File size: 6,889 Bytes
40797d1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4bafe39
 
 
 
 
 
 
 
 
 
40797d1
 
 
 
4bafe39
 
 
 
 
 
 
 
 
 
40797d1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
---
library_name: transformers
language:
- en
- fr
- de
- es
- it
- pt
- ja
- ko
- zh
- ar
license: cc-by-nc-4.0
tags:
- exl2
---

# c4ai-command-r-v01 - EXL2 6.0bpw

This is a 6.0bpw EXL2 quant of [CohereForAI/c4ai-command-r-v01](https://huggingface.co/CohereForAI/c4ai-command-r-v01)

Details about the model can be found at the above model page.


## EXL2 Version

These quants were made with exllamav2 version 0.0.18. Quants made on this version of EXL2 may not work on older versions of the exllamav2 library.

If you have problems loading these models, please update Text Generation WebUI to the latest version.

### RP Calibrated

The rpcal quants were made using data/PIPPA-cleaned/pippa_raw_fix.parquet for calibration.

## Perplexity Scoring

Below are the perplexity scores for the EXL2 models. A lower score is better. 
 
### Stock Quants

| Quant Level | Perplexity Score |
|-------------|------------------|
| 8.0 | 6.4436 |
| 7.0 | 6.4372 |
| 6.0 | 6.4391 |
| 5.0 | 6.4526 |
| 4.5 | 6.4629 |
| 4.0 | 6.5081 |
| 3.5 | 6.6301 |
| 3.0 | 6.7974 |

### RP Calibrated Quants

| Quant Level | Perplexity Score |
|-------------|------------------|
| 8.0 | 6.4331 |
| 7.0 | 6.4347 |
| 6.0 | 6.4356 |
| 5.0 | 6.4740 |
| 4.5 | 6.4875 |
| 4.0 | 6.5039 |
| 3.5 | 6.6928 |
| 3.0 | 6.8913 |

## EQ Bench

Here are the EQ Bench scores for the EXL2 quants using Alpaca, ChatML, Command-R and Command-R-Plus prompt templates. A higher score is better.

### Quants

| Quant Size | ChatML | Alpaca | Command-R | Command-R-Plus |
|------------|--------|--------|-----------|----------------|
| 8.0        | 47.28  | 56.67  | 58.46     | 58.49          |
| 7.0        | 46.86  | 57.5   | 57.29     | 57.91          |
| 6.0        | 48.61  | 56.5   | 57.8      | 58.64          |
| 5.0        | 48.48  | 54.64  | 57.14     | 56.63          |
| 4.5        | 48.1   | 57.75  | 57.08     | 56.7           |
| 4.0        | 50.99  | 53.41  | 57.46     | 57.99          |
| 3.5        | 52.72  | 56.68  | 60.91     | 60.91          |
| 3.0        | 39.19  | 36.45  | 49.17     | 49.68          |


### RP Calibrated Quants

| Quant Size | ChatML | Alpaca | Command-R | Command-R-Plus |
|------------|--------|--------|-----------|----------------|
| 8.0        | 48.42  | 56.23  | 58.41     | 58.41          |
| 7.0        | 48.47  | 57.01  | 57.85     | 57.67          |
| 6.0        | 50.93  | 58.33  | 60.32     | 59.83          |
| 5.0        | 50.29  | 55.28  | 58.96     | 59.23          |
| 4.5        | 46.63  | 55.01  | 57.7      | 59.24          |
| 4.0        | 47.13  | 49.76  | 54.76     | 55.5           |
| 3.5        | 52.98  | 56.39  | 59.19     | 58.32          |
| 3.0        | 47.94  | 50.36  | 54.89     | 53.61          |


### Command-R-Plus Template

This is the Command-R-Plus template yaml that was used in EQ bench(which uses Text Generation Web UI yaml templates). It adds BOS_TOKEN into the starter prompt.

_text-generation-webui/instruction-templates/Command-R-Plus.yaml_:
```yaml
instruction_template: |-
  {%- if messages[0]['role'] == 'system' -%}
      {%- set loop_messages = messages[1:] -%}
      {%- set system_message = messages[0]['content'] -%}
  {%- elif false == true -%}
      {%- set loop_messages = messages -%}
      {%- set system_message = 'You are Command-R, a brilliant, sophisticated, AI-assistant trained to assist human users by providing thorough responses. You are trained by Cohere.' -%}
  {%- else -%}
      {%- set loop_messages = messages -%}
      {%- set system_message = false -%}
  {%- endif -%}
  {%- if system_message != false -%}
      {{ '<BOS_TOKEN><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>' + system_message + '<|END_OF_TURN_TOKEN|>' }}
  {%- endif -%}
  {%- for message in loop_messages -%}
      {%- set content = message['content'] -%}
      {%- if message['role'] == 'user' -%}
          {{ '<|START_OF_TURN_TOKEN|><|USER_TOKEN|>' + content.strip() + '<|END_OF_TURN_TOKEN|>' }}
      {%- elif message['role'] == 'assistant' -%}
          {{ '<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>'  + content.strip() + '<|END_OF_TURN_TOKEN|>' }}
      {%- endif -%}
  {%- endfor -%}
  {%- if add_generation_prompt -%}
      {{ '<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>' }}
  {%- endif -%}

```

### Perplexity Script

This was the script used for perplexity testing.

```bash
#!/bin/bash

# Activate the conda environment
source ~/miniconda3/etc/profile.d/conda.sh
conda activate exllamav2

# Set the model name and bit size
MODEL_NAME="c4ai-command-r-v01"
BIT_PRECISIONS=(8.0 7.0 6.0 5.0 4.5 4.0 3.5 3.0)

# Print the markdown table header
echo "| Quant Level | Perplexity Score |"
echo "|-------------|------------------|"

for BIT_PRECISION in "${BIT_PRECISIONS[@]}"
do
  MODEL_DIR="models/${MODEL_NAME}_exl2_${BIT_PRECISION}bpw"
  # MODEL_DIR="models/${MODEL_NAME}_exl2_${BIT_PRECISION}bpw-rpcal"
  if [ -d "$MODEL_DIR" ]; then
    output=$(python test_inference.py -m "$MODEL_DIR" -gs 22,24 -ed data/wikitext/wikitext-2-v1.parquet)
    score=$(echo "$output" | grep -oP 'Evaluation perplexity: \K[\d.]+')
    echo "| $BIT_PRECISION | $score |"
  fi
done
```


## Quant Details

This is the script used for quantization.

```bash
#!/bin/bash

# Activate the conda environment
source ~/miniconda3/etc/profile.d/conda.sh
conda activate exllamav2

# Set the model name and bit size
MODEL_NAME="c4ai-command-r-v01"

# Define variables
MODEL_DIR="models/$MODEL_NAME"
OUTPUT_DIR="exl2_$MODEL_NAME"
MEASUREMENT_FILE="measurements/$MODEL_NAME.json"
# CALIBRATION_DATASET="data/PIPPA-cleaned/pippa_raw_fix.parquet"

# Create the measurement file if needed
if [ ! -f "$MEASUREMENT_FILE" ]; then
    echo "Creating $MEASUREMENT_FILE"
    # Create directories
    if [ -d "$OUTPUT_DIR" ]; then
        rm -r "$OUTPUT_DIR"
    fi
    mkdir "$OUTPUT_DIR"

    # python convert.py -i $MODEL_DIR -o $OUTPUT_DIR -nr -om $MEASUREMENT_FILE -c $CALIBRATION_DATASET
    python convert.py -i $MODEL_DIR -o $OUTPUT_DIR -nr -om $MEASUREMENT_FILE
fi

# Choose one of the below. Either create a single quant for testing or a batch of them.
# BIT_PRECISIONS=(5.0)
BIT_PRECISIONS=(8.0 7.0 6.0 5.0 4.5 4.0 3.5 3.0)

for BIT_PRECISION in "${BIT_PRECISIONS[@]}"
do
    CONVERTED_FOLDER="models/${MODEL_NAME}_exl2_${BIT_PRECISION}bpw"

    # If it doesn't already exist, make the quant
    if [ ! -d "$CONVERTED_FOLDER" ]; then

        echo "Creating $CONVERTED_FOLDER"

        # Create directories
        if [ -d "$OUTPUT_DIR" ]; then
            rm -r "$OUTPUT_DIR"
        fi
        mkdir "$OUTPUT_DIR"
        mkdir "$CONVERTED_FOLDER"
        
        # Run conversion commands  
        # python convert.py -i $MODEL_DIR -o $OUTPUT_DIR -nr -m $MEASUREMENT_FILE -b $BIT_PRECISION -c $CALIBRATION_DATASET -cf $CONVERTED_FOLDER
        python convert.py -i $MODEL_DIR -o $OUTPUT_DIR -nr -m $MEASUREMENT_FILE -b $BIT_PRECISION -cf $CONVERTED_FOLDER

    fi
done
```