Refined QDQ recipes of BERT/CLIP/VIT for QC and AMD. #1797
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1 @@ | ||
| 3.12.9 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,123 @@ | ||
| { | ||
| "input_model": { | ||
| "type": "PytorchModel", | ||
| "model_path": "google-bert/bert-base-multilingual-cased", | ||
| "io_config": { | ||
| "input_names": [ "input_ids", "attention_mask", "token_type_ids" ], | ||
| "input_shapes": [ [ 1, 512 ], [ 1, 512 ], [ 1, 512 ] ], | ||
| "input_types": [ "int32", "int32", "int32" ], | ||
| "output_names": [ "logits" ] | ||
| }, | ||
| "model_loader": "load_bert_nsp_model", | ||
| "model_script": "google_bert_script.py" | ||
| }, | ||
| "passes": { | ||
| "conversion": { "type": "OnnxConversion", "target_opset": 20, "dynamic": true, "use_dynamo_exporter": false }, | ||
|
There was a problem hiding this comment. dynamic and use_dynamo_exporter options are not required |
||
| "to_fixed_shape": { | ||
| "type": "DynamicToFixedShape", | ||
| "dim_param": [ "batch_size", "sequence_length" ], | ||
| "dim_value": [ 1, 512 ] | ||
| }, | ||
| "surgery": { | ||
| "type": "GraphSurgeries", | ||
| "surgeries": [ | ||
| { "surgeon": "ReplaceAttentionMaskValue", "replacement": -100.0 }, | ||
| { "surgeon": "MatMulAddToGemm" } | ||
|
There was a problem hiding this comment. this works well for qc. not sure if it's recommended for amd. |
||
| ] | ||
| }, | ||
| "transformer_optimizer": { | ||
|
There was a problem hiding this comment. I think this might not be required? with opset 20, the model gets exported with layernorm operator and quant_preprocess in onnxstatic quantization performs gelu fusion automatically |
||
| "type": "OrtTransformersOptimization", | ||
| "model_type": "bert", | ||
| "opt_level": 1, | ||
| "optimization_options": { | ||
| "enable_gelu": true, | ||
| "enable_bias_gelu": false, | ||
| "enable_layer_norm": true, | ||
| "enable_skip_layer_norm": false, | ||
| "enable_bias_skip_layer_norm": false, | ||
| "enable_attention": false | ||
| } | ||
| }, | ||
| "quantization": { | ||
| "type": "OnnxStaticQuantization", | ||
| "data_config": "calib_data", | ||
| "quant_preprocess": true, | ||
| "activation_type": "uint16", | ||
| "precision": "uint8" | ||
| }, | ||
| "addmetadata": { | ||
| "type": "VitisAIAddMetaData", | ||
| "config_meta_data_keys": [ "architectures", "model_type" ], | ||
| "activation_type": "uint16", | ||
| "weight_type": "uint8", | ||
| "quant_type": "OnnxStaticQuantization" | ||
| } | ||
| }, | ||
| "data_configs": [ | ||
| { | ||
| "name": "calib_data", | ||
| "type": "HuggingfaceContainer", | ||
| "load_dataset_config": { "data_name": "glue", "subset": "mrpc", "split": "train[:12]" }, | ||
| "pre_process_data_config": { | ||
| "model_name": "google-bert/bert-base-multilingual-cased", | ||
| "input_cols": [ "sentence1", "sentence2" ], | ||
| "max_length": 512, | ||
| "padding": "max_length" | ||
| }, | ||
| "dataloader_config": { "batch_size": 1 } | ||
| }, | ||
| { | ||
| "name": "wiki_data", | ||
| "type": "HuggingfaceContainer", | ||
| "load_dataset_config": { | ||
| "type": "dataset_to_nsp_dataset", | ||
| "data_name": "wikitext", | ||
| "subset": "wikitext-2-raw-v1", | ||
| "split": "test", | ||
| "input_cols": [ "sentence1", "sentence2" ], | ||
| "label_col": "label" | ||
| }, | ||
| "pre_process_data_config": { | ||
| "model_name": "google-bert/bert-base-multilingual-cased", | ||
| "input_cols": [ "sentence1", "sentence2" ], | ||
| "label_col": "label", | ||
| "max_length": 512, | ||
| "padding": "max_length" | ||
| }, | ||
| "post_process_data_config": { "type": "bert_scl_post_process" }, | ||
| "dataloader_config": { "batch_size": 1 }, | ||
| "user_script": "google_bert_script.py", | ||
| "script_dir": "." | ||
| } | ||
| ], | ||
| "evaluators": { | ||
| "nsp_evaluator": { | ||
| "metrics": [ | ||
| { | ||
| "name": "nsp", | ||
| "type": "accuracy", | ||
| "backend": "huggingface_metrics", | ||
| "data_config": "wiki_data", | ||
| "sub_types": [ { "name": "accuracy", "priority": 1 }, { "name": "f1" } ] | ||
| }, | ||
| { "name": "latency", "type": "latency", "sub_types": [ { "name": "avg" } ] } | ||
| ] | ||
| }, | ||
| "performance": { | ||
| "metrics": [ | ||
| { | ||
| "name": "latency", | ||
| "type": "latency", | ||
| "sub_types": [ | ||
| { "name": "avg", "priority": 1, "metric_config": { "warmup_num": 20, "repeat_test_num": 100 } }, | ||
| { "name": "p90", "metric_config": { "warmup_num": 20, "repeat_test_num": 100 } } | ||
| ] | ||
| } | ||
| ] | ||
| } | ||
| }, | ||
| "clean_cache": true, | ||
| "clean_evaluation_cache": true, | ||
| "evaluate_input_model": false, | ||
| "output_dir": "models/google/bert_base_multilingual_cased" | ||
| } | ||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,109 @@ | ||
| { | ||
| "input_model": { | ||
| "type": "HfModel", | ||
| "model_path": "google-bert/bert-large-cased-whole-word-masking-finetuned-squad", | ||
| "task": "question-answering", | ||
| "io_config": { | ||
| "input_names": [ "input_ids", "attention_mask" ], | ||
| "input_shapes": [ [ 1, 512 ], [ 1, 512 ] ], | ||
| "input_types": [ "int32", "int32" ], | ||
| "output_names": [ "start_logits", "end_logits" ] | ||
| } | ||
| }, | ||
| "passes": { | ||
| "conversion": { "type": "OnnxConversion", "target_opset": 20, "dynamic": true, "use_dynamo_exporter": false }, | ||
| "to_fixed_shape": { | ||
| "type": "DynamicToFixedShape", | ||
| "dim_param": [ "batch_size", "sequence_length" ], | ||
| "dim_value": [ 1, 512 ] | ||
| }, | ||
| "surgery": { | ||
| "type": "GraphSurgeries", | ||
| "surgeries": [ | ||
| { "surgeon": "ReplaceAttentionMaskValue", "replacement": -100.0 }, | ||
| { "surgeon": "MatMulAddToGemm" } | ||
| ] | ||
| }, | ||
| "transformer_optimizer": { | ||
| "type": "OrtTransformersOptimization", | ||
| "model_type": "bert", | ||
| "opt_level": 1, | ||
| "optimization_options": { | ||
| "enable_gelu": true, | ||
| "enable_bias_gelu": false, | ||
| "enable_layer_norm": true, | ||
| "enable_skip_layer_norm": false, | ||
| "enable_bias_skip_layer_norm": false, | ||
| "enable_attention": false | ||
| } | ||
| }, | ||
| "quantization": { | ||
| "type": "OnnxStaticQuantization", | ||
| "data_config": "calib_data", | ||
| "quant_preprocess": true, | ||
| "activation_type": "uint16", | ||
| "precision": "uint8" | ||
| }, | ||
| "addmetadata": { | ||
| "type": "VitisAIAddMetaData", | ||
| "config_meta_data_keys": [ "architectures", "model_type" ], | ||
| "activation_type": "uint16", | ||
| "weight_type": "uint8", | ||
| "quant_type": "OnnxStaticQuantization" | ||
| } | ||
| }, | ||
| "data_configs": [ | ||
| { | ||
| "name": "calib_data", | ||
| "type": "HuggingfaceContainer", | ||
| "load_dataset_config": { "data_name": "squad", "split": "train[:12]" }, | ||
| "pre_process_data_config": { | ||
| "input_cols": [ "question", "context" ], | ||
| "label_col": "id", | ||
| "padding": "max_length", | ||
| "max_length": 512 | ||
| }, | ||
| "dataloader_config": { "batch_size": 1 }, | ||
| "user_script": "google_bert_script.py" | ||
| } | ||
| ], | ||
| "evaluators": { | ||
| "squad_evaluator": { | ||
| "metrics": [ | ||
| { | ||
| "name": "squad", | ||
| "type": "custom", | ||
| "sub_types": [ | ||
| { "name": "exact_match", "priority": 1, "higher_is_better": true }, | ||
| { "name": "f1", "higher_is_better": true } | ||
| ], | ||
| "user_config": { | ||
| "evaluate_func": "eval_squad", | ||
| "evaluate_func_kwargs": { | ||
| "model_name": "google-bert/bert-large-cased-whole-word-masking-finetuned-squad", | ||
| "dataset_config": { "data_name": "squad", "split": "validation" } | ||
| }, | ||
| "user_script": "google_bert_script.py" | ||
| } | ||
| }, | ||
| { "name": "latency", "type": "latency", "sub_types": [ { "name": "avg" } ] } | ||
| ] | ||
| }, | ||
| "performance": { | ||
| "metrics": [ | ||
| { | ||
| "name": "latency", | ||
| "type": "latency", | ||
| "sub_types": [ | ||
| { "name": "avg", "priority": 1, "metric_config": { "warmup_num": 20, "repeat_test_num": 100 } }, | ||
| { "name": "p90", "metric_config": { "warmup_num": 20, "repeat_test_num": 100 } } | ||
| ] | ||
| } | ||
| ] | ||
| } | ||
| }, | ||
| "clean_cache": true, | ||
| "clean_evaluation_cache": true, | ||
| "evaluate_input_model": false, | ||
| "output_dir": "models/google/bert_large_cased_qa" | ||
| } |
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
since these are all new files, are still keeping the old qdq config jsons?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
bumping this again. Are we keeping both the old and new qdq configs?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@tezheng
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We have already released parts of these in AITK and I created a pr #1893 to align the recipes. It is part of this pr: CLIP text + vision, bert and vit qdq.
Please take a look @jambayk CC @tezheng
I updated the previous qdq configs directly
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
For others, we could take time to merge if feel needed