From 32a67532fd71d78a138a48040116f343c0ab2b94 Mon Sep 17 00:00:00 2001 From: Roi Tiefenbrunn Date: Wed, 7 Aug 2024 10:08:36 +0300 Subject: [PATCH 1/2] Simplify HQT config - Configure default behaviour as empty 'allowlist'/'blocklist' configuration (#326) --- .../act_maxabs_pow2_weights_pcs_opt_pow2_quant.json | 5 +---- .../text-generation/quantization_config/maxabs_measure.json | 3 --- .../quantization_config/maxabs_measure_include_outputs.json | 5 +---- .../text-generation/quantization_config/maxabs_quant.json | 5 +---- .../quantization_config/maxabs_quant_phi.json | 1 - .../quantization_config/unit_scale_quant.json | 3 --- 6 files changed, 3 insertions(+), 19 deletions(-) diff --git a/examples/text-generation/quantization_config/act_maxabs_pow2_weights_pcs_opt_pow2_quant.json b/examples/text-generation/quantization_config/act_maxabs_pow2_weights_pcs_opt_pow2_quant.json index 602a147baa..c7c2bd9621 100644 --- a/examples/text-generation/quantization_config/act_maxabs_pow2_weights_pcs_opt_pow2_quant.json +++ b/examples/text-generation/quantization_config/act_maxabs_pow2_weights_pcs_opt_pow2_quant.json @@ -3,8 +3,5 @@ "mode": "QUANTIZE", "observer": "maxabs", "scale_method": "ACT_MAXABS_POW2_WEIGHTS_PCS_OPT_POW2", - "allowlist": {"types": [], "names": []}, - "blocklist": {"types": [], "names": []}, - "dump_stats_path": "./hqt_output/measure", - "dump_stats_xlsx_path": "./hqt_output/measure/fp8stats.xlsx" + "dump_stats_path": "./hqt_output/measure" } diff --git a/examples/text-generation/quantization_config/maxabs_measure.json b/examples/text-generation/quantization_config/maxabs_measure.json index 3645fe743a..f586a79c88 100644 --- a/examples/text-generation/quantization_config/maxabs_measure.json +++ b/examples/text-generation/quantization_config/maxabs_measure.json @@ -2,8 +2,5 @@ "method": "HOOKS", "mode": "MEASURE", "observer": "maxabs", - "allowlist": {"types": [], "names": []}, - "blocklist": {"types": [], "names": []}, "dump_stats_path": "./hqt_output/measure", - "dump_stats_xlsx_path": "./hqt_output/measure/fp8stats.xlsx" } \ No newline at end of file diff --git a/examples/text-generation/quantization_config/maxabs_measure_include_outputs.json b/examples/text-generation/quantization_config/maxabs_measure_include_outputs.json index 6de845a54d..230884c3a1 100644 --- a/examples/text-generation/quantization_config/maxabs_measure_include_outputs.json +++ b/examples/text-generation/quantization_config/maxabs_measure_include_outputs.json @@ -3,8 +3,5 @@ "mode": "MEASURE", "observer": "maxabs", "measure_exclude": "NONE", - "allowlist": {"types": [], "names": []}, - "blocklist": {"types": [], "names": []}, - "dump_stats_path": "./hqt_output/measure", - "dump_stats_xlsx_path": "./hqt_output/measure/fp8stats.xlsx" + "dump_stats_path": "./hqt_output/measure" } \ No newline at end of file diff --git a/examples/text-generation/quantization_config/maxabs_quant.json b/examples/text-generation/quantization_config/maxabs_quant.json index 02314a728e..ce8bae27a8 100644 --- a/examples/text-generation/quantization_config/maxabs_quant.json +++ b/examples/text-generation/quantization_config/maxabs_quant.json @@ -3,8 +3,5 @@ "mode": "QUANTIZE", "observer": "maxabs", "scale_method": "maxabs_hw", - "allowlist": {"types": [], "names": []}, - "blocklist": {"types": [], "names": []}, - "dump_stats_path": "./hqt_output/measure", - "dump_stats_xlsx_path": "./hqt_output/measure/fp8stats.xlsx" + "dump_stats_path": "./hqt_output/measure" } \ No newline at end of file diff --git a/examples/text-generation/quantization_config/maxabs_quant_phi.json b/examples/text-generation/quantization_config/maxabs_quant_phi.json index 8f13c2aa38..c49b982d0f 100644 --- a/examples/text-generation/quantization_config/maxabs_quant_phi.json +++ b/examples/text-generation/quantization_config/maxabs_quant_phi.json @@ -3,7 +3,6 @@ "mode": "QUANTIZE", "observer": "maxabs", "scale_method": "maxabs_hw", - "allowlist": {"types": [], "names": []}, "blocklist": {"types": [], "names": [ "matmul_qk", "matmul_av", diff --git a/examples/text-generation/quantization_config/unit_scale_quant.json b/examples/text-generation/quantization_config/unit_scale_quant.json index caad4bb2a4..10474f26cf 100644 --- a/examples/text-generation/quantization_config/unit_scale_quant.json +++ b/examples/text-generation/quantization_config/unit_scale_quant.json @@ -3,8 +3,5 @@ "mode": "QUANTIZE", "observer": "maxabs", "scale_method": "unit_scale", - "allowlist": {"types": [], "names": []}, - "blocklist": {"types": [], "names": []}, "dump_stats_path": "./hqt_output/measure", - "dump_stats_xlsx_path": "./hqt_output/measure/fp8stats.xlsx" } From c1c0e3600a60db9bef5f005711a7682f991020f5 Mon Sep 17 00:00:00 2001 From: Roi Tiefenbrunn Date: Wed, 7 Aug 2024 15:37:33 +0300 Subject: [PATCH 2/2] [SW-195755] Fix trailing ',' in measure.json --- .../text-generation/quantization_config/maxabs_measure.json | 4 ++-- .../text-generation/quantization_config/maxabs_quant_phi.json | 3 +-- .../text-generation/quantization_config/unit_scale_quant.json | 2 +- 3 files changed, 4 insertions(+), 5 deletions(-) diff --git a/examples/text-generation/quantization_config/maxabs_measure.json b/examples/text-generation/quantization_config/maxabs_measure.json index f586a79c88..773bcc54b0 100644 --- a/examples/text-generation/quantization_config/maxabs_measure.json +++ b/examples/text-generation/quantization_config/maxabs_measure.json @@ -2,5 +2,5 @@ "method": "HOOKS", "mode": "MEASURE", "observer": "maxabs", - "dump_stats_path": "./hqt_output/measure", -} \ No newline at end of file + "dump_stats_path": "./hqt_output/measure" +} diff --git a/examples/text-generation/quantization_config/maxabs_quant_phi.json b/examples/text-generation/quantization_config/maxabs_quant_phi.json index c49b982d0f..e7c6b6ddd2 100644 --- a/examples/text-generation/quantization_config/maxabs_quant_phi.json +++ b/examples/text-generation/quantization_config/maxabs_quant_phi.json @@ -8,6 +8,5 @@ "matmul_av", "lm_head" ]}, - "dump_stats_path": "./hqt_output/measure", - "dump_stats_xlsx_path": "./hqt_output/measure/fp8stats.xlsx" + "dump_stats_path": "./hqt_output/measure" } diff --git a/examples/text-generation/quantization_config/unit_scale_quant.json b/examples/text-generation/quantization_config/unit_scale_quant.json index 10474f26cf..216cf27e68 100644 --- a/examples/text-generation/quantization_config/unit_scale_quant.json +++ b/examples/text-generation/quantization_config/unit_scale_quant.json @@ -3,5 +3,5 @@ "mode": "QUANTIZE", "observer": "maxabs", "scale_method": "unit_scale", - "dump_stats_path": "./hqt_output/measure", + "dump_stats_path": "./hqt_output/measure" }