File size: 7,590 Bytes
48e379e |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 |
---
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 4.0bpw
This is a 4.0bpw 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 | Instruct Template | Score |
|------------|-------------------|-------|
| 8.0 | Alpaca | 56.67 |
| 8.0 | ChatML | 47.28 |
| 8.0 | Command-R | 58.46 |
| 8.0 | Command-R-Plus | 58.49 |
| 7.0 | Alpaca | 57.5 |
| 7.0 | ChatML | 46.86 |
| 7.0 | Command-R | 57.29 |
| 7.0 | Command-R-Plus | 57.91 |
| 6.0 | Alpaca | 56.5 |
| 6.0 | ChatML | 48.61 |
| 6.0 | Command-R | 57.8 |
| 6.0 | Command-R-Plus | 58.64 |
| 5.0 | Alpaca | 54.64 |
| 5.0 | ChatML | 48.48 |
| 5.0 | Command-R | 57.14 |
| 5.0 | Command-R-Plus | 56.63 |
| 4.5 | Alpaca | 57.75 |
| 4.5 | ChatML | 48.1 |
| 4.5 | Command-R | 57.08 |
| 4.5 | Command-R-Plus | 56.7 |
| 4.0 | Alpaca | 53.41 |
| 4.0 | ChatML | 50.99 |
| 4.0 | Command-R | 57.46 |
| 4.0 | Command-R-Plus | 57.99 |
| 3.5 | Alpaca | 56.68 |
| 3.5 | ChatML | 52.72 |
| 3.5 | Command-R | 60.91 |
| 3.5 | Command-R-Plus | 60.91 |
| 3.0 | Alpaca | 36.45 |
| 3.0 | ChatML | 39.19 |
| 3.0 | Command-R | 49.17 |
| 3.0 | Command-R-Plus | 49.68 |
### RP Calibrated Quants
| Quant Size | Instruct Template | Score |
|------------|-------------------|-------|
| 8.0 | Alpaca | 56.23 |
| 8.0 | ChatML | 48.42 |
| 8.0 | Command-R | 58.41 |
| 8.0 | Command-R-Plus | 58.41 |
| 7.0 | Alpaca | 57.01 |
| 7.0 | ChatML | 48.47 |
| 7.0 | Command-R | 57.85 |
| 7.0 | Command-R-Plus | 57.67 |
| 6.0 | Alpaca | 58.33 |
| 6.0 | ChatML | 50.93 |
| 6.0 | Command-R | 60.32 |
| 6.0 | Command-R-Plus | 59.83 |
| 5.0 | Alpaca | 55.28 |
| 5.0 | ChatML | 50.29 |
| 5.0 | Command-R | 58.96 |
| 5.0 | Command-R-Plus | 59.23 |
| 4.5 | Alpaca | 55.01 |
| 4.5 | ChatML | 46.63 |
| 4.5 | Command-R | 57.7 |
| 4.5 | Command-R-Plus | 59.24 |
| 4.0 | Alpaca | 49.76 |
| 4.0 | ChatML | 47.13 |
| 4.0 | Command-R | 54.76 |
| 4.0 | Command-R-Plus | 55.5 |
| 3.5 | Alpaca | 56.39 |
| 3.5 | ChatML | 52.98 |
| 3.5 | Command-R | 59.19 |
| 3.5 | Command-R-Plus | 58.32 |
| 3.0 | Alpaca | 50.36 |
| 3.0 | ChatML | 47.94 |
| 3.0 | Command-R | 54.89 |
| 3.0 | Command-R-Plus | 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
```
|