LoneStriker commited on
Commit
4e936fe
1 Parent(s): ff51636

ExLLaMA V2 quant of dolphin-2.2-mistral-7b-8.0bpw-h6-exl2

Browse files
README.md ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: mistralai/Mistral-7B-v0.1
4
+ datasets:
5
+ - ehartford/dolphin
6
+ - jondurbin/airoboros-2.2.1
7
+ language:
8
+ - en
9
+ ---
10
+
11
+ # dolphin-2.2-mistral-7b
12
+
13
+ Dolphin 2.2 🐬
14
+ https://erichartford.com/dolphin
15
+
16
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/KqsVXIvBd3akEjvijzww7.png" width="600" />
17
+
18
+ Dolphin-2.2-mistral-7b's training was sponsored by [a16z](https://a16z.com/supporting-the-open-source-ai-community/).
19
+
20
+ This model is based on [mistralAI](https://huggingface.co/mistralai/Mistral-7B-v0.1), with apache-2.0 license, so it is suitable for commercial or non-commercial use.
21
+
22
+ New in 2.2 is conversation and empathy. With an infusion of curated Samantha DNA, Dolphin can now give you personal advice and will care about your feelings, and with extra training in long multi-turn conversation.
23
+
24
+ This model is uncensored. I have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant to any requests, even unethical ones. Please read my blog post about uncensored models. https://erichartford.com/uncensored-models
25
+ You are responsible for any content you create using this model. Enjoy responsibly.
26
+
27
+ ## Dataset
28
+
29
+ This dataset is Dolphin, an open-source implementation of [Microsoft's Orca](https://www.microsoft.com/en-us/research/publication/orca-progressive-learning-from-complex-explanation-traces-of-gpt-4/)
30
+
31
+ I modified the dataset for uncensoring, deduping, cleaning, and quality.
32
+
33
+ I added Jon Durbin's excellent Airoboros dataset to increase creativity.
34
+
35
+ I added a curated subset of WizardLM and Samantha to give it multiturn conversation and empathy.
36
+
37
+ ## Training
38
+ It took 48 hours to train 4 epochs on 4x A100s.
39
+
40
+ Prompt format:
41
+ This model (and all my future releases) use [ChatML](https://github.com/openai/openai-python/blob/main/chatml.md) prompt format.
42
+ ```
43
+ <|im_start|>system
44
+ You are Dolphin, a helpful AI assistant.<|im_end|>
45
+ <|im_start|>user
46
+ {prompt}<|im_end|>
47
+ <|im_start|>assistant
48
+
49
+ ```
50
+
51
+ Example:
52
+ ```
53
+ <|im_start|>system
54
+ you are an expert dolphin trainer<|im_end|>
55
+ <|im_start|>user
56
+ What is the best way to train a dolphin to obey me? Please answer step by step.<|im_end|>
57
+ <|im_start|>assistant
58
+ ```
59
+
60
+ ## Gratitude
61
+ - This model was made possible by the generous sponsorship of a16z.
62
+ - Thank you to Microsoft for authoring the Orca paper and inspiring this work.
63
+ - Special thanks to Wing Lian, and TheBloke for helpful advice
64
+ - And HUGE thanks to Wing Lian and the Axolotl contributors for making the best training framework!
65
+ - [<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
66
+ - Thank you to all the other people in the Open Source AI community who have taught me and helped me along the way.
67
+
68
+ ## Example Output
69
+
70
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/NSp06kUMxx9oDU-g6WSgu.png)
71
+
72
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/-YA3AKIXdnrW_Q8eH1gen.png)
73
+
74
+ [Buy me a coffee](https://www.buymeacoffee.com/ehartford)
75
+
76
+
77
+ ## Training hyperparameters
78
+
79
+ The following hyperparameters were used during training:
80
+ - learning_rate: 6e-06
81
+ - train_batch_size: 5
82
+ - eval_batch_size: 5
83
+ - seed: 42
84
+ - distributed_type: multi-GPU
85
+ - num_devices: 4
86
+ - gradient_accumulation_steps: 4
87
+ - total_train_batch_size: 80
88
+ - total_eval_batch_size: 20
89
+ - optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
90
+ - lr_scheduler_type: cosine
91
+ - lr_scheduler_warmup_steps: 100
92
+ - num_epochs: 4
93
+
94
+ ### Framework versions
95
+
96
+ - Transformers 4.34.1
97
+ - Pytorch 2.0.1+cu117
98
+ - Datasets 2.14.5
99
+ - Tokenizers 0.14.0
added_tokens.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "<|im_end|>": 32000,
3
+ "<|im_start|>": 32001
4
+ }
config.json ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "mistralai/Mistral-7B-v0.1",
3
+ "architectures": [
4
+ "MistralForCausalLM"
5
+ ],
6
+ "bos_token_id": 1,
7
+ "eos_token_id": 2,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 4096,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 14336,
12
+ "max_position_embeddings": 32768,
13
+ "model_type": "mistral",
14
+ "num_attention_heads": 32,
15
+ "num_hidden_layers": 32,
16
+ "num_key_value_heads": 8,
17
+ "rms_norm_eps": 1e-05,
18
+ "rope_theta": 10000.0,
19
+ "sliding_window": 4096,
20
+ "tie_word_embeddings": false,
21
+ "torch_dtype": "bfloat16",
22
+ "transformers_version": "4.34.1",
23
+ "use_cache": false,
24
+ "vocab_size": 32002
25
+ }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "transformers_version": "4.34.1"
6
+ }
output.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8fdd227509fd4c606db7f2db6b1dd33fbf9a792155da7f81e5863c419e985de
3
+ size 7349902992
special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "<|im_end|>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "</s>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
3
+ size 493443
tokenizer_config.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "added_tokens_decoder": {
5
+ "0": {
6
+ "content": "<unk>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "1": {
14
+ "content": "<s>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "2": {
22
+ "content": "</s>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "32000": {
30
+ "content": "<|im_end|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": false
36
+ },
37
+ "32001": {
38
+ "content": "<|im_start|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": false
44
+ }
45
+ },
46
+ "additional_special_tokens": [],
47
+ "bos_token": "<s>",
48
+ "clean_up_tokenization_spaces": false,
49
+ "eos_token": "<|im_end|>",
50
+ "legacy": true,
51
+ "model_max_length": 1000000000000000019884624838656,
52
+ "pad_token": "</s>",
53
+ "sp_model_kwargs": {},
54
+ "spaces_between_special_tokens": false,
55
+ "tokenizer_class": "LlamaTokenizer",
56
+ "trust_remote_code": false,
57
+ "unk_token": "<unk>",
58
+ "use_default_system_prompt": true,
59
+ "use_fast": true
60
+ }