Update README.md

README.md

@@ -25,7 +25,7 @@ library_name: transformers
 - **Model Developers:** Neural Magic
 
 Quantized version of [ibm-granite/granite-3.1-2b-instruct](https://huggingface.co/ibm-granite/granite-3.1-2b-instruct).
-It achieves an average score of xxxx on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves xxxx.
+It achieves an average score of 61.68 on the [OpenLLM](https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard) benchmark (version 1), whereas the unquantized model achieves 61.98.
 
 ### Model Optimizations
 
@@ -74,7 +74,7 @@ python quantize.py --model_path ibm-granite/granite-3.1-2b-instruct --quant_path
 
 ```python
 from datasets import load_dataset
-from transformers import AutoTokenizer
+from transformers import AutoTokenizer, AutoModelForCausalLM
 from llmcompressor.modifiers.quantization import GPTQModifier
 from llmcompressor.modifiers.smoothquant import SmoothQuantModifier
 from llmcompressor.transformers import SparseAutoModelForCausalLM, oneshot, apply
@@ -90,7 +90,7 @@ parser.add_argument('--dampening_frac', type=float, default=0.1)
 parser.add_argument('--observer', type=str, default="minmax")
 args = parser.parse_args()
 
-model = SparseAutoModelForCausalLM.from_pretrained(
+model = AutoModelForCausalLM.from_pretrained(
     args.model_path,
     device_map="auto",
     torch_dtype="auto",
@@ -223,3 +223,190 @@ evalplus.evaluate \
 |-----------------------------------------|:---------------------------------:|:-------------------------------------------:|
 | HumanEval Pass@1                        | 53.40                             | 54.9                                      |
 
+
+## Inference Performance
+
+
+This model achieves up to 1.4x speedup in single-stream deployment and up to 1.1x speedup in multi-stream asynchronous deployment, depending on hardware and use-case scenario.
+The following performance benchmarks were conducted with [vLLM](https://docs.vllm.ai/en/latest/) version 0.6.6.post1, and [GuideLLM](https://github.com/neuralmagic/guidellm).
+
+### Single-stream performance (measured with vLLM version 0.6.6.post1)
+<table>
+  <tr>
+    <td></td>
+    <td></td>
+    <td></td>
+    <th style="text-align: center;" colspan="7" >Latency (s)</th>
+  </tr>
+  <tr>
+    <th>GPU class</th>
+    <th>Model</th>
+    <th>Speedup</th>
+    <th>Code Completion<br>prefill: 256 tokens<br>decode: 1024 tokens</th>
+    <th>Docstring Generation<br>prefill: 768 tokens<br>decode: 128 tokens</th>
+    <th>Code Fixing<br>prefill: 1024 tokens<br>decode: 1024 tokens</th>
+    <th>RAG<br>prefill: 1024 tokens<br>decode: 128 tokens</th>
+    <th>Instruction Following<br>prefill: 256 tokens<br>decode: 128 tokens</th>
+    <th>Multi-turn Chat<br>prefill: 512 tokens<br>decode: 256 tokens</th>
+    <th>Large Summarization<br>prefill: 4096 tokens<br>decode: 512 tokens</th>
+  </tr>
+  <tr>
+    <td style="vertical-align: middle;" rowspan="3" >A5000</td>
+    <td>granite-3.1-2b-instruct</td>
+    <td></td>
+    <td>10.9</td>
+    <td>1.4</td>
+    <td>11.0</td>
+    <td>1.5</td>
+    <td>1.4</td>
+    <td>2.8</td>
+    <td>6.1</td>
+  </tr>
+  <tr>
+    <td>granite-3.1-2b-instruct-quantized.w8a8<br>(this model)</td>
+    <td>1.37</td>
+    <td>7.9</td>
+    <td>1.0</td>
+    <td>8.0</td>
+    <td>1.1</td>
+    <td>1.0</td>
+    <td>2.0</td>
+    <td>4.7</td>
+  </tr>
+  <tr>
+    <td>granite-3.1-2b-instruct-quantized.w4a16</td>
+    <td>1.94</td>
+    <td>5.4</td>
+    <td>0.7</td>
+    <td>5.5</td>
+    <td>0.8</td>
+    <td>0.7</td>
+    <td>1.4</td>
+    <td>3.4</td>
+  </tr>
+  <tr>
+    <td style="vertical-align: middle;" rowspan="3" >A6000</td>
+    <td>granite-3.1-2b-instruct</td>
+    <td></td>
+    <td>9.8</td>
+    <td>1.3</td>
+    <td>10.0</td>
+    <td>1.3</td>
+    <td>1.3</td>
+    <td>2.6</td>
+    <td>5.4</td>
+  </tr>
+  <tr>
+    <td>granite-3.1-2b-instruct-quantized.w8a8<br>(this model)</td>
+    <td>1.31</td>
+    <td>7.8</td>
+    <td>1.0</td>
+    <td>7.6</td>
+    <td>1.0</td>
+    <td>0.9</td>
+    <td>1.9</td>
+    <td>4.5</td>
+  </tr>
+  <tr>
+    <td>granite-3.1-2b-instruct-quantized.w4a16</td>
+    <td>1.87</td>
+    <td>5.1</td>
+    <td>0.7</td>
+    <td>5.2</td>
+    <td>0.7</td>
+    <td>0.7</td>
+    <td>1.3</td>
+    <td>3.1</td>
+  </tr>
+</table>
+
+
+### Multi-stream asynchronous performance (measured with vLLM version 0.6.6.post1)
+<table>
+  <tr>
+    <td></td>
+    <td></td>
+    <td></td>
+    <th style="text-align: center;" colspan="7" >Maximum Throughput (Queries per Second)</th>
+  </tr>
+  <tr>
+    <th>GPU class</th>
+    <th>Model</th>
+    <th>Speedup</th>
+    <th>Code Completion<br>prefill: 256 tokens<br>decode: 1024 tokens</th>
+    <th>Docstring Generation<br>prefill: 768 tokens<br>decode: 128 tokens</th>
+    <th>Code Fixing<br>prefill: 1024 tokens<br>decode: 1024 tokens</th>
+    <th>RAG<br>prefill: 1024 tokens<br>decode: 128 tokens</th>
+    <th>Instruction Following<br>prefill: 256 tokens<br>decode: 128 tokens</th>
+    <th>Multi-turn Chat<br>prefill: 512 tokens<br>decode: 256 tokens</th>
+    <th>Large Summarization<br>prefill: 4096 tokens<br>decode: 512 tokens</th>
+  </tr>
+  <tr>
+    <td style="vertical-align: middle;" rowspan="3" >A5000</td>
+    <td>granite-3.1-2b-instruct</td>
+    <td></td>
+    <td>2.9</td>
+    <td>10.2</td>
+    <td>1.8</td>
+    <td>8.2</td>
+    <td>19.3</td>
+    <td>9.1</td>
+    <td>1.3</td>
+  </tr>
+  <tr>
+    <td>granite-3.1-2b-instruct-quantized.w8a8<br>(this model)</td>
+    <td>1.13</td>
+    <td>3.1</td>
+    <td>12.1</td>
+    <td>2.0</td>
+    <td>9.6</td>
+    <td>22.2</td>
+    <td>10.2</td>
+    <td>1.4</td>
+  </tr>
+  <tr>
+    <td>granite-3.1-2b-instruct-quantized.w4a16</td>
+    <td>0.98</td>
+    <td>2.8</td>
+    <td>10.0</td>
+    <td>1.8</td>
+    <td>8.1</td>
+    <td>18.6</td>
+    <td>9.0</td>
+    <td>1.2</td>
+  </tr>
+  <tr>
+    <td style="vertical-align: middle;" rowspan="3" >A6000</td>
+    <td>granite-3.1-2b-instruct</td>
+    <td></td>
+    <td>3.7</td>
+    <td>12.4</td>
+    <td>2.4</td>
+    <td>10.3</td>
+    <td>23.6</td>
+    <td>11.0</td>
+    <td>1.6</td>
+  </tr>
+  <tr>
+    <td>granite-3.1-2b-instruct-quantized.w8a8<br>(this model)</td>
+    <td>1.12</td>
+    <td>3.6</td>
+    <td>14.4</td>
+    <td>2.7</td>
+    <td>12.0</td>
+    <td>28.3</td>
+    <td>12.9</td>
+    <td>1.7</td>
+  </tr>
+  <tr>
+    <td>granite-3.1-2b-instruct-quantized.w4a16</td>
+    <td>0.95</td>
+    <td>3.7</td>
+    <td>11.4</td>
+    <td>2.5</td>
+    <td>9.8</td>
+    <td>22.1</td>
+    <td>10.4</td>
+    <td>1.4</td>
+  </tr>
+</table>