File size: 6,594 Bytes
352cc88
 
 
4ba7237
dc0317d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
352cc88
 
f27250b
95e08f9
023161c
 
dd7eb79
 
f27250b
dd7eb79
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
023161c
95e08f9
 
 
 
 
 
54d0bf8
c49e8ec
54d0bf8
 
 
023161c
 
 
92f3267
 
 
 
dc0317d
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
language:
- en
license: mit
datasets:
- anon8231489123/ShareGPT_Vicuna_unfiltered
model-index:
- name: yi6B_Vicuna
  results:
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: AI2 Reasoning Challenge (25-Shot)
      type: ai2_arc
      config: ARC-Challenge
      split: test
      args:
        num_few_shot: 25
    metrics:
    - type: acc_norm
      value: 46.16
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lorinma/yi6B_Vicuna
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: HellaSwag (10-Shot)
      type: hellaswag
      split: validation
      args:
        num_few_shot: 10
    metrics:
    - type: acc_norm
      value: 69.3
      name: normalized accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lorinma/yi6B_Vicuna
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: MMLU (5-Shot)
      type: cais/mmlu
      config: all
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 58.43
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lorinma/yi6B_Vicuna
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: TruthfulQA (0-shot)
      type: truthful_qa
      config: multiple_choice
      split: validation
      args:
        num_few_shot: 0
    metrics:
    - type: mc2
      value: 48.11
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lorinma/yi6B_Vicuna
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: Winogrande (5-shot)
      type: winogrande
      config: winogrande_xl
      split: validation
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 65.67
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lorinma/yi6B_Vicuna
      name: Open LLM Leaderboard
  - task:
      type: text-generation
      name: Text Generation
    dataset:
      name: GSM8k (5-shot)
      type: gsm8k
      config: main
      split: test
      args:
        num_few_shot: 5
    metrics:
    - type: acc
      value: 18.42
      name: accuracy
    source:
      url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=lorinma/yi6B_Vicuna
      name: Open LLM Leaderboard
---


**Bug**: Having a bit issue with the tokenizer, still figuring out...You can use the original Yi tokenizer configuratin.


Reproduce Vicuna, but based on yi-6B. The training data I used was ShareGPT_V3_unfiltered_cleaned_split_no_imsorry.json. 

The training framework I used https://github.com/shibing624/MedicalGPT , train shell:
```
CUDA_VISIBLE_DEVICES=0,1,2,3,5 torchrun --nproc_per_node 5 ../supervised_finetuning.py \
    --model_type auto \
    --model_name_or_path /data/llm/models/Pretrained/yi-6B/01ai/Yi-6B \
    --tokenizer_name_or_path /data/llm/models/Pretrained/yi-6B/01ai/Yi-6B \
    --train_file_dir ../data/finetune/vicuna/ \
    --per_device_train_batch_size 2\
    --do_train \
    --max_train_samples -1 \
    --num_train_epochs 3 \
    --learning_rate 2e-5 \
    --weight_decay 0. \
    --bf16 \
    --use_peft False \
    --logging_strategy steps \
    --logging_steps 10 \
    --save_strategy epoch \
    --save_total_limit 5 \
    --gradient_accumulation_steps 1 \
    --preprocessing_num_workers 8 \
    --output_dir ../outputs/20240106_yi6B_vicuna \
    --overwrite_output_dir \
    --ddp_timeout 30000 \
    --logging_first_step True \
    --torch_dtype bfloat16 \
    --device_map auto \
    --report_to tensorboard \
    --ddp_find_unused_parameters False \
    --gradient_checkpointing True \
    --cache_dir ./cache \
    --model_max_length 4096 \
    --deepspeed ../deepspeed_zero_stage2_config_no16.json \
    --template_name yi   
```

The training used 5*A800 for 3 epochs
```
***** train metrics *****
  epoch                    =                3.0
  train_loss               =             0.3785
  train_runtime            = 1 day, 10:01:13.95
  train_samples            =              93204
  train_samples_per_second =               2.24
  train_steps_per_second   =              0.224
```

Post-training inference is also using this repository:
```
CUDA_VISIBLE_DEVICES=4 python gradio_demo.py  --model_type auto --base_model /data/mn/shibing624/MedicalGPT-1.6.3-231215/outputs/20240106_yi6B_vicuna    --tokenizer_path /data/mn/shibing624/MedicalGPT-1.6.3-231215/outputs/20240106_yi6B_vicuna --template_name yi --gpus 4
CUDA_VISIBLE_DEVICES=6 python inference.py --model_type auto --base_model /data/mn/shibing624/MedicalGPT-1.6.3-231215/outputs/20240106_yi6B_vicuna      --template_name yi --gpus 6     --interactive --tokenizer_path /data/llm/models/Pretrained/yi-6B/01ai/Yi-6B
```

We can see from some preliminary results, the conversation is natural and informative (unsurprisingly).

![image/png](https://cdn-uploads.huggingface.co/production/uploads/6413d7be996b2e426f230fb7/WfQYyyLxtXA2KlePmIPQJ.png)

Also we observe the unfiltering seems to be working! **Heads up** some examples are unsafe and inappropriate, this is entirely for research purposes, to test how alignment-filtered SFT data affect LLM's final output.

![image/png](https://cdn-uploads.huggingface.co/production/uploads/6413d7be996b2e426f230fb7/pklSsljCRN34QuL2ZF2zU.png)

![image/png](https://cdn-uploads.huggingface.co/production/uploads/6413d7be996b2e426f230fb7/22pTSVkBCVlQ5N8A8JBkF.png)

**Update:** Evaluate on Open LLM Leaderboard:

![image/png](https://cdn-uploads.huggingface.co/production/uploads/6413d7be996b2e426f230fb7/Xp11HLQqwh0HMSJgpr19n.png)
# [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)
Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_lorinma__yi6B_Vicuna)

|             Metric              |Value|
|---------------------------------|----:|
|Avg.                             |51.02|
|AI2 Reasoning Challenge (25-Shot)|46.16|
|HellaSwag (10-Shot)              |69.30|
|MMLU (5-Shot)                    |58.43|
|TruthfulQA (0-shot)              |48.11|
|Winogrande (5-shot)              |65.67|
|GSM8k (5-shot)                   |18.42|