Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix warnings and errors on cpu benchmark #20967

Merged
merged 5 commits into from
Jun 13, 2024
Merged

Fix warnings and errors on cpu benchmark #20967

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
df91085
Fix warnings and errors on cpu benchmark
hanbitmyths Jun 7, 2024
bc1a595
Fix lint errors
hanbitmyths Jun 7, 2024
676e4e5
Remove checking position_ids
hanbitmyths Jun 12, 2024
b56d8d1
Delete warnings of unnecessary inputs
hanbitmyths Jun 12, 2024
9c8208a
Fix a lint error
hanbitmyths Jun 12, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
81 changes: 58 additions & 23 deletions onnxruntime/python/tools/transformers/models/llama/benchmark_e2e.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,22 +69,36 @@ def get_model(args: argparse.Namespace):
cache_dir=args.cache_dir,
torch_dtype=args.torch_dtype,
use_auth_token=args.auth,
trust_remote_code=args.auth,
trust_remote_code=args.trust,
use_cache=True,
attn_implementation="flash_attention_2",
quantization_config=bnb_config,
max_memory={args.device_id: "80GB"},
)
else:
model = AutoModelForCausalLM.from_pretrained(
args.hf_dir_path if args.hf_dir_path != "" else args.model_name,
cache_dir=args.cache_dir,
torch_dtype=args.torch_dtype,
use_auth_token=args.auth,
trust_remote_code=args.auth,
use_cache=True,
attn_implementation=("flash_attention_2" if args.device == "cuda" else "sdpa"),
).to(args.target_device)
try:
model = AutoModelForCausalLM.from_pretrained(
args.hf_dir_path if args.hf_dir_path != "" else args.model_name,
cache_dir=args.cache_dir,
torch_dtype=args.torch_dtype,
use_auth_token=args.auth,
trust_remote_code=args.trust,
use_cache=True,
attn_implementation=("flash_attention_2" if args.device == "cuda" else "sdpa"),
).to(args.target_device)
except Exception as e:
# When flash_attention or sdpa doesn't support a model, it throws an exception.
# Rather than stopping a process, run as eager mode.
print("Try to load a model using eager mode: ", e)
model = AutoModelForCausalLM.from_pretrained(
args.hf_dir_path if args.hf_dir_path != "" else args.model_name,
cache_dir=args.cache_dir,
torch_dtype=args.torch_dtype,
use_auth_token=args.auth,
trust_remote_code=args.trust,
use_cache=True,
attn_implementation="eager",
).to(args.target_device)

model.eval()

Expand All @@ -103,6 +117,18 @@ def get_model(args: argparse.Namespace):
return model


# When it runs a model without position_ids input, ORT keep printint out a complaint.
# Check if a model has a position_ids input and suppress if not.
def has_position_ids(args):
hanbitmyths marked this conversation as resolved.
Show resolved Hide resolved
if args.benchmark_type != "ort":
return True

import onnx

model = onnx.load(args.onnx_model_path, load_external_data=False)
return any(input.name == "position_ids" for input in model.graph.input)


def run_inference(args, model, runs, inputs, outputs):
if args.benchmark_type == "pt-compile":
with torch.no_grad():
Expand Down Expand Up @@ -134,10 +160,10 @@ def run_inference(args, model, runs, inputs, outputs):
return avg, outputs


def prepare_model_for_inference(args, model, config, tokenizer, prompt_length, prompt):
def prepare_model_for_inference(args, model, config, tokenizer, prompt_length, prompt, use_position_ids):
clear_cache()
inputs, outputs = get_initial_inputs_and_outputs(
config, tokenizer, prompt_length, prompt, args.target_device, args.use_fp16, args.use_buffer_share, args.engine
config, tokenizer, prompt_length, prompt, args.target_device, args.use_fp16, args.use_buffer_share, args.engine, use_position_ids
)
_, outputs = run_inference(args, model, args.warmup_runs, inputs, outputs)
return inputs, outputs
Expand Down Expand Up @@ -200,6 +226,14 @@ def get_args():
help="Use Hugging Face authentication token to access model",
)

parser.add_argument(
"-t",
"--trust",
default=False,
action="store_true",
help="Whether or not to allow for custom models defined on the Hugging Face Hub in their own modeling files",
)

parser.add_argument(
"-c",
"--cache-dir",
Expand Down Expand Up @@ -322,6 +356,8 @@ def get_args():
setattr(args, "engine", engine) # noqa: B010
setattr(args, "use_fp16", args.precision == "fp16") # noqa: B010

args.use_buffer_share = args.use_buffer_share and engine == "ort"

return args


Expand All @@ -340,16 +376,18 @@ def main():
args.hf_dir_path if args.hf_dir_path != "" else args.model_name,
cache_dir=args.cache_dir,
use_auth_token=args.auth,
trust_remote_code=args.auth,
trust_remote_code=args.trust,
)
tokenizer = AutoTokenizer.from_pretrained(
args.hf_dir_path if args.hf_dir_path != "" else args.model_name,
cache_dir=args.cache_dir,
use_auth_token=args.auth,
trust_remote_code=args.auth,
trust_remote_code=args.trust,
)
model = get_model(args)

use_position_ids = has_position_ids(args)

all_csv_metrics = []
for batch_size, prompt_length in itertools.product(args.batch_sizes, args.prompt_lengths):
batch_size, prompt_length = int(batch_size), int(prompt_length) # noqa: PLW2901
Expand All @@ -375,7 +413,7 @@ def main():
try:
# Measure prompt processing
logger.info("Measuring prompt processing...")
inputs, outputs = prepare_model_for_inference(args, model, config, tokenizer, prompt_length, prompt)
inputs, outputs = prepare_model_for_inference(args, model, config, tokenizer, prompt_length, prompt, use_position_ids)
accelerator_prompt_latency_s, outputs = run_inference(args, model, args.num_runs, inputs, outputs)

# Calculate prompt metrics
Expand All @@ -390,7 +428,7 @@ def main():
# Measure token generation
logger.info("Measuring token generation...")
clear_cache()
inputs, outputs = prepare_model_for_inference(args, model, config, tokenizer, prompt_length, prompt)
inputs, outputs = prepare_model_for_inference(args, model, config, tokenizer, prompt_length, prompt, use_position_ids)

all_token_ids = inputs["input_ids"].clone()
current_length = all_token_ids.shape[-1]
Expand Down Expand Up @@ -442,11 +480,8 @@ def main():
inputs["attention_mask"] = torch.cat(
[inputs["attention_mask"], (~has_eos).to(torch.int64).reshape(batch_size, 1)], 1
)
inputs["position_ids"] = (
None
if "position_ids" not in inputs
else torch.max(inputs["position_ids"], dim=1)[0].reshape(batch_size, 1) + 1
)
if use_position_ids:
inputs["position_ids"] = torch.max(inputs["position_ids"], dim=1)[0].reshape(batch_size, 1) + 1

# Set logits to zeros for next inference run and re-use memory buffer
if outputs["logits"].shape[1] != 1:
Expand Down Expand Up @@ -574,8 +609,8 @@ def main():
)
all_csv_metrics.append(csv_metrics)

except: # noqa: E722
logger.info(f"Could not benchmark at batch size = {batch_size}, prompt length = {prompt_length}")
except Exception as e:
logger.info(f"Could not benchmark at batch size = {batch_size}, prompt length = {prompt_length} - {e}")

filename = f"benchmark_{args.engine}_e2e_{datetime.datetime.now():%Y-%m-%d_%H:%M:%S}.csv"
save_results(all_csv_metrics, filename, args.generation_length)
Expand Down
48 changes: 23 additions & 25 deletions onnxruntime/python/tools/transformers/models/llama/llama_inputs.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,7 @@ def get_sample_inputs(
seq_len: int,
engine: str = "pt",
return_dict: bool = False,
use_position_ids: bool = True,
):
input_ids = torch.randint(low=0, high=config.vocab_size, size=(batch_size, seq_len), dtype=torch.int64)
attention_mask = torch.ones(batch_size, seq_len, dtype=torch.int64)
Expand All @@ -52,8 +53,10 @@ def get_sample_inputs(
inputs = {
"input_ids": input_ids,
"attention_mask": attention_mask,
"position_ids": position_ids,
}
if use_position_ids:
inputs["position_ids"] = position_ids

return inputs


Expand All @@ -72,6 +75,7 @@ def get_sample_with_past_kv_inputs(
engine: str = "pt",
return_dict: bool = False,
world_size: int = 1,
use_position_ids: bool = True,
):
input_ids = torch.randint(low=0, high=config.vocab_size, size=(batch_size, 1), dtype=torch.int64)
attention_mask = torch.ones(batch_size, past_seq_len + 1, dtype=torch.int64)
Expand All @@ -97,8 +101,9 @@ def get_sample_with_past_kv_inputs(
inputs = {
"input_ids": input_ids,
"attention_mask": attention_mask,
"position_ids": position_ids,
}
if use_position_ids:
inputs["position_ids"] = position_ids
if engine == "ort":
assert isinstance(past_kv, dict)
inputs.update(past_kv)
Expand Down Expand Up @@ -131,6 +136,7 @@ def get_merged_sample_with_past_kv_inputs(
engine: str = "pt",
return_dict: bool = False,
world_size: int = 1,
use_position_ids: bool = True,
):
input_ids = torch.randint(low=0, high=config.vocab_size, size=(batch_size, seq_len), dtype=torch.int64)
attention_mask = torch.ones(batch_size, past_seq_len + seq_len, dtype=torch.int64)
Expand All @@ -156,8 +162,9 @@ def get_merged_sample_with_past_kv_inputs(
inputs = {
"input_ids": input_ids,
"attention_mask": attention_mask,
"position_ids": position_ids,
}
if use_position_ids:
inputs["position_ids"] = position_ids
if engine == "ort":
assert isinstance(past_kv, dict)
inputs.update(past_kv)
Expand Down Expand Up @@ -382,27 +389,16 @@ def add_io_bindings_as_tensors(

for output in model.get_outputs():
name = output.name
if use_buffer_share and "present" in name:
# Bind KV cache outputs to KV cache inputs
v = inputs[name.replace("present", "past_key_values")]
io_binding.bind_output(
name=name,
device_type=v.device.type,
device_id=v.device.index,
element_type=np.float16,
shape=tuple(v.shape),
buffer_ptr=v.data_ptr(),
)
else:
v = outputs[name]
io_binding.bind_output(
name=name,
device_type=device.type,
device_id=0 if device.type == "cpu" else device.index,
element_type=(np.float16 if use_fp16 else np.float32),
shape=tuple(v.shape),
buffer_ptr=v.data_ptr(),
)
# Bind KV cache outputs to KV cache inputs
v = inputs[name.replace("present", "past_key_values")] if use_buffer_share and "present" in name else outputs[name]
io_binding.bind_output(
name=name,
device_type=device.type,
device_id=0 if device.type == "cpu" else device.index,
element_type=(np.float16 if use_fp16 else np.float32),
shape=tuple(v.shape),
buffer_ptr=v.data_ptr(),
)

return io_binding

Expand All @@ -417,6 +413,7 @@ def get_initial_inputs_and_outputs(
use_fp16: bool,
use_buffer_share: bool,
engine: str,
use_position_ids: bool = True,
):
tokenizer.pad_token = tokenizer.eos_token
encodings_dict = tokenizer.batch_encode_plus(prompt, padding=True)
Expand Down Expand Up @@ -452,8 +449,9 @@ def get_initial_inputs_and_outputs(
inputs = {
"input_ids": input_ids.contiguous() if engine == "ort" else input_ids,
"attention_mask": attention_mask.contiguous() if engine == "ort" else attention_mask,
"position_ids": position_ids.contiguous() if engine == "ort" else position_ids,
}
if use_position_ids:
inputs["position_ids"] = position_ids.contiguous() if engine == "ort" else position_ids
if engine != "ort":
inputs["past_key_values"] = []

Expand Down
Loading