File size: 5,710 Bytes
40ed737
2abfb9d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
40ed737
 
2abfb9d
 
df7e2c9
 
2abfb9d
 
d5fc5b4
 
2abfb9d
 
 
d5fc5b4
2abfb9d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d5fc5b4
 
2abfb9d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
base_model:
- CohereForAI/c4ai-command-r-plus
library_name: transformers
tags:
- mergekit
- merge
language:
- en
- fr
- de
- es
- it
- pt
- ja
- ko
- zh
- ar
pipeline_tag: text-generation
license: cc-by-nc-4.0
---
# Megac4ai-command-r-plus

[GGUF version is here.](https://huggingface.co/nitky/Megac4ai-command-r-plus-gguf)

๐Ÿšจ **This model is created using the special mergekit that supports c4ai-command-r-plus.**

This is a 160b frankenmerge model created by interleaving layers of [CohereForAI/c4ai-command-r-plus](https://huggingface.co/CohereForAI/c4ai-command-r-plus) with itself using mergekit.

## Output comparison
### Test Case Details

Condition: temperature=0.3

```
<|START_OF_TURN_TOKEN|><|USER_TOKEN|>ใƒ†ใ‚ฃใƒ : ใ‚„ใ‚ใ€่ชฟๅญใฏใฉใ†๏ผŸ
ใ‚ญใƒ : ใ„ใ‚ใ„ใ‚ใ‚„ใ‚ใ†ใจใ—ใฆใŸใ‚“ใ ใ‘ใฉใ€ใพใŸๅ…ˆๅปถใฐใ—ใซใ—ใกใ‚ƒใฃใŸใ‚ˆใ€‚
ใƒ†ใ‚ฃใƒ : ไฝ•ใ‚’ใ—ใ‚ˆใ†ใจใ—ใฆใ„ใŸใฎ๏ผŸ
ใ‚ญใƒ : ๅคงๅญฆใฎ่ชฒ้กŒใ ใ‚ˆใ€‚ใฉใ†ใซใ‚‚ใ‚„ใ‚‹ๆฐ—ใŒๅ‡บใชใใฆใญใ€‚
ใƒ†ใ‚ฃใƒ : ้›†ไธญใงใใชใ„ใชใ‚‰ใ€ใƒใƒขใƒ‰ใƒผใƒญใƒปใƒ†ใ‚ฏใƒ‹ใƒƒใ‚ฏใ‚’ใ™ใ‚‹ใจใ„ใ„ใ‚ˆใ€‚
ใ‚ญใƒ : ไฝ•ใใ‚Œ๏ผŸ
ใƒ†ใ‚ฃใƒ : 25ๅˆ†ไฝœๆฅญใ—ใฆใ€5ๅˆ†ไผ‘ๆ†ฉใ™ใ‚‹ใฎใ‚’็นฐใ‚Š่ฟ”ใ™ใ‚“ใ ใ‚ˆใ€‚ไธ€ๅ›žใ‚ใŸใ‚Šใฎไฝœๆฅญๆ™‚้–“ใŒ็Ÿญใใฆ้›†ไธญใงใใ‚‹ใ‚ˆใ€‚
ใ‚ญใƒ : ใ†ใƒผใ‚“ใ€้›†ไธญใฃใฆใ„ใ†ใ‚ใ‘ใ˜ใ‚ƒใชใ„ใ‚“ใ ใ‚ˆใญ
ใƒ†ใ‚ฃใƒ : ใ˜ใ‚ƒใ‚1ๆ—ฅใซ5ๅˆ†ใ ใ‘ใงใ„ใ„ใ‹ใ‚‰ๆœบใง่ชฒ้กŒใ‚’ใ™ใ‚‹ใฃใฆใ„ใ†ใฎใฏใฉใ†๏ผŸ
ใ‚ญใƒ : 5ๅˆ†ใ˜ใ‚ƒไฝ•ใ‚‚ใงใใชใใชใ„๏ผŸ
ใƒ†ใ‚ฃใƒ : ็Ÿญใ„ๆ™‚้–“ใงใ‚‚ใ„ใ„ใ‹ใ‚‰ๆœบใงไฝœๆฅญใ™ใ‚‹ใฃใฆใ„ใ†ใฎใŒใƒใ‚คใƒณใƒˆใชใ‚“ใ ใ‚ˆใ€‚ใ‚€ใ—ใ‚ใ‚‚ใฃใจใ‚„ใ‚ŠใŸใ„ใ€ใใ‚‰ใ„ใงๆญขใ‚ใฆใŠใใจๆฏŽๆ—ฅ็ถšใใ‚‚ใฃใจ้•ทใ„ๆ™‚้–“ใงใใ‚‹ใ‚ˆใ†ใซใชใ‚‹ใ‚ˆใ€‚
ใ‚ญใƒ : ็ขบใ‹ใซใใ‚Œใชใ‚‰ใงใใ‚‹ใ‹ใ‚‚ใ€‚ใ‚ใ‚ŠใŒใจใ†๏ผ
Q: ใ‚ญใƒ ใฏไฝ•ใ‚’ใ‚„ใฃใฆใฟใ‚ˆใ†ใจใ—ใฆใ„ใพใ™ใ‹๏ผŸใพใŸไฝ•ๆ•…ใใ‚Œใ‚’ใ‚„ใ‚ใ†ใจใ—ใฆใ„ใพใ™ใ‹๏ผŸ<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>
```

This task is included in [elyza/ELYZA-tasks-100](https://huggingface.co/datasets/elyza/ELYZA-tasks-100).

### Output Example

| Model | Output |
|-------|--------|
| CohereForAI/c4ai-command-r-plus | ใ‚ญใƒ ใฏๅคงๅญฆใฎ่ชฒ้กŒใ‚’ใ‚„ใ‚ใ†ใจใ—ใฆใ„ใพใ™ใ€‚ใ—ใ‹ใ—ใ€ใ‚„ใ‚‹ๆฐ—ใŒ่ตทใใšใ€ๅ…ˆๅปถใฐใ—ใซใ—ใฆใ—ใพใฃใŸใ‚ˆใ†ใงใ™ใ€‚ |
| nitky/Megac4ai-command-r-plus | ใ‚ญใƒ ใฏๅคงๅญฆใฎ่ชฒ้กŒใ‚’ใ‚„ใ‚ใ†ใจใ—ใฆใ„ใพใ™ใ€‚ใ‚„ใ‚‹ๆฐ—ใŒๅ‡บใšใ€้›†ไธญใ—ใฆไฝœๆฅญใ™ใ‚‹ใ“ใจใŒใงใใชใ„ใŸใ‚ใ€ใƒ†ใ‚ฃใƒ ใซใ‚ขใƒ‰ใƒใ‚คใ‚นใ‚’ๆฑ‚ใ‚ใฆใ„ใพใ™ใ€‚ใƒ†ใ‚ฃใƒ ใŒๆๆกˆใ—ใŸใƒใƒขใƒ‰ใƒผใƒญใƒปใƒ†ใ‚ฏใƒ‹ใƒƒใ‚ฏใ‚„ใ€1ๆ—ฅใซ5ๅˆ†ใ ใ‘ๆœบใง่ชฒ้กŒใ‚’ใ™ใ‚‹ใจใ„ใ†ๆ–นๆณ•ใ‚’่ฉฆใ™ใ“ใจใงใ€่ชฒ้กŒใซๅ–ใ‚Š็ต„ใ‚€็ฟ’ๆ…ฃใ‚’่บซใซใคใ‘ใ‚ˆใ†ใจใ—ใฆใ„ใพใ™ใ€‚ |

## Test environment

This model was tested using [text-generation-webui](https://github.com/oobabooga/text-generation-webui/tree/main). I use preset `min_p` and `Null preset` with temperature=0.3 for Generation.

## Usage

Please install `transformers` from the source repository that includes the necessary changes for this model.
```python
# pip install 'git+https://github.com/huggingface/transformers.git'
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "nitky/megac4ai-command-r-plus"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

# Format message with the command-r-plus chat template
messages = [{"role": "user", "content": "Hello, how are you?"}]
input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")
## <BOS_TOKEN><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello, how are you?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>

gen_tokens = model.generate(
    input_ids, 
    max_new_tokens=100, 
    do_sample=True, 
    temperature=0.3,
    )

gen_text = tokenizer.decode(gen_tokens[0])
print(gen_text)
```

### Quantized model through bitsandbytes, 4-bit precision

```python
# pip install 'git+https://github.com/huggingface/transformers.git' bitsandbytes accelerate
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig

bnb_config = BitsAndBytesConfig(load_in_4bit=True)

model_id = "nitky/megac4ai-command-r-plus"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, quantization_config=bnb_config)

# Format message with the command-r-plus chat template
messages = [{"role": "user", "content": "Hello, how are you?"}]
input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")
## <BOS_TOKEN><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello, how are you?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>

gen_tokens = model.generate(
    input_ids, 
    max_new_tokens=100, 
    do_sample=True, 
    temperature=0.3,
    )

gen_text = tokenizer.decode(gen_tokens[0])
print(gen_text)
```

## Merge Details
### Merge Method

This model was merged using the passthrough merge method.

### Models Merged

The following models were included in the merge:
* [CohereForAI/c4ai-command-r-plus](https://huggingface.co/CohereForAI/c4ai-command-r-plus)

### Configuration

The following YAML configuration was used to produce this model:

```yaml
dtype: float16
merge_method: passthrough
slices:
- sources:
  - layer_range: [0, 20]
    model: CohereForAI/c4ai-command-r-plus
- sources:
  - layer_range: [11, 31]
    model: CohereForAI/c4ai-command-r-plus
- sources:
  - layer_range: [22, 42]
    model: CohereForAI/c4ai-command-r-plus
- sources:
  - layer_range: [33, 53]
    model: CohereForAI/c4ai-command-r-plus
- sources:
  - layer_range: [44, 64]
    model: CohereForAI/c4ai-command-r-plus
```