+                        self.__class__.__name__=="AutoRoundMLLM") :
                     logger.warning(f"we strongly recommend using additional CUDA/HPU devices,e.g. "
-                                   f"'CUDA_VISIBLE_DEVICES=0,1 python xxx',"
-                                   f" for optimal performance during calibration when enabling lm-head quantization. "
+                                   f"set `--device '0,1'` in our cmd line usage or "
+                                   f"load the model with `device_mapping=auto`,"
+                                   f" for optimal performance during calibration "
                                    f"Otherwise, the process may be significantly slower.")
                 self.model = mv_module_from_gpu(self.model, self.low_cpu_mem_usage)
                 clear_memory()

auto_round/script/llm.py

@@ -25,24 +25,8 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
-import os
-import re
 import argparse
 
-import torch
-import transformers
-
-os.environ["CUBLAS_WORKSPACE_CONFIG"] = ":4096:8"
-torch.use_deterministic_algorithms(True, warn_only=True)
-from transformers import AutoModelForCausalLM, AutoTokenizer, AutoModel, AutoConfig, AutoProcessor
-from lm_eval.utils import make_table  # pylint: disable=E0401
-
-from auto_round import AutoRoundConfig
-from auto_round.eval.evaluation import simple_evaluate
-from auto_round.utils import detect_device, get_library_version, detect_device_count
-from auto_round.utils import logger
-
 
 class BasicArgumentParser(argparse.ArgumentParser):
     def __init__(self, *args, **kwargs):
@@ -59,7 +43,7 @@ def __init__(self, *args, **kwargs):
         self.add_argument("--eval_bs", default=None, type=int,
                           help="batch size in evaluation")
 
-        self.add_argument("--device", default="auto", type=str,
+        self.add_argument("--device", "--devices", default="auto", type=str,
                           help="the device to be used for tuning. The default is set to auto,"
                                "allowing for automatic detection."
                                "Currently, device settings support CPU, GPU, and HPU.")
@@ -234,7 +218,7 @@ def tune(args):
     supported_formats = ["auto_round", "auto_gptq", "auto_awq", "auto_round:gptq", "auto_round:auto_gptq",
                          "auto_round:auto_gptq:marlin", "auto_round:gptq:marlin", "auto_round:auto_awq",
                          "auto_round:awq", "auto_gptq:marlin", "itrex", "iterx_xpu", "fake"]
-    formats =  args.format.replace(' ', '').split(",")
+    formats = args.format.replace(' ', '').split(",")
     for format in formats:
         if format not in supported_formats:
             raise ValueError(f"{format} is not supported, we only support {supported_formats}")
@@ -247,12 +231,33 @@ def tune(args):
     if "marlin" in args.format and args.asym is True:
         assert False, "marlin backend only supports sym quantization, please remove --asym"
 
+    ##must set this before import torch
+    import os
+    devices = args.device.split(',')
+    use_auto_mapping = False
+    if all(s.isdigit() for s in devices):
+        os.environ["CUDA_VISIBLE_DEVICES"] = args.device
+        use_auto_mapping = True
+
+    import re
+    import torch
+    import transformers
+
+    os.environ["CUBLAS_WORKSPACE_CONFIG"] = ":4096:8"
+    torch.use_deterministic_algorithms(True, warn_only=True)
+    from transformers import AutoModelForCausalLM, AutoTokenizer, AutoModel, AutoConfig, AutoProcessor
+    from lm_eval.utils import make_table  # pylint: disable=E0401
+
+    from auto_round import AutoRoundConfig
+    from auto_round.eval.evaluation import simple_evaluate
+    from auto_round.utils import detect_device, get_library_version, detect_device_count
+    from auto_round.utils import logger
+
     model_name = args.model
     if model_name[-1] == "/":
         model_name = model_name[:-1]
     logger.info(f"start to quantize {model_name}")
-
-    device_str = detect_device(args.device)
+    device_str = detect_device(devices[0])
     torch_dtype = "auto"
     if "hpu" in device_str:
         torch_dtype = torch.bfloat16
@@ -289,7 +294,7 @@ def tune(args):
             trust_remote_code=not args.disable_trust_remote_code
         )
     else:
-        if detect_device_count() > 1:
+        if use_auto_mapping:
             model = model_cls.from_pretrained(
                 model_name, low_cpu_mem_usage=True, torch_dtype=torch_dtype,
                 trust_remote_code=not args.disable_trust_remote_code, device_map="auto"
@@ -370,7 +375,6 @@ def tune(args):
                 raise ValueError(
                     f"{format} is not supported for lm-head quantization, please change to {auto_round_formats}")
 
-
     autoround = round(
         model, tokenizer, args.bits, args.group_size, sym=not args.asym, batch_size=args.batch_size,
         dataset=args.dataset, seqlen=seqlen, nblocks=args.nblocks, iters=args.iters, lr=args.lr,
@@ -433,8 +437,18 @@ def tune(args):
 
 
 def eval(args):
-    device_str = detect_device(args.device)
+    import os
+    devices = args.device.split(",")
+    parallelism = False
+    if all(s.isdigit() for s in devices):
+        os.environ["CUDA_VISIBLE_DEVICES"] = args.device
+        parallelism = True
+        device_str = None
+    from auto_round.eval.evaluation import simple_evaluate
+
     model_args = f"pretrained={args.model},trust_remote_code={not args.disable_trust_remote_code}"
+    if parallelism:
+        model_args += ",parallelize=True"
     if isinstance(args.tasks, str):
         tasks = args.tasks.split(',')
     res = simple_evaluate(

auto_round/script/mllm.py

@@ -42,7 +42,7 @@ def __init__(self, *args, **kwargs):
         self.add_argument("--eval_bs", default=None, type=int,
                           help="batch size in evaluation")
 
-        self.add_argument("--device", default="auto", type=str,
+        self.add_argument("--device", "--devices", default="auto", type=str,
                           help="the device to be used for tuning. The default is set to auto,"
                                "allowing for automatic detection."
                                "Currently, device settings support CPU, GPU, and HPU.")
@@ -206,7 +206,7 @@ def tune(args):
                          "auto_round:auto_gptq:marlin", "auto_round:gptq:marlin", "auto_round:auto_awq",
                          "auto_round:awq"]
     if not args.quant_nontext_module:
-        supported_formats.append("auto_gptq","auto_gptq:marlin")
+        supported_formats.extend(["auto_gptq", "auto_gptq:marlin"])
 
     formats = args.format.replace(' ', '').split(",")
     for format in formats:
@@ -220,7 +220,13 @@ def tune(args):
 
     assert args.dataset is not None, "dataset should not be None."
 
-    device_str = detect_device(args.device)
+    devices = args.device.split(',')
+    use_auto_mapping = False
+    if torch.cuda.is_available() and all(s.isdigit() for s in devices):
+        os.environ["CUDA_VISIBLE_DEVICES"] = args.device
+        use_auto_mapping = True
+    device_str = detect_device(devices[0])
+
     torch_dtype = "auto"
     if "hpu" in device_str:
         torch_dtype = torch.bfloat16
@@ -239,8 +245,13 @@ def tune(args):
         cls = MllamaForConditionalGeneration
     else:
         cls = AutoModelForCausalLM
-    model = cls.from_pretrained(
-        model_name, trust_remote_code=not args.disable_trust_remote_code, torch_dtype=torch_dtype)
+    if use_auto_mapping:
+        model = cls.from_pretrained(
+            model_name, trust_remote_code=not args.disable_trust_remote_code, torch_dtype=torch_dtype)
+    else:
+        model = cls.from_pretrained(
+            model_name, trust_remote_code=not args.disable_trust_remote_code, torch_dtype=torch_dtype,
+            device_map="auto")
 
     if "cogvlm2" in model_name:
         model.config.model_type = "cogvlm2"