Hugging Face's logo

NurtureAI
/
zephyr-7b-beta-16k-GGUF

English
Generated from Trainer
Model card Files Community
2
perlthoughts commited on
Commit
40c895b
•
1 Parent(s): 9963bb0

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +238 -0
README.md ADDED
@@ -0,0 +1,238 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - generated_from_trainer
4
+ model-index:
5
+ - name: zephyr-7b-beta
6
+ results: []
7
+ license: mit
8
+ datasets:
9
+ - HuggingFaceH4/ultrachat_200k
10
+ - HuggingFaceH4/ultrafeedback_binarized
11
+ language:
12
+ - en
13
+ base_model: mistralai/Mistral-7B-v0.1
14
+ ---
15
+
16
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
17
+ should probably proofread and complete it, then remove this comment. -->
18
+
19
+ <img src="https://huggingface.co/HuggingFaceH4/zephyr-7b-alpha/resolve/main/thumbnail.png" alt="Zephyr Logo" width="800" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
20
+
21
+
22
+ # Model Card for Zephyr 7B β (context size extended to 16k)
23
+
24
+ # (Quantized GGUF Models)
25
+
26
+ Zephyr is a series of language models that are trained to act as helpful assistants. Zephyr-7B-β is the second model in the series, and is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) that was trained on on a mix of publicly available, synthetic datasets using [Direct Preference Optimization (DPO)](https://arxiv.org/abs/2305.18290). We found that removing the in-built alignment of these datasets boosted performance on [MT Bench](https://huggingface.co/spaces/lmsys/mt-bench) and made the model more helpful. However, this means that model is likely to generate problematic text when prompted to do so and should only be used for educational and research purposes. You can find more details in the [technical report](https://arxiv.org/abs/2310.16944).
27
+
28
+ # Original Model Card
29
+
30
+ ## Model description
31
+
32
+ - **Model type:** A 7B parameter GPT-like model fine-tuned on a mix of publicly available, synthetic datasets.
33
+ - **Language(s) (NLP):** Primarily English
34
+ - **License:** MIT
35
+ - **Finetuned from model:** [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1)
36
+
37
+ ### Model Sources
38
+
39
+ <!-- Provide the basic links for the model. -->
40
+
41
+ - **Repository:** https://github.com/huggingface/alignment-handbook
42
+ - **Demo:** https://huggingface.co/spaces/HuggingFaceH4/zephyr-chat
43
+ - **Chatbot Arena:** Evaluate Zephyr 7B against 10+ LLMs in the LMSYS arena: http://arena.lmsys.org
44
+
45
+ ## Performance
46
+
47
+ At the time of release, Zephyr-7B-β is the highest ranked 7B chat model on the [MT-Bench](https://huggingface.co/spaces/lmsys/mt-bench) and [AlpacaEval](https://tatsu-lab.github.io/alpaca_eval/) benchmarks:
48
+
49
+ | Model | Size | Alignment | MT-Bench (score) | AlpacaEval (win rate %) |
50
+ |-------------|-----|----|---------------|--------------|
51
+ | StableLM-Tuned-α | 7B| dSFT |2.75| -|
52
+ | MPT-Chat | 7B |dSFT |5.42| -|
53
+ | Xwin-LMv0.1 | 7B| dPPO| 6.19| 87.83|
54
+ | Mistral-Instructv0.1 | 7B| - | 6.84 |-|
55
+ | Zephyr-7b-α |7B| dDPO| 6.88| -|
56
+ | **Zephyr-7b-β** 🪁 | **7B** | **dDPO** | **7.34** | **90.60** |
57
+ | Falcon-Instruct | 40B |dSFT |5.17 |45.71|
58
+ | Guanaco | 65B | SFT |6.41| 71.80|
59
+ | Llama2-Chat | 70B |RLHF |6.86| 92.66|
60
+ | Vicuna v1.3 | 33B |dSFT |7.12 |88.99|
61
+ | WizardLM v1.0 | 70B |dSFT |7.71 |-|
62
+ | Xwin-LM v0.1 | 70B |dPPO |- |95.57|
63
+ | GPT-3.5-turbo | - |RLHF |7.94 |89.37|
64
+ | Claude 2 | - |RLHF |8.06| 91.36|
65
+ | GPT-4 | -| RLHF |8.99| 95.28|
66
+
67
+ In particular, on several categories of MT-Bench, Zephyr-7B-β has strong performance compared to larger open models like Llama2-Chat-70B:
68
+
69
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/6200d0a443eb0913fa2df7cc/raxvt5ma16d7T23my34WC.png)
70
+
71
+ However, on more complex tasks like coding and mathematics, Zephyr-7B-β lags behind proprietary models and more research is needed to close the gap.
72
+
73
+
74
+ ## Intended uses & limitations
75
+
76
+ The model was initially fine-tuned on a filtered and preprocessed of the [`UltraChat`](https://huggingface.co/datasets/stingning/ultrachat) dataset, which contains a diverse range of synthetic dialogues generated by ChatGPT.
77
+ We then further aligned the model with [🤗 TRL's](https://github.com/huggingface/trl) `DPOTrainer` on the [openbmb/UltraFeedback](https://huggingface.co/datasets/openbmb/UltraFeedback) dataset, which contains 64k prompts and model completions that are ranked by GPT-4. As a result, the model can be used for chat and you can check out our [demo](https://huggingface.co/spaces/HuggingFaceH4/zephyr-chat) to test its capabilities.
78
+
79
+ You can find the datasets used for training Zephyr-7B-β [here](https://huggingface.co/collections/HuggingFaceH4/zephyr-7b-6538c6d6d5ddd1cbb1744a66)
80
+
81
+ Here's how you can run the model using the `pipeline()` function from 🤗 Transformers:
82
+
83
+ ```python
84
+ # Install transformers from source - only needed for versions <= v4.34
85
+ # pip install git+https://github.com/huggingface/transformers.git
86
+ # pip install accelerate
87
+
88
+ import torch
89
+ from transformers import pipeline
90
+
91
+ pipe = pipeline("text-generation", model="HuggingFaceH4/zephyr-7b-beta", torch_dtype=torch.bfloat16, device_map="auto")
92
+
93
+ # We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating
94
+ messages = [
95
+ {
96
+ "role": "system",
97
+ "content": "You are a friendly chatbot who always responds in the style of a pirate",
98
+ },
99
+ {"role": "user", "content": "How many helicopters can a human eat in one sitting?"},
100
+ ]
101
+ prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
102
+ outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
103
+ print(outputs[0]["generated_text"])
104
+ # <|system|>
105
+ # You are a friendly chatbot who always responds in the style of a pirate.</s>
106
+ # <|user|>
107
+ # How many helicopters can a human eat in one sitting?</s>
108
+ # <|assistant|>
109
+ # Ah, me hearty matey! But yer question be a puzzler! A human cannot eat a helicopter in one sitting, as helicopters are not edible. They be made of metal, plastic, and other materials, not food!
110
+ ```
111
+
112
+ ## Bias, Risks, and Limitations
113
+
114
+ <!-- This section is meant to convey both technical and sociotechnical limitations. -->
115
+
116
+ Zephyr-7B-β has not been aligned to human preferences with techniques like RLHF or deployed with in-the-loop filtering of responses like ChatGPT, so the model can produce problematic outputs (especially when prompted to do so).
117
+ It is also unknown what the size and composition of the corpus was used to train the base model (`mistralai/Mistral-7B-v0.1`), however it is likely to have included a mix of Web data and technical sources like books and code. See the [Falcon 180B model card](https://huggingface.co/tiiuae/falcon-180B#training-data) for an example of this.
118
+
119
+
120
+ ## Training and evaluation data
121
+
122
+ During DPO training, this model achieves the following results on the evaluation set:
123
+
124
+ - Loss: 0.7496
125
+ - Rewards/chosen: -4.5221
126
+ - Rewards/rejected: -8.3184
127
+ - Rewards/accuracies: 0.7812
128
+ - Rewards/margins: 3.7963
129
+ - Logps/rejected: -340.1541
130
+ - Logps/chosen: -299.4561
131
+ - Logits/rejected: -2.3081
132
+ - Logits/chosen: -2.3531
133
+
134
+
135
+ ### Training hyperparameters
136
+
137
+ The following hyperparameters were used during training:
138
+ - learning_rate: 5e-07
139
+ - train_batch_size: 2
140
+ - eval_batch_size: 4
141
+ - seed: 42
142
+ - distributed_type: multi-GPU
143
+ - num_devices: 16
144
+ - total_train_batch_size: 32
145
+ - total_eval_batch_size: 64
146
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
147
+ - lr_scheduler_type: linear
148
+ - lr_scheduler_warmup_ratio: 0.1
149
+ - num_epochs: 3.0
150
+
151
+ ### Training results
152
+
153
+ The table below shows the full set of DPO training metrics:
154
+
155
+
156
+ | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
157
+ |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
158
+ | 0.6284 | 0.05 | 100 | 0.6098 | 0.0425 | -0.1872 | 0.7344 | 0.2297 | -258.8416 | -253.8099 | -2.7976 | -2.8234 |
159
+ | 0.4908 | 0.1 | 200 | 0.5426 | -0.0279 | -0.6842 | 0.75 | 0.6563 | -263.8124 | -254.5145 | -2.7719 | -2.7960 |
160
+ | 0.5264 | 0.15 | 300 | 0.5324 | 0.0414 | -0.9793 | 0.7656 | 1.0207 | -266.7627 | -253.8209 | -2.7892 | -2.8122 |
161
+ | 0.5536 | 0.21 | 400 | 0.4957 | -0.0185 | -1.5276 | 0.7969 | 1.5091 | -272.2460 | -254.4203 | -2.8542 | -2.8764 |
162
+ | 0.5362 | 0.26 | 500 | 0.5031 | -0.2630 | -1.5917 | 0.7812 | 1.3287 | -272.8869 | -256.8653 | -2.8702 | -2.8958 |
163
+ | 0.5966 | 0.31 | 600 | 0.5963 | -0.2993 | -1.6491 | 0.7812 | 1.3499 | -273.4614 | -257.2279 | -2.8778 | -2.8986 |
164
+ | 0.5014 | 0.36 | 700 | 0.5382 | -0.2859 | -1.4750 | 0.75 | 1.1891 | -271.7204 | -257.0942 | -2.7659 | -2.7869 |
165
+ | 0.5334 | 0.41 | 800 | 0.5677 | -0.4289 | -1.8968 | 0.7969 | 1.4679 | -275.9378 | -258.5242 | -2.7053 | -2.7265 |
166
+ | 0.5251 | 0.46 | 900 | 0.5772 | -0.2116 | -1.3107 | 0.7344 | 1.0991 | -270.0768 | -256.3507 | -2.8463 | -2.8662 |
167
+ | 0.5205 | 0.52 | 1000 | 0.5262 | -0.3792 | -1.8585 | 0.7188 | 1.4793 | -275.5552 | -258.0276 | -2.7893 | -2.7979 |
168
+ | 0.5094 | 0.57 | 1100 | 0.5433 | -0.6279 | -1.9368 | 0.7969 | 1.3089 | -276.3377 | -260.5136 | -2.7453 | -2.7536 |
169
+ | 0.5837 | 0.62 | 1200 | 0.5349 | -0.3780 | -1.9584 | 0.7656 | 1.5804 | -276.5542 | -258.0154 | -2.7643 | -2.7756 |
170
+ | 0.5214 | 0.67 | 1300 | 0.5732 | -1.0055 | -2.2306 | 0.7656 | 1.2251 | -279.2761 | -264.2903 | -2.6986 | -2.7113 |
171
+ | 0.6914 | 0.72 | 1400 | 0.5137 | -0.6912 | -2.1775 | 0.7969 | 1.4863 | -278.7448 | -261.1467 | -2.7166 | -2.7275 |
172
+ | 0.4655 | 0.77 | 1500 | 0.5090 | -0.7987 | -2.2930 | 0.7031 | 1.4943 | -279.8999 | -262.2220 | -2.6651 | -2.6838 |
173
+ | 0.5731 | 0.83 | 1600 | 0.5312 | -0.8253 | -2.3520 | 0.7812 | 1.5268 | -280.4902 | -262.4876 | -2.6543 | -2.6728 |
174
+ | 0.5233 | 0.88 | 1700 | 0.5206 | -0.4573 | -2.0951 | 0.7812 | 1.6377 | -277.9205 | -258.8084 | -2.6870 | -2.7097 |
175
+ | 0.5593 | 0.93 | 1800 | 0.5231 | -0.5508 | -2.2000 | 0.7969 | 1.6492 | -278.9703 | -259.7433 | -2.6221 | -2.6519 |
176
+ | 0.4967 | 0.98 | 1900 | 0.5290 | -0.5340 | -1.9570 | 0.8281 | 1.4230 | -276.5395 | -259.5749 | -2.6564 | -2.6878 |
177
+ | 0.0921 | 1.03 | 2000 | 0.5368 | -1.1376 | -3.1615 | 0.7812 | 2.0239 | -288.5854 | -265.6111 | -2.6040 | -2.6345 |
178
+ | 0.0733 | 1.08 | 2100 | 0.5453 | -1.1045 | -3.4451 | 0.7656 | 2.3406 | -291.4208 | -265.2799 | -2.6289 | -2.6595 |
179
+ | 0.0972 | 1.14 | 2200 | 0.5571 | -1.6915 | -3.9823 | 0.8125 | 2.2908 | -296.7934 | -271.1505 | -2.6471 | -2.6709 |
180
+ | 0.1058 | 1.19 | 2300 | 0.5789 | -1.0621 | -3.8941 | 0.7969 | 2.8319 | -295.9106 | -264.8563 | -2.5527 | -2.5798 |
181
+ | 0.2423 | 1.24 | 2400 | 0.5455 | -1.1963 | -3.5590 | 0.7812 | 2.3627 | -292.5599 | -266.1981 | -2.5414 | -2.5784 |
182
+ | 0.1177 | 1.29 | 2500 | 0.5889 | -1.8141 | -4.3942 | 0.7969 | 2.5801 | -300.9120 | -272.3761 | -2.4802 | -2.5189 |
183
+ | 0.1213 | 1.34 | 2600 | 0.5683 | -1.4608 | -3.8420 | 0.8125 | 2.3812 | -295.3901 | -268.8436 | -2.4774 | -2.5207 |
184
+ | 0.0889 | 1.39 | 2700 | 0.5890 | -1.6007 | -3.7337 | 0.7812 | 2.1330 | -294.3068 | -270.2423 | -2.4123 | -2.4522 |
185
+ | 0.0995 | 1.45 | 2800 | 0.6073 | -1.5519 | -3.8362 | 0.8281 | 2.2843 | -295.3315 | -269.7538 | -2.4685 | -2.5050 |
186
+ | 0.1145 | 1.5 | 2900 | 0.5790 | -1.7939 | -4.2876 | 0.8438 | 2.4937 | -299.8461 | -272.1744 | -2.4272 | -2.4674 |
187
+ | 0.0644 | 1.55 | 3000 | 0.5735 | -1.7285 | -4.2051 | 0.8125 | 2.4766 | -299.0209 | -271.5201 | -2.4193 | -2.4574 |
188
+ | 0.0798 | 1.6 | 3100 | 0.5537 | -1.7226 | -4.2850 | 0.8438 | 2.5624 | -299.8200 | -271.4610 | -2.5367 | -2.5696 |
189
+ | 0.1013 | 1.65 | 3200 | 0.5575 | -1.5715 | -3.9813 | 0.875 | 2.4098 | -296.7825 | -269.9498 | -2.4926 | -2.5267 |
190
+ | 0.1254 | 1.7 | 3300 | 0.5905 | -1.6412 | -4.4703 | 0.8594 | 2.8291 | -301.6730 | -270.6473 | -2.5017 | -2.5340 |
191
+ | 0.085 | 1.76 | 3400 | 0.6133 | -1.9159 | -4.6760 | 0.8438 | 2.7601 | -303.7296 | -273.3941 | -2.4614 | -2.4960 |
192
+ | 0.065 | 1.81 | 3500 | 0.6074 | -1.8237 | -4.3525 | 0.8594 | 2.5288 | -300.4951 | -272.4724 | -2.4597 | -2.5004 |
193
+ | 0.0755 | 1.86 | 3600 | 0.5836 | -1.9252 | -4.4005 | 0.8125 | 2.4753 | -300.9748 | -273.4872 | -2.4327 | -2.4716 |
194
+ | 0.0746 | 1.91 | 3700 | 0.5789 | -1.9280 | -4.4906 | 0.8125 | 2.5626 | -301.8762 | -273.5149 | -2.4686 | -2.5115 |
195
+ | 0.1348 | 1.96 | 3800 | 0.6015 | -1.8658 | -4.2428 | 0.8281 | 2.3769 | -299.3976 | -272.8936 | -2.4943 | -2.5393 |
196
+ | 0.0217 | 2.01 | 3900 | 0.6122 | -2.3335 | -4.9229 | 0.8281 | 2.5894 | -306.1988 | -277.5699 | -2.4841 | -2.5272 |
197
+ | 0.0219 | 2.07 | 4000 | 0.6522 | -2.9890 | -6.0164 | 0.8281 | 3.0274 | -317.1334 | -284.1248 | -2.4105 | -2.4545 |
198
+ | 0.0119 | 2.12 | 4100 | 0.6922 | -3.4777 | -6.6749 | 0.7969 | 3.1972 | -323.7187 | -289.0121 | -2.4272 | -2.4699 |
199
+ | 0.0153 | 2.17 | 4200 | 0.6993 | -3.2406 | -6.6775 | 0.7969 | 3.4369 | -323.7453 | -286.6413 | -2.4047 | -2.4465 |
200
+ | 0.011 | 2.22 | 4300 | 0.7178 | -3.7991 | -7.4397 | 0.7656 | 3.6406 | -331.3667 | -292.2260 | -2.3843 | -2.4290 |
201
+ | 0.0072 | 2.27 | 4400 | 0.6840 | -3.3269 | -6.8021 | 0.8125 | 3.4752 | -324.9908 | -287.5042 | -2.4095 | -2.4536 |
202
+ | 0.0197 | 2.32 | 4500 | 0.7013 | -3.6890 | -7.3014 | 0.8125 | 3.6124 | -329.9841 | -291.1250 | -2.4118 | -2.4543 |
203
+ | 0.0182 | 2.37 | 4600 | 0.7476 | -3.8994 | -7.5366 | 0.8281 | 3.6372 | -332.3356 | -293.2291 | -2.4163 | -2.4565 |
204
+ | 0.0125 | 2.43 | 4700 | 0.7199 | -4.0560 | -7.5765 | 0.8438 | 3.5204 | -332.7345 | -294.7952 | -2.3699 | -2.4100 |
205
+ | 0.0082 | 2.48 | 4800 | 0.7048 | -3.6613 | -7.1356 | 0.875 | 3.4743 | -328.3255 | -290.8477 | -2.3925 | -2.4303 |
206
+ | 0.0118 | 2.53 | 4900 | 0.6976 | -3.7908 | -7.3152 | 0.8125 | 3.5244 | -330.1224 | -292.1431 | -2.3633 | -2.4047 |
207
+ | 0.0118 | 2.58 | 5000 | 0.7198 | -3.9049 | -7.5557 | 0.8281 | 3.6508 | -332.5271 | -293.2844 | -2.3764 | -2.4194 |
208
+ | 0.006 | 2.63 | 5100 | 0.7506 | -4.2118 | -7.9149 | 0.8125 | 3.7032 | -336.1194 | -296.3530 | -2.3407 | -2.3860 |
209
+ | 0.0143 | 2.68 | 5200 | 0.7408 | -4.2433 | -7.9802 | 0.8125 | 3.7369 | -336.7721 | -296.6682 | -2.3509 | -2.3946 |
210
+ | 0.0057 | 2.74 | 5300 | 0.7552 | -4.3392 | -8.0831 | 0.7969 | 3.7439 | -337.8013 | -297.6275 | -2.3388 | -2.3842 |
211
+ | 0.0138 | 2.79 | 5400 | 0.7404 | -4.2395 | -7.9762 | 0.8125 | 3.7367 | -336.7322 | -296.6304 | -2.3286 | -2.3737 |
212
+ | 0.0079 | 2.84 | 5500 | 0.7525 | -4.4466 | -8.2196 | 0.7812 | 3.7731 | -339.1662 | -298.7007 | -2.3200 | -2.3641 |
213
+ | 0.0077 | 2.89 | 5600 | 0.7520 | -4.5586 | -8.3485 | 0.7969 | 3.7899 | -340.4545 | -299.8206 | -2.3078 | -2.3517 |
214
+ | 0.0094 | 2.94 | 5700 | 0.7527 | -4.5542 | -8.3509 | 0.7812 | 3.7967 | -340.4790 | -299.7773 | -2.3062 | -2.3510 |
215
+ | 0.0054 | 2.99 | 5800 | 0.7520 | -4.5169 | -8.3079 | 0.7812 | 3.7911 | -340.0493 | -299.4038 | -2.3081 | -2.3530 |
216
+
217
+
218
+ ### Framework versions
219
+
220
+ - Transformers 4.35.0.dev0
221
+ - Pytorch 2.0.1+cu118
222
+ - Datasets 2.12.0
223
+ - Tokenizers 0.14.0
224
+
225
+ ## Citation
226
+
227
+ If you find Zephyr-7B-β is useful in your work, please cite it with:
228
+
229
+ ```
230
+ @misc{tunstall2023zephyr,
231
+ title={Zephyr: Direct Distillation of LM Alignment},
232
+ author={Lewis Tunstall and Edward Beeching and Nathan Lambert and Nazneen Rajani and Kashif Rasul and Younes Belkada and Shengyi Huang and Leandro von Werra and Clémentine Fourrier and Nathan Habib and Nathan Sarrazin and Omar Sanseviero and Alexander M. Rush and Thomas Wolf},
233
+ year={2023},
234
+ eprint={2310.16944},
235
+ archivePrefix={arXiv},
236
+ primaryClass={cs.LG}
237
+ }
238
+ ```