From ddda03b0c7cabf1d3e6c9048da22aa96ac295d32 Mon Sep 17 00:00:00 2001
From: Sara Adkins <sara@neuralmagic.com>
Date: Fri, 12 Apr 2024 20:11:52 +0000
Subject: [PATCH] add examples dir

---
 .../llama_1.1b/ex_sparseml_quantization.py    | 41 +++++++++++++++++++
 examples/llama_1.1b/example_quant_config.json | 39 ++++++++++++++++++
 examples/llama_1.1b/example_quant_recipe.yaml |  0
 makefile                                      |  2 +-
 4 files changed, 81 insertions(+), 1 deletion(-)
 create mode 100644 examples/llama_1.1b/ex_sparseml_quantization.py
 create mode 100644 examples/llama_1.1b/example_quant_config.json
 create mode 100644 examples/llama_1.1b/example_quant_recipe.yaml

diff --git a/examples/llama_1.1b/ex_sparseml_quantization.py b/examples/llama_1.1b/ex_sparseml_quantization.py
new file mode 100644
index 00000000..3c66f5e8
--- /dev/null
+++ b/examples/llama_1.1b/ex_sparseml_quantization.py
@@ -0,0 +1,41 @@
+# Copyright (c) 2021 - present / Neuralmagic, Inc. All Rights Reserved.
+# 
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+# 
+#    http://www.apache.org/licenses/LICENSE-2.0
+# 
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an "AS IS" BASIS,
+# 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.
+
+from sparseml.transformers import oneshot, SparseAutoModelForCausalLM
+
+dataset_name = "open_platypus"
+overwrite_output_dir = True
+splits = {"calibration": "train"}
+seed = 42
+output_dir = "./llama_1.1b_quant_mod_only"
+num_calibration_samples = 1024
+recipe = "example_quant_recipe.yaml"
+model_name = "TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T"
+max_seq_length = 1024
+pad_to_max_length = False
+
+model = SparseAutoModelForCausalLM.from_pretrained(model_name, device_map="cuda:0")
+
+oneshot(
+    model=model_name,
+    dataset=dataset_name,
+    output_dir=output_dir,
+    overwrite_output_dir=overwrite_output_dir,
+    splits = splits,
+    max_seq_length = max_seq_length,
+    seed=seed,
+    num_calibration_samples=num_calibration_samples,
+    recipe=recipe,
+    pad_to_max_length=pad_to_max_length
+)
\ No newline at end of file
diff --git a/examples/llama_1.1b/example_quant_config.json b/examples/llama_1.1b/example_quant_config.json
new file mode 100644
index 00000000..65d15740
--- /dev/null
+++ b/examples/llama_1.1b/example_quant_config.json
@@ -0,0 +1,39 @@
+{
+	"quant_method": "sparseml",
+	"format": "fakequant",
+	"quantization_status": "frozen",
+	"global_compression_ratio": null,
+	"config_groups": {
+        "group_1": {
+            "weights": {
+                "num_bits": 8,
+                "type": "int",
+                "symmetric": true,
+                "strategy": "tensor"
+            },
+            "input_activations": {
+                "num_bits": 8,
+                "type": "int",
+                "symmetric": true,
+                "strategy": "tensor"
+            },
+            "targets": ["Linear"]
+        },
+        "group_2": {
+            "weights": {
+                "num_bits": 8,
+                "type": "int",
+                "symmetric": false,
+                "strategy": "tensor"
+            },
+            "input_activations": null,
+            "targets": ["Embedding"]
+        }
+    },
+	"ignore": [
+        "LlamaRotaryEmbedding", "LlamaRMSNorm", "SiLUActivation", 
+        "model.layers.1.mlp.down_proj", "MatMulLeftInput_QK", "MatMulRightInput_QK",
+        "MatMulOutput_QK", "MatMulLeftInput_PV", "MatMulRightInput_PV", 
+        "MatMulOutput_PV"
+    ]
+}
\ No newline at end of file
diff --git a/examples/llama_1.1b/example_quant_recipe.yaml b/examples/llama_1.1b/example_quant_recipe.yaml
new file mode 100644
index 00000000..e69de29b
diff --git a/makefile b/makefile
index 02872127..435a37b9 100644
--- a/makefile
+++ b/makefile
@@ -1,6 +1,6 @@
 BUILDDIR := $(PWD)
 PYCHECKDIRS := src tests
-PYCHECKGLOBS := 'src/**/*.py' 'tests/**/*.py' 'utils/**/*.py' setup.py
+PYCHECKGLOBS := 'src/**/*.py' 'tests/**/*.py' 'utils/**/*.py' 'examples/**/*.py' setup.py
 # run checks on all files for the repo
 quality:
 	@echo "Running copyright checks";