docs: update conformer transducer/ctc results

Author: nglehuy

examples/models/ctc/conformer/results/sentencepiece/README.md

@@ -1,23 +1,7 @@
-**Table of Contents**
-- [LibriSpeech](#librispeech)
+- [\[English\] LibriSpeech](#english-librispeech)
   - [I. Small + SentencePiece 256](#i-small--sentencepiece-256)
-    - [Training](#training)
-      - [1. Epoch Loss](#1-epoch-loss)
-      - [2. Batch Loss](#2-batch-loss)
-      - [3. Learning Rate](#3-learning-rate)
-    - [Pretrained Model](#pretrained-model)
-    - [Results](#results)
-- [VietBud500](#vietbud500)
-  - [I. Small + SentencePiece 256](#i-small--sentencepiece-256-1)
-    - [Training](#training-1)
-      - [1. Epoch Loss](#1-epoch-loss-1)
-      - [2. Batch Loss](#2-batch-loss-1)
-      - [3. Learning Rate](#3-learning-rate-1)
-    - [Pretrained Model](#pretrained-model-1)
-    - [Results](#results-1)
 
-
-# LibriSpeech
+# [English] LibriSpeech
 
 ## I. Small + SentencePiece 256
 
@@ -30,75 +14,45 @@
 | Global Batch Size | 8 * 4 * 8 = 256 (as 4 TPUs, 8 Gradient Accumulation Steps) |
 | Max Epochs        | 450                                                        |
 
+**Config:**
 
-### Training
-
-#### 1. Epoch Loss
-
-![Epoch Loss](./figs/)
-
-#### 2. Batch Loss
-
-![Batch Loss](./figs/)
-
-#### 3. Learning Rate
-
-![Learning Rate](./figs/)
-
-### Pretrained Model
-
-[Link]()
-
-### Results
-
-
-```json
-[
-  {
-    "epoch": 157,
-    "test-clean": {
-    },
-    "test-other": {
-    }
-  }
-]
+```jinja2
+{% import "examples/datasets/librispeech/sentencepiece/sp.256.yml.j2" as decoder_config with context %}
+{{decoder_config}}
+{% import "examples/models/ctc/conformer/small.yml.j2" as config with context %}
+{{config}}
 ```
 
-# VietBud500
+**Results:**
 
-## I. Small + SentencePiece 256
+| Epoch | Dataset    | decoding | wer       | cer       | mer       | wil      | wip      |
+| :---- | :--------- | :------- | :-------- | :-------- | :-------- | :------- | :------- |
+| 170   | test-clean | greedy   | 0.0967171 | 0.031954  | 0.0958403 | 0.168307 | 0.831693 |
+| 170   | test-other | greedy   | 0.201612  | 0.0812955 | 0.197415  | 0.330207 | 0.669793 |
+
+<!--
+## II. Small + Streaming + SentencePiece 256
 
 | Category          | Description                                                |
 | :---------------- | :--------------------------------------------------------- |
-| Config            | [small.yml.j2](../../small.yml.j2)                         |
+| Config            | [small-streaming.yml.j2](../../small-streaming.yml.j2)     |
 | Tensorflow        | **2.18.0**                                                 |
 | Device            | Google Cloud TPUs v4-8                                     |
 | Mixed Precision   | strict                                                     |
 | Global Batch Size | 8 * 4 * 8 = 256 (as 4 TPUs, 8 Gradient Accumulation Steps) |
 | Max Epochs        | 450                                                        |
 
-### Training
-
-#### 1. Epoch Loss
-
-![Epoch Loss](./figs/)
-
-#### 2. Batch Loss
+**Config:**
 
-![Batch Loss](./figs/)
-
-#### 3. Learning Rate
-
-![Learning Rate](./figs/)
-
-### Pretrained Model
-
-[Link]()
-
-### Results
+```jinja2
+{% import "examples/datasets/librispeech/sentencepiece/sp.256.yml.j2" as decoder_config with context %}
+{{decoder_config}}
+{% import "examples/models/ctc/conformer/small-streaming.yml.j2" as config with context %}
+{{config}}
+```
 
-```json
-[
+**Results:**
 
-]
-```
+| Epoch | Dataset | decoding | wer  | cer  | mer  | wil  | wip  |
+| :---- | :------ | :------- | :--- | :--- | :--- | :--- | :--- |
+-->

examples/models/transducer/conformer/results/sentencepiece/README.md

@@ -1,12 +1,8 @@
-**Table of Contents**
 - [\[English\] LibriSpeech](#english-librispeech)
   - [I. Small + SentencePiece 1k](#i-small--sentencepiece-1k)
-    - [Config](#config)
-    - [Results](#results)
-- [\[Vietnamese\] VietBud500](#vietnamese-vietbud500)
   - [II. Small + Streaming + SentencePiece 1k](#ii-small--streaming--sentencepiece-1k)
-    - [Config](#config-1)
-    - [Results](#results-1)
+- [\[Vietnamese\] VietBud500](#vietnamese-vietbud500)
+  - [I. Small + Streaming + SentencePiece 1k](#i-small--streaming--sentencepiece-1k)
 
 <!-- ----------------------------------------------------- EN ------------------------------------------------------ -->
 
@@ -23,7 +19,7 @@
 | Global Batch Size | 4 * 4 * 8 = 128 (as 4 TPUs, 8 Gradient Accumulation Steps) |
 | Max Epochs        | 300                                                        |
 
-### Config
+**Config:**
 
 ```jinja2
 {% import "examples/datasets/librispeech/sentencepiece/sp.yml.j2" as decoder_config with context %}
@@ -32,18 +28,45 @@
 {{config}}
 ```
 
-### Results
+**Results:**
 
 | Epoch | Dataset    | decoding | wer      | cer      | mer      | wil      | wip      |
 | :---- | :--------- | :------- | :------- | :------- | :------- | :------- | :------- |
 | 157   | test-clean | greedy   | 0.062918 | 0.025361 | 0.062527 | 0.109992 | 0.890007 |
 | 157   | test-other | greedy   | 0.142616 | 0.066839 | 0.140610 | 0.239201 | 0.760798 |
 
+## II. Small + Streaming + SentencePiece 1k
+
+| Category          | Description                                                |
+| :---------------- | :--------------------------------------------------------- |
+| Config            | [small-streaming.yml.j2](../../small-streaming.yml.j2)     |
+| Tensorflow        | **2.18.0**                                                 |
+| Device            | Google Cloud TPUs v4-8                                     |
+| Mixed Precision   | strict                                                     |
+| Global Batch Size | 4 * 4 * 8 = 128 (as 4 TPUs, 8 Gradient Accumulation Steps) |
+| Max Epochs        | 300                                                        |
+
+**Config:**
+
+```jinja2
+{% import "examples/datasets/librispeech/sentencepiece/sp.yml.j2" as decoder_config with context %}
+{{decoder_config}}
+{% import "examples/models/transducer/conformer/small-streaming.yml.j2" as config with context %}
+{{config}}
+```
+
+**Results:**
+
+| Epoch | Dataset    | decoding | wer      | cer       | mer      | wil      | wip      |
+| :---- | :--------- | :------- | :------- | :-------- | :------- | :------- | :------- |
+| 45    | test-clean | greedy   | 0.110564 | 0.0460022 | 0.109064 | 0.186109 | 0.813891 |
+| 45    | test-other | greedy   | 0.267772 | 0.139369  | 0.260952 | 0.417361 | 0.582639 |
+
 <!-- ----------------------------------------------------- VN ------------------------------------------------------ -->
 
 # [Vietnamese] VietBud500
 
-## II. Small + Streaming + SentencePiece 1k
+## I. Small + Streaming + SentencePiece 1k
 
 | Category          | Description                                                |
 | :---------------- | :--------------------------------------------------------- |
@@ -54,7 +77,7 @@
 | Global Batch Size | 8 * 4 * 8 = 256 (as 4 TPUs, 8 Gradient Accumulation Steps) |
 | Max Epochs        | 300                                                        |
 
-### Config
+**Config:**
 
 ```jinja2
 {% import "examples/datasets/vietbud500/sentencepiece/sp.yml.j2" as decoder_config with context %}
@@ -63,13 +86,26 @@
 {{config}}
 ```
 
-### Results
-
-| Training      | Image                                                           |
-| :------------ | :-------------------------------------------------------------- |
-| Epoch Loss    | ![Epoch Loss](./figs/vietbud500-small-streaming-epoch-loss.svg) |
-| Batch Loss    | ![Batch Loss](./figs/vietbud500-small-streaming-batch-loss.svg) |
-| Learning Rate | ![Learning Rate](./figs/vietbud500-small-streaming-lr.svg)      |
+**Tensorboard:**
+
+<table>
+  <tr>
+    <td align="center">
+      <img src="./figs/vietbud500-small-streaming-epoch-loss.jpg" width="200px"><br>
+      <sub><strong>Epoch Loss</strong></sub>
+    </td>
+    <td align="center">
+      <img src="./figs/vietbud500-small-streaming-batch-loss.jpg" width="200px"><br>
+      <sub><strong>Batch Loss</strong></sub>
+    </td>
+    <td align="center">
+      <img src="./figs/vietbud500-small-streaming-lr.jpg " width="200px"><br>
+      <sub><strong>Learning Rate</strong></sub>
+    </td>
+  </tr>
+</table>
+
+**Results:**
 
 | Epoch | decoding | wer      | cer      | mer     | wil      | wip      |
 | :---- | :------- | :------- | :------- | :------ | :------- | :------- |