Adding modes, graphs and metadata.

README.md

@@ -0,0 +1,102 @@
+---
+language: en
+thumbnail:
+license: mit
+tags:
+- question-answering
+- bert
+- bert-base
+datasets:
+- squad
+metrics:
+- squad
+widget:
+- text: "Where is the Eiffel Tower located?"
+  context: "The Eiffel Tower is a wrought-iron lattice tower on the Champ de Mars in Paris, France. It is named after the engineer Gustave Eiffel, whose company designed and built the tower."
+- text: "Who is Frederic Chopin?"
+  context: "Frédéric François Chopin, born Fryderyk Franciszek Chopin (1 March 1810 – 17 October 1849), was a Polish composer and virtuoso pianist of the Romantic era who wrote primarily for solo piano."
+---
+
+## BERT-base uncased model fine-tuned on SQuAD v1
+
+This model was created using the [nn_pruning](https://github.com/huggingface/nn_pruning) python library: the **linear layers contains 30.0%** of the original weights.
+
+This model **CANNOT be used without using nn_pruning `optimize_model`** function, as it uses NoNorms instead of LayerNorms and this is not currently supported by the Transformers library.
+
+
+It uses ReLUs instead of GeLUs as in the initial BERT network, to speedup inference.
+This does not need special handling, as it is supported by the Transformers library, and flagged in the model config by the ```"hidden_act": "relu"``` entry.
+
+
+The model contains **45.0%** of the original weights **overall** (the embeddings account for a significant part of the model, and they are not pruned by this method).
+
+With a simple resizing of the linear matrices it ran **2.01x as fast as BERT-base** on the evaluation.
+This is possible because the pruning method lead to structured matrices: to visualize them, hover below on the plot to see the non-zero/zero parts of each matrix.
+
+<div class="graph"><script src="/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/density_info.js" id="f93fdd05-71ca-4d3e-8bc5-a38cfcb56b7a"></script></div>
+
+In terms of accuracy, its **F1 is 89.19**, compared with 88.5 for BERT-base, a **F1 gain of 0.69**.
+
+## Fine-Pruning details
+This model was fine-tuned from the HuggingFace [BERT](https://www.aclweb.org/anthology/N19-1423/) base uncased checkpoint on [SQuAD1.1](https://rajpurkar.github.io/SQuAD-explorer), and distilled from the model [bert-large-uncased-whole-word-masking-finetuned-squad](https://huggingface.co/bert-large-uncased-whole-word-masking-finetuned-squad).
+This model is case-insensitive: it does not make a difference between english and English.
+
+A side-effect of the block pruning is that some of the attention heads are completely removed: 55 heads were removed on a total of 144 (38.2%).
+Here is a detailed view on how the remaining heads are distributed in the network after pruning.
+<div class="graph"><script src="/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/pruning_info.js" id="ad7eb4ee-5e94-4088-872e-d6d32e758312"></script></div>
+
+## Details of the SQuAD1.1 dataset
+
+| Dataset  | Split | # samples |
+| -------- | ----- | --------- |
+| SQuAD1.1 | train | 90.6K      |
+| SQuAD1.1 | eval  | 11.1k     |
+
+### Fine-tuning
+- Python: `3.8.5`
+
+- Machine specs:
+
+```CPU: Intel(R) Core(TM) i7-6700K CPU
+Memory: 64 GiB
+GPUs: 1 GeForce GTX 3090, with 24GiB memory
+GPU driver: 455.23.05, CUDA: 11.1
+```
+
+### Results
+
+**Pytorch model file size**: `374M` (original BERT: `438M`)
+
+| Metric | # Value   | # Original ([Table 2](https://www.aclweb.org/anthology/N19-1423.pdf))| Variation |
+| ------ | --------- | --------- | --------- |
+| **EM** | **82.21** | **80.8** | **+1.41**|
+| **F1** | **89.19** | **88.5** | **+0.69**|
+
+## Example Usage
+Install nn_pruning: it contains the optimization script, which just pack the linear layers into smaller ones by removing empty rows/columns.
+
+`pip install nn_pruning`
+
+Then you can use the `transformers library` almost as usual: you just have to call `optimize_model` when the pipeline has loaded.
+
+```python
+from transformers import pipeline
+from nn_pruning.inference_model_patcher import optimize_model
+
+qa_pipeline = pipeline(
+    "question-answering",
+    model="madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1",
+    tokenizer="madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1"
+)
+
+print("BERT-base parameters: 110M")
+print(f"Parameters count (includes head pruning)={int(qa_pipeline.model.num_parameters() / 1E6)}M")
+qa_pipeline.model = optimize_model(qa_pipeline.model, "dense")
+
+print(f"Parameters count after optimization={int(qa_pipeline.model.num_parameters() / 1E6)}M")
+predictions = qa_pipeline({
+    'context': "Frédéric François Chopin, born Fryderyk Franciszek Chopin (1 March 1810 – 17 October 1849), was a Polish composer and virtuoso pianist of the Romantic era who wrote primarily for solo piano.",
+    'question': "Who is Frederic Chopin?",
+})
+print("Predictions", predictions)
+```

config.json

@@ -0,0 +1,104 @@
+{
+  "_name_or_path": "/tmp/tmpcklouvey",
+  "architectures": [
+    "BertForQuestionAnswering"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "gradient_checkpointing": false,
+  "hidden_act": "relu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "layer_norm_type": "no_norm",
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "pruned_heads": {
+    "0": [
+      0,
+      2,
+      4,
+      5,
+      6
+    ],
+    "1": [
+      0,
+      2,
+      3,
+      5,
+      6,
+      7,
+      8
+    ],
+    "2": [
+      4,
+      7,
+      8
+    ],
+    "3": [
+      2,
+      4,
+      6
+    ],
+    "4": [
+      1,
+      2
+    ],
+    "5": [
+      1,
+      2,
+      6,
+      7,
+      11
+    ],
+    "6": [
+      2,
+      3,
+      10
+    ],
+    "7": [
+      1,
+      3,
+      6,
+      7,
+      11
+    ],
+    "8": [
+      0,
+      3,
+      4
+    ],
+    "9": [
+      1,
+      4,
+      5,
+      7,
+      9,
+      10
+    ],
+    "10": [
+      1,
+      2,
+      4,
+      5,
+      6,
+      7,
+      8
+    ],
+    "11": [
+      0,
+      5,
+      7,
+      8,
+      10,
+      11
+    ]
+  },
+  "type_vocab_size": 2,
+  "vocab_size": 30522
+}

model_card/density_info.js

@@ -0,0 +1,174 @@
+(function() {
+  var fn = function() {
+    
+    (function(root) {
+      function now() {
+        return new Date();
+      }
+    
+      var force = false;
+    
+      if (typeof root._bokeh_onload_callbacks === "undefined" || force === true) {
+        root._bokeh_onload_callbacks = [];
+        root._bokeh_is_loading = undefined;
+      }
+    
+      
+      
+    
+      var element = document.getElementById("f93fdd05-71ca-4d3e-8bc5-a38cfcb56b7a");
+        if (element == null) {
+          console.warn("Bokeh: autoload.js configured with elementid 'f93fdd05-71ca-4d3e-8bc5-a38cfcb56b7a' but no matching script tag was found.")
+        }
+      
+    
+      function run_callbacks() {
+        try {
+          root._bokeh_onload_callbacks.forEach(function(callback) {
+            if (callback != null)
+              callback();
+          });
+        } finally {
+          delete root._bokeh_onload_callbacks
+        }
+        console.debug("Bokeh: all callbacks have finished");
+      }
+    
+      function load_libs(css_urls, js_urls, callback) {
+        if (css_urls == null) css_urls = [];
+        if (js_urls == null) js_urls = [];
+    
+        root._bokeh_onload_callbacks.push(callback);
+        if (root._bokeh_is_loading > 0) {
+          console.debug("Bokeh: BokehJS is being loaded, scheduling callback at", now());
+          return null;
+        }
+        if (js_urls == null || js_urls.length === 0) {
+          run_callbacks();
+          return null;
+        }
+        console.debug("Bokeh: BokehJS not loaded, scheduling load and callback at", now());
+        root._bokeh_is_loading = css_urls.length + js_urls.length;
+    
+        function on_load() {
+          root._bokeh_is_loading--;
+          if (root._bokeh_is_loading === 0) {
+            console.debug("Bokeh: all BokehJS libraries/stylesheets loaded");
+            run_callbacks()
+          }
+        }
+    
+        function on_error() {
+          console.error("failed to load " + url);
+        }
+    
+        for (var i = 0; i < css_urls.length; i++) {
+          var url = css_urls[i];
+          const element = document.createElement("link");
+          element.onload = on_load;
+          element.onerror = on_error;
+          element.rel = "stylesheet";
+          element.type = "text/css";
+          element.href = url;
+          console.debug("Bokeh: injecting link tag for BokehJS stylesheet: ", url);
+          document.body.appendChild(element);
+        }
+    
+        const hashes = {"https://cdn.bokeh.org/bokeh/release/bokeh-2.2.3.min.js": "T2yuo9Oe71Cz/I4X9Ac5+gpEa5a8PpJCDlqKYO0CfAuEszu1JrXLl8YugMqYe3sM", "https://cdn.bokeh.org/bokeh/release/bokeh-widgets-2.2.3.min.js": "98GDGJ0kOMCUMUePhksaQ/GYgB3+NH9h996V88sh3aOiUNX3N+fLXAtry6xctSZ6", "https://cdn.bokeh.org/bokeh/release/bokeh-tables-2.2.3.min.js": "89bArO+nlbP3sgakeHjCo1JYxYR5wufVgA3IbUvDY+K7w4zyxJqssu7wVnfeKCq8"};
+    
+        for (var i = 0; i < js_urls.length; i++) {
+          var url = js_urls[i];
+          var element = document.createElement('script');
+          element.onload = on_load;
+          element.onerror = on_error;
+          element.async = false;
+          element.src = url;
+          if (url in hashes) {
+            element.crossOrigin = "anonymous";
+            element.integrity = "sha384-" + hashes[url];
+          }
+          console.debug("Bokeh: injecting script tag for BokehJS library: ", url);
+          document.head.appendChild(element);
+        }
+      };
+    
+      function inject_raw_css(css) {
+        const element = document.createElement("style");
+        element.appendChild(document.createTextNode(css));
+        document.body.appendChild(element);
+      }
+    
+      
+      var js_urls = ["https://cdn.bokeh.org/bokeh/release/bokeh-2.2.3.min.js", "https://cdn.bokeh.org/bokeh/release/bokeh-widgets-2.2.3.min.js", "https://cdn.bokeh.org/bokeh/release/bokeh-tables-2.2.3.min.js"];
+      var css_urls = [];
+      
+    
+      var inline_js = [
+        function(Bokeh) {
+          Bokeh.set_log_level("info");
+        },
+        
+        function(Bokeh) {
+          (function() {
+            var fn = function() {
+              Bokeh.safely(function() {
+                (function(root) {
+                  function embed_document(root) {
+                    
+                  var docs_json = '{"c4d44eda-2785-430d-b26f-a12b834e83c2":{"roots":{"references":[{"attributes":{},"id":"1154","type":"UnionRenderers"},{"attributes":{},"id":"1152","type":"UnionRenderers"},{"attributes":{},"id":"1155","type":"Selection"},{"attributes":{"fill_color":{"value":"#aa69f7"},"line_color":{"value":"#aa69f7"},"top":{"field":"height"},"width":{"value":0.125},"x":{"field":"x"}},"id":"1136","type":"VBar"},{"attributes":{"data_source":{"id":"1116"},"glyph":{"id":"1118"},"hover_glyph":null,"muted_glyph":null,"name":"query","nonselection_glyph":{"id":"1119"},"selection_glyph":null,"view":{"id":"1121"}},"id":"1120","type":"GlyphRenderer"},{"attributes":{},"id":"1146","type":"BasicTickFormatter"},{"attributes":{"fill_alpha":{"value":0.1},"fill_color":{"value":"#20cb97"},"line_alpha":{"value":0.1},"line_color":{"value":"#20cb97"},"top":{"field":"height"},"width":{"value":0.125},"x":{"field":"x"}},"id":"1131","type":"VBar"},{"attributes":{"data_source":{"id":"1134"},"glyph":{"id":"1136"},"hover_glyph":null,"muted_glyph":null,"name":"fully connected","nonselection_glyph":{"id":"1137"},"selection_glyph":null,"view":{"id":"1139"}},"id":"1138","type":"GlyphRenderer"},{"attributes":{"source":{"id":"1116"}},"id":"1121","type":"CDSView"},{"attributes":{},"id":"1150","type":"UnionRenderers"},{"attributes":{"source":{"id":"1134"}},"id":"1139","type":"CDSView"},{"attributes":{"data":{"density":["58.3%","41.7%","75.0%","75.0%","83.3%","58.3%","75.0%","58.3%","75.0%","50.0%","41.7%","50.0%"],"height":[0.344064,0.24576,0.442368,0.442368,0.49152,0.344064,0.442368,0.344064,0.442368,0.294912,0.24576,0.294912],"img_height":["96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px"],"img_width":["96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px"],"name":["0.attention.value","1.attention.value","2.attention.value","3.attention.value","4.attention.value","5.attention.value","6.attention.value","7.attention.value","8.attention.value","9.attention.value","10.attention.value","11.attention.value"],"parameters":["0.34","0.25","0.44","0.44","0.49","0.34","0.44","0.34","0.44","0.29","0.25","0.29"],"url":["/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_0_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_1_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_2_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_3_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_4_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_5_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_6_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_7_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_8_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_9_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_10_attention_self_value.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_11_attention_self_value.png"],"x":[0.41666666666666663,1.4166666666666665,2.416666666666667,3.416666666666667,4.416666666666666,5.416666666666666,6.416666666666666,7.416666666666666,8.416666666666668,9.416666666666668,10.416666666666668,11.416666666666668]},"selected":{"id":"1155"},"selection_policy":{"id":"1154"}},"id":"1128","type":"ColumnDataSource"},{"attributes":{"active_drag":"auto","active_inspect":"auto","active_multi":null,"active_scroll":"auto","active_tap":"auto","tools":[{"id":"1094"}]},"id":"1114","type":"Toolbar"},{"attributes":{"axis":{"id":"1106"},"grid_line_color":null,"ticker":null},"id":"1109","type":"Grid"},{"attributes":{"axis_label":"Layer","formatter":{"id":"1146"},"minor_tick_line_color":null,"ticker":{"id":"1107"}},"id":"1106","type":"LinearAxis"},{"attributes":{"data_source":{"id":"1128"},"glyph":{"id":"1130"},"hover_glyph":null,"muted_glyph":null,"name":"value","nonselection_glyph":{"id":"1131"},"selection_glyph":null,"view":{"id":"1133"}},"id":"1132","type":"GlyphRenderer"},{"attributes":{},"id":"1151","type":"Selection"},{"attributes":{"axis_label":"Parameters (M)","formatter":{"id":"1148"},"minor_tick_line_color":null,"ticker":{"id":"1111"}},"id":"1110","type":"LinearAxis"},{"attributes":{"data":{"density":["58.3%","15.6%","15.6%","41.7%","17.3%","17.3%","75.0%","21.6%","21.6%","75.0%","20.4%","20.4%","83.3%","17.7%","17.7%","58.3%","18.2%","18.2%","75.0%","15.7%","15.7%","58.3%","11.9%","11.9%","75.0%","6.0%","6.0%","50.0%","4.3%","4.3%","41.7%","10.7%","10.7%","50.0%","15.4%","15.4%"],"height":[0.344064,0.367104,0.367104,0.24576,0.40704,0.40704,0.442368,0.51072,0.51072,0.442368,0.482304,0.482304,0.49152,0.417792,0.417792,0.344064,0.428544,0.428544,0.442368,0.370176,0.370176,0.344064,0.281856,0.281856,0.442368,0.141312,0.141312,0.294912,0.100608,0.100608,0.24576,0.252672,0.252672,0.294912,0.364032,0.364032],"img_height":["96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px"],"img_width":["96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px","96px","96px","384px"],"name":["0.attention.output","0.intermediate","0.output","1.attention.output","1.intermediate","1.output","2.attention.output","2.intermediate","2.output","3.attention.output","3.intermediate","3.output","4.attention.output","4.intermediate","4.output","5.attention.output","5.intermediate","5.output","6.attention.output","6.intermediate","6.output","7.attention.output","7.intermediate","7.output","8.attention.output","8.intermediate","8.output","9.attention.output","9.intermediate","9.output","10.attention.output","10.intermediate","10.output","11.attention.output","11.intermediate","11.output"],"parameters":["0.34","0.37","0.37","0.25","0.41","0.41","0.44","0.51","0.51","0.44","0.48","0.48","0.49","0.42","0.42","0.34","0.43","0.43","0.44","0.37","0.37","0.34","0.28","0.28","0.44","0.14","0.14","0.29","0.10","0.10","0.25","0.25","0.25","0.29","0.36","0.36"],"url":["/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_0_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_0_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_0_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_1_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_1_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_1_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_2_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_2_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_2_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_3_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_3_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_3_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_4_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_4_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_4_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_5_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_5_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_5_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_6_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_6_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_6_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_7_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_7_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_7_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_8_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_8_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_8_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_9_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_9_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_9_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_10_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_10_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_10_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_11_attention_output_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_11_intermediate_dense.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_11_output_dense.png"],"x":[0.5833333333333334,0.75,0.9166666666666667,1.5833333333333333,1.75,1.9166666666666665,2.5833333333333335,2.75,2.916666666666667,3.5833333333333335,3.75,3.916666666666667,4.583333333333333,4.75,4.916666666666666,5.583333333333333,5.75,5.916666666666666,6.583333333333333,6.75,6.916666666666666,7.583333333333333,7.75,7.916666666666666,8.583333333333334,8.75,8.916666666666668,9.583333333333334,9.75,9.916666666666668,10.583333333333334,10.75,10.916666666666668,11.583333333333334,11.75,11.916666666666668]},"selected":{"id":"1157"},"selection_policy":{"id":"1156"}},"id":"1134","type":"ColumnDataSource"},{"attributes":{"above":[{"id":"1140"}],"below":[{"id":"1106"}],"center":[{"id":"1109"},{"id":"1113"}],"left":[{"id":"1110"}],"outline_line_color":null,"plot_height":300,"plot_width":505,"renderers":[{"id":"1120"},{"id":"1126"},{"id":"1132"},{"id":"1138"}],"title":{"id":"1096"},"toolbar":{"id":"1114"},"x_range":{"id":"1098"},"x_scale":{"id":"1102"},"y_range":{"id":"1100"},"y_scale":{"id":"1104"}},"id":"1095","subtype":"Figure","type":"Plot"},{"attributes":{"label":{"value":"fully connected"},"renderers":[{"id":"1138"}]},"id":"1144","type":"LegendItem"},{"attributes":{"fill_alpha":{"value":0.1},"fill_color":{"value":"#aa69f7"},"line_alpha":{"value":0.1},"line_color":{"value":"#aa69f7"},"top":{"field":"height"},"width":{"value":0.125},"x":{"field":"x"}},"id":"1137","type":"VBar"},{"attributes":{"data_source":{"id":"1122"},"glyph":{"id":"1124"},"hover_glyph":null,"muted_glyph":null,"name":"key","nonselection_glyph":{"id":"1125"},"selection_glyph":null,"view":{"id":"1127"}},"id":"1126","type":"GlyphRenderer"},{"attributes":{"axis":{"id":"1110"},"dimension":1,"ticker":null},"id":"1113","type":"Grid"},{"attributes":{"fill_color":{"value":"#20cb97"},"line_color":{"value":"#20cb97"},"top":{"field":"height"},"width":{"value":0.125},"x":{"field":"x"}},"id":"1130","type":"VBar"},{"attributes":{"label":{"value":"query"},"renderers":[{"id":"1120"}]},"id":"1141","type":"LegendItem"},{"attributes":{"label":{"value":"value"},"renderers":[{"id":"1132"}]},"id":"1143","type":"LegendItem"},{"attributes":{"fill_alpha":{"value":0.1},"fill_color":{"value":"#6573f7"},"line_alpha":{"value":0.1},"line_color":{"value":"#6573f7"},"top":{"field":"height"},"width":{"value":0.125},"x":{"field":"x"}},"id":"1119","type":"VBar"},{"attributes":{"items":[{"id":"1141"},{"id":"1142"},{"id":"1143"},{"id":"1144"}],"location":[10,0],"orientation":"horizontal"},"id":"1140","type":"Legend"},{"attributes":{"callback":null,"tooltips":"\\n        &lt;div&gt;\\n            &lt;div style=\\"margin-bottom:10px\\"&gt;\\n                &lt;span style=\\"font-size: 15px;\\"&gt;&lt;b&gt;@name&lt;/b&gt;&lt;br/&gt;density=@density&lt;/span&gt;\\n            &lt;/div&gt;\\n            &lt;div&gt;            \\n                &lt;img\\n                    src=\\"@url\\" height=\\"@img_height\\" width=\\"@img_width\\" alt=\\"@url\\"\\n                    style=\\"float: left; margin: 0px 15px 15px 0px;\\"\\n                    border=\\"0\\"\\n                /&gt;\\n            &lt;/div&gt;\\n        &lt;/div&gt;\\n        "},"id":"1094","type":"HoverTool"},{"attributes":{},"id":"1107","type":"BasicTicker"},{"attributes":{},"id":"1156","type":"UnionRenderers"},{"attributes":{"text":"Transformer Layers"},"id":"1096","type":"Title"},{"attributes":{"label":{"value":"key"},"renderers":[{"id":"1126"}]},"id":"1142","type":"LegendItem"},{"attributes":{},"id":"1102","type":"LinearScale"},{"attributes":{},"id":"1104","type":"LinearScale"},{"attributes":{"data":{"density":["58.3%","41.7%","75.0%","75.0%","83.3%","58.3%","75.0%","58.3%","75.0%","50.0%","41.7%","50.0%"],"height":[0.344064,0.24576,0.442368,0.442368,0.49152,0.344064,0.442368,0.344064,0.442368,0.294912,0.24576,0.294912],"img_height":["96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px"],"img_width":["96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px"],"name":["0.attention.query","1.attention.query","2.attention.query","3.attention.query","4.attention.query","5.attention.query","6.attention.query","7.attention.query","8.attention.query","9.attention.query","10.attention.query","11.attention.query"],"parameters":["0.34","0.25","0.44","0.44","0.49","0.34","0.44","0.34","0.44","0.29","0.25","0.29"],"url":["/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_0_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_1_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_2_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_3_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_4_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_5_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_6_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_7_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_8_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_9_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_10_attention_self_query.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_11_attention_self_query.png"],"x":[0.08333333333333333,1.0833333333333333,2.0833333333333335,3.0833333333333335,4.083333333333333,5.083333333333333,6.083333333333333,7.083333333333333,8.083333333333334,9.083333333333334,10.083333333333334,11.083333333333334]},"selected":{"id":"1151"},"selection_policy":{"id":"1150"}},"id":"1116","type":"ColumnDataSource"},{"attributes":{"data":{"density":["58.3%","41.7%","75.0%","75.0%","83.3%","58.3%","75.0%","58.3%","75.0%","50.0%","41.7%","50.0%"],"height":[0.344064,0.24576,0.442368,0.442368,0.49152,0.344064,0.442368,0.344064,0.442368,0.294912,0.24576,0.294912],"img_height":["96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px"],"img_width":["96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px","96px"],"name":["0.attention.key","1.attention.key","2.attention.key","3.attention.key","4.attention.key","5.attention.key","6.attention.key","7.attention.key","8.attention.key","9.attention.key","10.attention.key","11.attention.key"],"parameters":["0.34","0.25","0.44","0.44","0.49","0.34","0.44","0.34","0.44","0.29","0.25","0.29"],"url":["/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_0_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_1_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_2_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_3_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_4_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_5_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_6_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_7_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_8_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_9_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_10_attention_self_key.png","/madlag/bert-base-uncased-squadv1-x2.01-f89.2-d30-hybrid-rewind-opt-v1/raw/main/model_card/images/layer_11_attention_self_key.png"],"x":[0.25,1.25,2.25,3.25,4.25,5.25,6.25,7.25,8.25,9.25,10.25,11.25]},"selected":{"id":"1153"},"selection_policy":{"id":"1152"}},"id":"1122","type":"ColumnDataSource"},{"attributes":{},"id":"1148","type":"BasicTickFormatter"},{"attributes":{"fill_color":{"value":"#6573f7"},"line_color":{"value":"#6573f7"},"top":{"field":"height"},"width":{"value":0.125},"x":{"field":"x"}},"id":"1118","type":"VBar"},{"attributes":{"fill_alpha":{"value":0.1},"fill_color":{"value":"#ed5642"},"line_alpha":{"value":0.1},"line_color":{"value":"#ed5642"},"top":{"field":"height"},"width":{"value":0.125},"x":{"field":"x"}},"id":"1125","type":"VBar"},{"attributes":{},"id":"1153","type":"Selection"},{"attributes":{},"id":"1157","type":"Selection"},{"attributes":{"start":0},"id":"1100","type":"DataRange1d"},{"attributes":{"fill_color":{"value":"#ed5642"},"line_color":{"value":"#ed5642"},"top":{"field":"height"},"width":{"value":0.125},"x":{"field":"x"}},"id":"1124","type":"VBar"},{"attributes":{},"id":"1111","type":"BasicTicker"},{"attributes":{"source":{"id":"1122"}},"id":"1127","type":"CDSView"},{"attributes":{},"id":"1098","type":"DataRange1d"},{"attributes":{"source":{"id":"1128"}},"id":"1133","type":"CDSView"}],"root_ids":["1095"]},"title":"Bokeh Application","version":"2.2.3"}}';
+                  var render_items = [{"docid":"c4d44eda-2785-430d-b26f-a12b834e83c2","root_ids":["1095"],"roots":{"1095":"f93fdd05-71ca-4d3e-8bc5-a38cfcb56b7a"}}];
+                  root.Bokeh.embed.embed_items(docs_json, render_items);
+                
+                  }
+                  if (root.Bokeh !== undefined) {
+                    embed_document(root);
+                  } else {
+                    var attempts = 0;
+                    var timer = setInterval(function(root) {
+                      if (root.Bokeh !== undefined) {
+                        clearInterval(timer);
+                        embed_document(root);
+                      } else {
+                        attempts++;
+                        if (attempts > 100) {
+                          clearInterval(timer);
+                          console.log("Bokeh: ERROR: Unable to run BokehJS code because BokehJS library is missing");
+                        }
+                      }
+                    }, 10, root)
+                  }
+                })(window);
+              });
+            };
+            if (document.readyState != "loading") fn();
+            else document.addEventListener("DOMContentLoaded", fn);
+          })();
+        },
+        function(Bokeh) {
+        
+        
+        }
+      ];
+    
+      function run_inline_js() {
+        
+        for (var i = 0; i < inline_js.length; i++) {
+          inline_js[i].call(root, root.Bokeh);
+        }
+        
+      }
+    
+      if (root._bokeh_is_loading === 0) {
+        console.debug("Bokeh: BokehJS loaded, going straight to plotting");
+        run_inline_js();
+      } else {
+        load_libs(css_urls, js_urls, function() {
+          console.debug("Bokeh: BokehJS plotting callback run at", now());
+          run_inline_js();
+        });
+      }
+    }(window));
+  };
+  if (document.readyState != "loading") fn();
+  else document.addEventListener("DOMContentLoaded", fn);
+})();