TheBloke commited on
Commit
ba35a88
β€’
1 Parent(s): edfb620

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +131 -69
README.md CHANGED
@@ -1,6 +1,6 @@
1
  ---
2
  inference: false
3
- license: other
4
  model_creator: WizardLM
5
  model_link: https://huggingface.co/WizardLM/WizardMath-70B-V1.0
6
  model_name: WizardMath 70B V1.0
@@ -29,18 +29,24 @@ quantized_by: TheBloke
29
  - Model creator: [WizardLM](https://huggingface.co/WizardLM)
30
  - Original model: [WizardMath 70B V1.0](https://huggingface.co/WizardLM/WizardMath-70B-V1.0)
31
 
 
32
  ## Description
33
 
34
  This repo contains GPTQ model files for [WizardLM's WizardMath 70B V1.0](https://huggingface.co/WizardLM/WizardMath-70B-V1.0).
35
 
36
  Multiple GPTQ parameter permutations are provided; see Provided Files below for details of the options provided, their parameters, and the software used to create them.
37
 
 
 
38
  ## Repositories available
39
 
40
  * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ)
41
- * [2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GGML)
 
42
  * [WizardLM's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/WizardLM/WizardMath-70B-V1.0)
 
43
 
 
44
  ## Prompt template: Alpaca-CoT
45
 
46
  ```
@@ -52,22 +58,26 @@ Below is an instruction that describes a task. Write a response that appropriate
52
 
53
 
54
  ### Response: Let's think step by step.
 
55
  ```
56
 
 
 
 
57
  ## Provided files and GPTQ parameters
58
 
59
  Multiple quantisation parameters are provided, to allow you to choose the best one for your hardware and requirements.
60
 
61
  Each separate quant is in a different branch. See below for instructions on fetching from different branches.
62
 
63
- All GPTQ files are made with AutoGPTQ.
64
 
65
  <details>
66
  <summary>Explanation of GPTQ parameters</summary>
67
 
68
  - Bits: The bit size of the quantised model.
69
  - GS: GPTQ group size. Higher numbers use less VRAM, but have lower quantisation accuracy. "None" is the lowest possible value.
70
- - Act Order: True or False. Also known as `desc_act`. True results in better quantisation accuracy. Some GPTQ clients have issues with models that use Act Order plus Group Size.
71
  - Damp %: A GPTQ parameter that affects how samples are processed for quantisation. 0.01 is default, but 0.1 results in slightly better accuracy.
72
  - GPTQ dataset: The dataset used for quantisation. Using a dataset more appropriate to the model's training can improve quantisation accuracy. Note that the GPTQ dataset is not the same as the dataset used to train the model - please refer to the original model repo for details of the training dataset(s).
73
  - Sequence Length: The length of the dataset sequences used for quantisation. Ideally this is the same as the model sequence length. For some very long sequence models (16+K), a lower sequence length may have to be used. Note that a lower sequence length does not limit the sequence length of the quantised model. It only impacts the quantisation accuracy on longer inference sequences.
@@ -77,13 +87,16 @@ All GPTQ files are made with AutoGPTQ.
77
 
78
  | Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
79
  | ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
80
- | [main](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/main) | 4 | None | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 35.33 GB | Yes | Most compatible option. Good inference speed in AutoGPTQ and GPTQ-for-LLaMa. Lower inference quality than other options. |
81
- | [gptq-4bit-32g-actorder_True](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 40.66 GB | Yes | 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage. Poor AutoGPTQ CUDA speed. |
82
- | [gptq-4bit-64g-actorder_True](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/gptq-4bit-64g-actorder_True) | 4 | 64 | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 37.99 GB | Yes | 4-bit, with Act Order and group size 64g. Uses less VRAM than 32g, but with slightly lower accuracy. Poor AutoGPTQ CUDA speed. |
83
- | [gptq-4bit-128g-actorder_True](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/gptq-4bit-128g-actorder_True) | 4 | 128 | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 36.65 GB | Yes | 4-bit, with Act Order and group size 128g. Uses even less VRAM than 64g, but with slightly lower accuracy. Poor AutoGPTQ CUDA speed. |
84
- | [gptq-3bit--1g-actorder_True](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/gptq-3bit--1g-actorder_True) | 3 | None | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 26.78 GB | No | 3-bit, with Act Order and no group size. Lowest possible VRAM requirements. May be lower quality than 3-bit 128g. |
85
  | [gptq-3bit-128g-actorder_True](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/gptq-3bit-128g-actorder_True) | 3 | 128 | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 28.03 GB | No | 3-bit, with group size 128g and act-order. Higher quality than 128g-False but poor AutoGPTQ CUDA speed. |
86
 
 
 
 
87
  ## How to download from branches
88
 
89
  - In text-generation-webui, you can add `:branch` to the end of the download name, eg `TheBloke/WizardMath-70B-V1.0-GPTQ:gptq-4bit-32g-actorder_True`
@@ -92,73 +105,72 @@ All GPTQ files are made with AutoGPTQ.
92
  git clone --single-branch --branch gptq-4bit-32g-actorder_True https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ
93
  ```
94
  - In Python Transformers code, the branch is the `revision` parameter; see below.
95
-
 
96
  ## How to easily download and use this model in [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
97
 
98
  Please make sure you're using the latest version of [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
99
 
100
- It is strongly recommended to use the text-generation-webui one-click-installers unless you know how to make a manual install.
101
 
102
  1. Click the **Model tab**.
103
  2. Under **Download custom model or LoRA**, enter `TheBloke/WizardMath-70B-V1.0-GPTQ`.
104
  - To download from a specific branch, enter for example `TheBloke/WizardMath-70B-V1.0-GPTQ:gptq-4bit-32g-actorder_True`
105
  - see Provided Files above for the list of branches for each option.
106
  3. Click **Download**.
107
- 4. The model will start downloading. Once it's finished it will say "Done"
108
  5. In the top left, click the refresh icon next to **Model**.
109
  6. In the **Model** dropdown, choose the model you just downloaded: `WizardMath-70B-V1.0-GPTQ`
110
  7. The model will automatically load, and is now ready for use!
111
  8. If you want any custom settings, set them and then click **Save settings for this model** followed by **Reload the Model** in the top right.
112
- * Note that you do not need to set GPTQ parameters any more. These are set automatically from the file `quantize_config.json`.
113
  9. Once you're ready, click the **Text Generation tab** and enter a prompt to get started!
 
114
 
 
115
  ## How to use this GPTQ model from Python code
116
 
117
- First make sure you have [AutoGPTQ](https://github.com/PanQiWei/AutoGPTQ) 0.3.1 or later installed:
118
 
119
- ```
120
- pip3 install auto-gptq
121
- ```
122
 
123
- If you have problems installing AutoGPTQ, please build from source instead:
 
 
124
  ```
 
 
 
 
125
  pip3 uninstall -y auto-gptq
126
  git clone https://github.com/PanQiWei/AutoGPTQ
127
  cd AutoGPTQ
128
  pip3 install .
129
  ```
130
 
131
- Then try the following example code:
 
 
 
 
 
 
 
 
132
 
133
  ```python
134
- from transformers import AutoTokenizer, pipeline, logging
135
- from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
136
 
137
  model_name_or_path = "TheBloke/WizardMath-70B-V1.0-GPTQ"
138
-
139
- use_triton = False
 
 
 
 
140
 
141
  tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
142
 
143
- model = AutoGPTQForCausalLM.from_quantized(model_name_or_path,
144
- use_safetensors=True,
145
- trust_remote_code=False,
146
- device="cuda:0",
147
- use_triton=use_triton,
148
- quantize_config=None)
149
-
150
- """
151
- # To download from a specific branch, use the revision parameter, as in this example:
152
- # Note that `revision` requires AutoGPTQ 0.3.1 or later!
153
-
154
- model = AutoGPTQForCausalLM.from_quantized(model_name_or_path,
155
- revision="gptq-4bit-32g-actorder_True",
156
- use_safetensors=True,
157
- trust_remote_code=False,
158
- device="cuda:0",
159
- quantize_config=None)
160
- """
161
-
162
  prompt = "Tell me about AI"
163
  prompt_template=f'''Below is an instruction that describes a task. Write a response that appropriately completes the request.
164
 
@@ -168,6 +180,7 @@ prompt_template=f'''Below is an instruction that describes a task. Write a respo
168
 
169
 
170
  ### Response: Let's think step by step.
 
171
  '''
172
 
173
  print("\n\n*** Generate:")
@@ -178,9 +191,6 @@ print(tokenizer.decode(output[0]))
178
 
179
  # Inference can also be done using transformers' pipeline
180
 
181
- # Prevent printing spurious transformers error when using pipeline with AutoGPTQ
182
- logging.set_verbosity(logging.CRITICAL)
183
-
184
  print("*** Pipeline:")
185
  pipe = pipeline(
186
  "text-generation",
@@ -194,12 +204,17 @@ pipe = pipeline(
194
 
195
  print(pipe(prompt_template)[0]['generated_text'])
196
  ```
 
197
 
 
198
  ## Compatibility
199
 
200
- The files provided will work with AutoGPTQ (CUDA and Triton modes), GPTQ-for-LLaMa (only CUDA has been tested), and Occ4m's GPTQ-for-LLaMa fork.
201
 
202
- ExLlama works with Llama models in 4-bit. Please see the Provided Files table above for per-file compatibility.
 
 
 
203
 
204
  <!-- footer start -->
205
  <!-- 200823 -->
@@ -224,7 +239,7 @@ Donaters will get priority support on any and all AI/LLM/model questions and req
224
 
225
  **Special thanks to**: Aemon Algiz.
226
 
227
- **Patreon special mentions**: Sam, theTransient, Jonathan Leane, Steven Wood, webtim, Johann-Peter Hartmann, Geoffrey Montalvo, Gabriel Tamborski, Willem Michiel, John Villwock, Derek Yates, Mesiah Bishop, Eugene Pentland, Pieter, Chadd, Stephen Murray, Daniel P. Andersen, terasurfer, Brandon Frisco, Thomas Belote, Sid, Nathan LeClaire, Magnesian, Alps Aficionado, Stanislav Ovsiannikov, Alex, Joseph William Delisle, Nikolai Manek, Michael Davis, Junyu Yang, K, J, Spencer Kim, Stefan Sabev, Olusegun Samson, transmissions 11, Michael Levine, Cory Kujawski, Rainer Wilmers, zynix, Kalila, Luke @flexchar, Ajan Kanaga, Mandus, vamX, Ai Maven, Mano Prime, Matthew Berman, subjectnull, Vitor Caleffi, Clay Pascal, biorpg, alfie_i, 阿明, Jeffrey Morgan, ya boyyy, Raymond Fosdick, knownsqashed, Olakabola, Leonard Tan, ReadyPlayerEmma, Enrico Ros, Dave, Talal Aujan, Illia Dulskyi, Sean Connelly, senxiiz, Artur Olbinski, Elle, Raven Klaugh, Fen Risland, Deep Realms, Imad Khwaja, Fred von Graf, Will Dee, usrbinkat, SuperWojo, Alexandros Triantafyllidis, Swaroop Kallakuri, Dan Guido, John Detwiler, Pedro Madruga, Iucharbius, Viktor Bowallius, Asp the Wyvern, Edmond Seymore, Trenton Dambrowitz, Space Cruiser, Spiking Neurons AB, Pyrater, LangChain4j, Tony Hughes, Kacper WikieΕ‚, Rishabh Srivastava, David Ziegler, Luke Pendergrass, Andrey, Gabriel Puliatti, Lone Striker, Sebastain Graf, Pierre Kircher, Randy H, NimbleBox.ai, Vadim, danny, Deo Leter
228
 
229
 
230
  Thank you to all my generous patrons and donaters!
@@ -237,53 +252,100 @@ And thank you again to a16z for their generous grant.
237
 
238
 
239
 
240
- ## WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct
241
-
242
 
243
 
244
  <p align="center">
245
- πŸ€— <a href="https://huggingface.co/WizardLM" target="_blank">HF Repo</a> β€’ 🐦 <a href="https://twitter.com/WizardLM_AI" target="_blank">Twitter</a> β€’ πŸ“ƒ <a href="https://arxiv.org/abs/2304.12244" target="_blank">[WizardLM]</a> β€’ πŸ“ƒ <a href="https://arxiv.org/abs/2306.08568" target="_blank">[WizardCoder]</a> <br>
246
  </p>
247
  <p align="center">
248
- πŸ‘‹ Join our <a href="https://discord.gg/bpmeZD7V" target="_blank">Discord</a>
249
  </p>
250
 
251
-
252
-
253
-
254
-
255
- | Model | Checkpoint | Paper | GSM8k | MATH | License|
256
- | ----- |------| ---- |------|-------| ----- |
257
- | WizardMath-70B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardMath-70B-V1.0" target="_blank">HF Link</a> | πŸ“ƒComing Soon| **81.6** | **22.7** | <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama 2 License </a> |
258
- | WizardMath-13B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardMath-13B-V1.0" target="_blank">HF Link</a> | πŸ“ƒComing Soon| **63.9** | **14.0** | <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama 2 License </a> |
259
- | WizardMath-7B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardMath-7B-V1.0" target="_blank">HF Link</a> | πŸ“ƒComing Soon| **54.9** | **10.7** | <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama 2 License </a>|
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
260
 
261
  **Github Repo**: https://github.com/nlpxucan/WizardLM/tree/main/WizardMath
262
 
263
- **Twitter**: https://twitter.com/WizardLM_AI/status/1689990201467432960
264
 
265
- **Discord**: https://discord.gg/bpmeZD7V
266
 
 
267
 
 
 
 
 
 
268
 
269
  ❗<b>Note for model system prompts usage:</b>
270
 
271
- ## CoT Version:
 
 
272
 
273
  ```
274
- Below is an instruction that describes a task. Write a response that appropriately completes the request.
 
275
 
276
 
277
- ### Instruction:
278
- {instruction}
279
 
280
 
281
- ### Response: Let's think step by step.
 
282
  ```
283
 
 
 
 
 
 
284
  ❗<b>To commen concern about dataset:</b>
285
 
286
- Recently, there have been clear changes in the open-source policy and regulations of our overall organization's code, data, and models.
287
  Despite this, we have still worked hard to obtain opening the weights of the model first, but the data involves stricter auditing and is in review with our legal team .
288
  Our researchers have no authority to publicly release them without authorization.
289
  Thank you for your understanding.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  inference: false
3
+ license: llama2
4
  model_creator: WizardLM
5
  model_link: https://huggingface.co/WizardLM/WizardMath-70B-V1.0
6
  model_name: WizardMath 70B V1.0
 
29
  - Model creator: [WizardLM](https://huggingface.co/WizardLM)
30
  - Original model: [WizardMath 70B V1.0](https://huggingface.co/WizardLM/WizardMath-70B-V1.0)
31
 
32
+ <!-- description start -->
33
  ## Description
34
 
35
  This repo contains GPTQ model files for [WizardLM's WizardMath 70B V1.0](https://huggingface.co/WizardLM/WizardMath-70B-V1.0).
36
 
37
  Multiple GPTQ parameter permutations are provided; see Provided Files below for details of the options provided, their parameters, and the software used to create them.
38
 
39
+ <!-- description end -->
40
+ <!-- repositories-available start -->
41
  ## Repositories available
42
 
43
  * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ)
44
+ * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GGUF)
45
+ * [2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference (deprecated)](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GGML)
46
  * [WizardLM's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/WizardLM/WizardMath-70B-V1.0)
47
+ <!-- repositories-available end -->
48
 
49
+ <!-- prompt-template start -->
50
  ## Prompt template: Alpaca-CoT
51
 
52
  ```
 
58
 
59
 
60
  ### Response: Let's think step by step.
61
+
62
  ```
63
 
64
+ <!-- prompt-template end -->
65
+
66
+ <!-- README_GPTQ.md-provided-files start -->
67
  ## Provided files and GPTQ parameters
68
 
69
  Multiple quantisation parameters are provided, to allow you to choose the best one for your hardware and requirements.
70
 
71
  Each separate quant is in a different branch. See below for instructions on fetching from different branches.
72
 
73
+ All recent GPTQ files are made with AutoGPTQ, and all files in non-main branches are made with AutoGPTQ. Files in the `main` branch which were uploaded before August 2023 were made with GPTQ-for-LLaMa.
74
 
75
  <details>
76
  <summary>Explanation of GPTQ parameters</summary>
77
 
78
  - Bits: The bit size of the quantised model.
79
  - GS: GPTQ group size. Higher numbers use less VRAM, but have lower quantisation accuracy. "None" is the lowest possible value.
80
+ - Act Order: True or False. Also known as `desc_act`. True results in better quantisation accuracy. Some GPTQ clients have had issues with models that use Act Order plus Group Size, but this is generally resolved now.
81
  - Damp %: A GPTQ parameter that affects how samples are processed for quantisation. 0.01 is default, but 0.1 results in slightly better accuracy.
82
  - GPTQ dataset: The dataset used for quantisation. Using a dataset more appropriate to the model's training can improve quantisation accuracy. Note that the GPTQ dataset is not the same as the dataset used to train the model - please refer to the original model repo for details of the training dataset(s).
83
  - Sequence Length: The length of the dataset sequences used for quantisation. Ideally this is the same as the model sequence length. For some very long sequence models (16+K), a lower sequence length may have to be used. Note that a lower sequence length does not limit the sequence length of the quantised model. It only impacts the quantisation accuracy on longer inference sequences.
 
87
 
88
  | Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
89
  | ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
90
+ | [main](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/main) | 4 | None | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 35.33 GB | Yes | Most compatible option. Good inference speed in AutoGPTQ and GPTQ-for-LLaMa. Lower inference quality than other options. |
91
+ | [gptq-4bit-32g-actorder_True](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 40.66 GB | Yes | 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage. Poor AutoGPTQ CUDA speed. |
92
+ | [gptq-4bit-64g-actorder_True](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/gptq-4bit-64g-actorder_True) | 4 | 64 | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 37.99 GB | Yes | 4-bit, with Act Order and group size 64g. Uses less VRAM than 32g, but with slightly lower accuracy. Poor AutoGPTQ CUDA speed. |
93
+ | [gptq-4bit-128g-actorder_True](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/gptq-4bit-128g-actorder_True) | 4 | 128 | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 36.65 GB | Yes | 4-bit, with Act Order and group size 128g. Uses even less VRAM than 64g, but with slightly lower accuracy. Poor AutoGPTQ CUDA speed. |
94
+ | [gptq-3bit--1g-actorder_True](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/gptq-3bit--1g-actorder_True) | 3 | None | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 26.78 GB | No | 3-bit, with Act Order and no group size. Lowest possible VRAM requirements. May be lower quality than 3-bit 128g. |
95
  | [gptq-3bit-128g-actorder_True](https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ/tree/gptq-3bit-128g-actorder_True) | 3 | 128 | Yes | 0.1 | [CamelAI Math](https://huggingface.co/datasets/andersonbcdefg/math) | 4096 | 28.03 GB | No | 3-bit, with group size 128g and act-order. Higher quality than 128g-False but poor AutoGPTQ CUDA speed. |
96
 
97
+ <!-- README_GPTQ.md-provided-files end -->
98
+
99
+ <!-- README_GPTQ.md-download-from-branches start -->
100
  ## How to download from branches
101
 
102
  - In text-generation-webui, you can add `:branch` to the end of the download name, eg `TheBloke/WizardMath-70B-V1.0-GPTQ:gptq-4bit-32g-actorder_True`
 
105
  git clone --single-branch --branch gptq-4bit-32g-actorder_True https://huggingface.co/TheBloke/WizardMath-70B-V1.0-GPTQ
106
  ```
107
  - In Python Transformers code, the branch is the `revision` parameter; see below.
108
+ <!-- README_GPTQ.md-download-from-branches end -->
109
+ <!-- README_GPTQ.md-text-generation-webui start -->
110
  ## How to easily download and use this model in [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
111
 
112
  Please make sure you're using the latest version of [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
113
 
114
+ It is strongly recommended to use the text-generation-webui one-click-installers unless you're sure you know how to make a manual install.
115
 
116
  1. Click the **Model tab**.
117
  2. Under **Download custom model or LoRA**, enter `TheBloke/WizardMath-70B-V1.0-GPTQ`.
118
  - To download from a specific branch, enter for example `TheBloke/WizardMath-70B-V1.0-GPTQ:gptq-4bit-32g-actorder_True`
119
  - see Provided Files above for the list of branches for each option.
120
  3. Click **Download**.
121
+ 4. The model will start downloading. Once it's finished it will say "Done".
122
  5. In the top left, click the refresh icon next to **Model**.
123
  6. In the **Model** dropdown, choose the model you just downloaded: `WizardMath-70B-V1.0-GPTQ`
124
  7. The model will automatically load, and is now ready for use!
125
  8. If you want any custom settings, set them and then click **Save settings for this model** followed by **Reload the Model** in the top right.
126
+ * Note that you do not need to and should not set manual GPTQ parameters any more. These are set automatically from the file `quantize_config.json`.
127
  9. Once you're ready, click the **Text Generation tab** and enter a prompt to get started!
128
+ <!-- README_GPTQ.md-text-generation-webui end -->
129
 
130
+ <!-- README_GPTQ.md-use-from-python start -->
131
  ## How to use this GPTQ model from Python code
132
 
133
+ ### Install the necessary packages
134
 
135
+ Requires: Transformers 4.32.0 or later, Optimum 1.12.0 or later, and AutoGPTQ 0.4.2 or later.
 
 
136
 
137
+ ```shell
138
+ pip3 install transformers>=4.32.0 optimum>=1.12.0
139
+ pip3 install auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/ # Use cu117 if on CUDA 11.7
140
  ```
141
+
142
+ If you have problems installing AutoGPTQ using the pre-built wheels, install it from source instead:
143
+
144
+ ```shell
145
  pip3 uninstall -y auto-gptq
146
  git clone https://github.com/PanQiWei/AutoGPTQ
147
  cd AutoGPTQ
148
  pip3 install .
149
  ```
150
 
151
+ ### For CodeLlama models only: you must use Transformers 4.33.0 or later.
152
+
153
+ If 4.33.0 is not yet released when you read this, you will need to install Transformers from source:
154
+ ```shell
155
+ pip3 uninstall -y transformers
156
+ pip3 install git+https://github.com/huggingface/transformers.git
157
+ ```
158
+
159
+ ### You can then use the following code
160
 
161
  ```python
162
+ from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 
163
 
164
  model_name_or_path = "TheBloke/WizardMath-70B-V1.0-GPTQ"
165
+ # To use a different branch, change revision
166
+ # For example: revision="gptq-4bit-32g-actorder_True"
167
+ model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
168
+ torch_dtype=torch.float16,
169
+ device_map="auto",
170
+ revision="main")
171
 
172
  tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
173
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
174
  prompt = "Tell me about AI"
175
  prompt_template=f'''Below is an instruction that describes a task. Write a response that appropriately completes the request.
176
 
 
180
 
181
 
182
  ### Response: Let's think step by step.
183
+
184
  '''
185
 
186
  print("\n\n*** Generate:")
 
191
 
192
  # Inference can also be done using transformers' pipeline
193
 
 
 
 
194
  print("*** Pipeline:")
195
  pipe = pipeline(
196
  "text-generation",
 
204
 
205
  print(pipe(prompt_template)[0]['generated_text'])
206
  ```
207
+ <!-- README_GPTQ.md-use-from-python end -->
208
 
209
+ <!-- README_GPTQ.md-compatibility start -->
210
  ## Compatibility
211
 
212
+ The files provided are tested to work with AutoGPTQ, both via Transformers and using AutoGPTQ directly. They should also work with [Occ4m's GPTQ-for-LLaMa fork](https://github.com/0cc4m/KoboldAI).
213
 
214
+ [ExLlama](https://github.com/turboderp/exllama) is compatible with Llama models in 4-bit. Please see the Provided Files table above for per-file compatibility.
215
+
216
+ [Huggingface Text Generation Inference (TGI)](https://github.com/huggingface/text-generation-inference) is compatible with all GPTQ models.
217
+ <!-- README_GPTQ.md-compatibility end -->
218
 
219
  <!-- footer start -->
220
  <!-- 200823 -->
 
239
 
240
  **Special thanks to**: Aemon Algiz.
241
 
242
+ **Patreon special mentions**: Russ Johnson, J, alfie_i, Alex, NimbleBox.ai, Chadd, Mandus, Nikolai Manek, Ken Nordquist, ya boyyy, Illia Dulskyi, Viktor Bowallius, vamX, Iucharbius, zynix, Magnesian, Clay Pascal, Pierre Kircher, Enrico Ros, Tony Hughes, Elle, Andrey, knownsqashed, Deep Realms, Jerry Meng, Lone Striker, Derek Yates, Pyrater, Mesiah Bishop, James Bentley, Femi Adebogun, Brandon Frisco, SuperWojo, Alps Aficionado, Michael Dempsey, Vitor Caleffi, Will Dee, Edmond Seymore, usrbinkat, LangChain4j, Kacper WikieΕ‚, Luke Pendergrass, John Detwiler, theTransient, Nathan LeClaire, Tiffany J. Kim, biorpg, Eugene Pentland, Stanislav Ovsiannikov, Fred von Graf, terasurfer, Kalila, Dan Guido, Nitin Borwankar, 阿明, Ai Maven, John Villwock, Gabriel Puliatti, Stephen Murray, Asp the Wyvern, danny, Chris Smitley, ReadyPlayerEmma, S_X, Daniel P. Andersen, Olakabola, Jeffrey Morgan, Imad Khwaja, Caitlyn Gatomon, webtim, Alicia Loh, Trenton Dambrowitz, Swaroop Kallakuri, Erik BjΓ€reholt, Leonard Tan, Spiking Neurons AB, Luke @flexchar, Ajan Kanaga, Thomas Belote, Deo Leter, RoA, Willem Michiel, transmissions 11, subjectnull, Matthew Berman, Joseph William Delisle, David Ziegler, Michael Davis, Johann-Peter Hartmann, Talal Aujan, senxiiz, Artur Olbinski, Rainer Wilmers, Spencer Kim, Fen Risland, Cap'n Zoog, Rishabh Srivastava, Michael Levine, Geoffrey Montalvo, Sean Connelly, Alexandros Triantafyllidis, Pieter, Gabriel Tamborski, Sam, Subspace Studios, Junyu Yang, Pedro Madruga, Vadim, Cory Kujawski, K, Raven Klaugh, Randy H, Mano Prime, Sebastain Graf, Space Cruiser
243
 
244
 
245
  Thank you to all my generous patrons and donaters!
 
252
 
253
 
254
 
255
+ ## WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct (RLEIF)
 
256
 
257
 
258
  <p align="center">
259
+ πŸ€— <a href="https://huggingface.co/WizardLM" target="_blank">HF Repo</a> β€’πŸ± <a href="https://github.com/nlpxucan/WizardLM" target="_blank">Github Repo</a> β€’ 🐦 <a href="https://twitter.com/WizardLM_AI" target="_blank">Twitter</a> β€’ πŸ“ƒ <a href="https://arxiv.org/abs/2304.12244" target="_blank">[WizardLM]</a> β€’ πŸ“ƒ <a href="https://arxiv.org/abs/2306.08568" target="_blank">[WizardCoder]</a> β€’ πŸ“ƒ <a href="https://arxiv.org/abs/2308.09583" target="_blank">[WizardMath]</a> <br>
260
  </p>
261
  <p align="center">
262
+ πŸ‘‹ Join our <a href="https://discord.gg/VZjjHtWrKs" target="_blank">Discord</a>
263
  </p>
264
 
265
+ | Model | Checkpoint | Paper | HumanEval | MBPP | Demo | License |
266
+ | ----- |------| ---- |------|-------| ----- | ----- |
267
+ | WizardCoder-Python-34B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardCoder-Python-34B-V1.0" target="_blank">HF Link</a> | πŸ“ƒ <a href="https://arxiv.org/abs/2306.08568" target="_blank">[WizardCoder]</a> | 73.2 | 61.2 | [Demo](http://47.103.63.15:50085/) | <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama2</a> |
268
+ | WizardCoder-15B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardCoder-15B-V1.0" target="_blank">HF Link</a> | πŸ“ƒ <a href="https://arxiv.org/abs/2306.08568" target="_blank">[WizardCoder]</a> | 59.8 |50.6 | -- | <a href="https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement" target="_blank">OpenRAIL-M</a> |
269
+ | WizardCoder-Python-13B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardCoder-Python-13B-V1.0" target="_blank">HF Link</a> | πŸ“ƒ <a href="https://arxiv.org/abs/2306.08568" target="_blank">[WizardCoder]</a> | 64.0 | 55.6 | -- | <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama2</a> |
270
+ | WizardCoder-Python-7B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardCoder-Python-7B-V1.0" target="_blank">HF Link</a> | πŸ“ƒ <a href="https://arxiv.org/abs/2306.08568" target="_blank">[WizardCoder]</a> | 55.5 | 51.6 | [Demo](http://47.103.63.15:50088/) | <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama2</a> |
271
+ | WizardCoder-3B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardCoder-3B-V1.0" target="_blank">HF Link</a> | πŸ“ƒ <a href="https://arxiv.org/abs/2306.08568" target="_blank">[WizardCoder]</a> | 34.8 |37.4 | -- | <a href="https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement" target="_blank">OpenRAIL-M</a> |
272
+ | WizardCoder-1B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardCoder-1B-V1.0" target="_blank">HF Link</a> | πŸ“ƒ <a href="https://arxiv.org/abs/2306.08568" target="_blank">[WizardCoder]</a> | 23.8 |28.6 | -- | <a href="https://huggingface.co/spaces/bigcode/bigcode-model-license-agreement" target="_blank">OpenRAIL-M</a> |
273
+
274
+
275
+ | Model | Checkpoint | Paper | GSM8k | MATH |Online Demo| License|
276
+ | ----- |------| ---- |------|-------| ----- | ----- |
277
+ | WizardMath-70B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardMath-70B-V1.0" target="_blank">HF Link</a> | πŸ“ƒ <a href="https://arxiv.org/abs/2308.09583" target="_blank">[WizardMath]</a>| **81.6** | **22.7** |[Demo](http://47.103.63.15:50083/)| <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama 2 </a> |
278
+ | WizardMath-13B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardMath-13B-V1.0" target="_blank">HF Link</a> | πŸ“ƒ <a href="https://arxiv.org/abs/2308.09583" target="_blank">[WizardMath]</a>| **63.9** | **14.0** |[Demo](http://47.103.63.15:50082/)| <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama 2 </a> |
279
+ | WizardMath-7B-V1.0 | πŸ€— <a href="https://huggingface.co/WizardLM/WizardMath-7B-V1.0" target="_blank">HF Link</a> | πŸ“ƒ <a href="https://arxiv.org/abs/2308.09583" target="_blank">[WizardMath]</a>| **54.9** | **10.7** | [Demo](http://47.103.63.15:50080/)| <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama 2 </a>|
280
+
281
+
282
+
283
+ <font size=4>
284
+
285
+ | <sup>Model</sup> | <sup>Checkpoint</sup> | <sup>Paper</sup> |<sup>MT-Bench</sup> | <sup>AlpacaEval</sup> | <sup>GSM8k</sup> | <sup>HumanEval</sup> | <sup>License</sup>|
286
+ | ----- |------| ---- |------|-------| ----- | ----- | ----- |
287
+ | <sup>**WizardLM-70B-V1.0**</sup> | <sup>πŸ€— <a href="https://huggingface.co/WizardLM/WizardLM-70B-V1.0" target="_blank">HF Link</a> </sup>|<sup>πŸ“ƒ**Coming Soon**</sup>| <sup>**7.78**</sup> | <sup>**92.91%**</sup> |<sup>**77.6%**</sup> | <sup> **50.6 pass@1**</sup>|<sup> <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama 2 License </a></sup> |
288
+ | <sup>WizardLM-13B-V1.2</sup> | <sup>πŸ€— <a href="https://huggingface.co/WizardLM/WizardLM-13B-V1.2" target="_blank">HF Link</a> </sup>| | <sup>7.06</sup> | <sup>89.17%</sup> |<sup>55.3%</sup> | <sup>36.6 pass@1</sup>|<sup> <a href="https://ai.meta.com/resources/models-and-libraries/llama-downloads/" target="_blank">Llama 2 License </a></sup> |
289
+ | <sup>WizardLM-13B-V1.1</sup> |<sup> πŸ€— <a href="https://huggingface.co/WizardLM/WizardLM-13B-V1.1" target="_blank">HF Link</a> </sup> | | <sup>6.76</sup> |<sup>86.32%</sup> | | <sup>25.0 pass@1</sup>| <sup>Non-commercial</sup>|
290
+ | <sup>WizardLM-30B-V1.0</sup> | <sup>πŸ€— <a href="https://huggingface.co/WizardLM/WizardLM-30B-V1.0" target="_blank">HF Link</a></sup> | | <sup>7.01</sup> | | | <sup>37.8 pass@1</sup>| <sup>Non-commercial</sup> |
291
+ | <sup>WizardLM-13B-V1.0</sup> | <sup>πŸ€— <a href="https://huggingface.co/WizardLM/WizardLM-13B-V1.0" target="_blank">HF Link</a> </sup> | | <sup>6.35</sup> | <sup>75.31%</sup> | | <sup> 24.0 pass@1 </sup> | <sup>Non-commercial</sup>|
292
+ | <sup>WizardLM-7B-V1.0 </sup>| <sup>πŸ€— <a href="https://huggingface.co/WizardLM/WizardLM-7B-V1.0" target="_blank">HF Link</a> </sup> |<sup> πŸ“ƒ <a href="https://arxiv.org/abs/2304.12244" target="_blank">[WizardLM]</a> </sup>| | | |<sup>19.1 pass@1 </sup>|<sup> Non-commercial</sup>|
293
+ </font>
294
 
295
  **Github Repo**: https://github.com/nlpxucan/WizardLM/tree/main/WizardMath
296
 
297
+ **Twitter**: https://twitter.com/WizardLM_AI/status/1689998428200112128
298
 
299
+ **Discord**: https://discord.gg/VZjjHtWrKs
300
 
301
+ ## Comparing WizardMath-V1.0 with Other LLMs.
302
 
303
+ πŸ”₯ The following figure shows that our **WizardMath-70B-V1.0 attains the fifth position in this benchmark**, surpassing ChatGPT (81.6 vs. 80.8) , Claude Instant (81.6 vs. 80.9), PaLM 2 540B (81.6 vs. 80.7).
304
+
305
+ <p align="center" width="100%">
306
+ <a ><img src="https://raw.githubusercontent.com/nlpxucan/WizardLM/main/WizardMath/images/wizardmath_gsm8k.png" alt="WizardMath" style="width: 96%; min-width: 300px; display: block; margin: auto;"></a>
307
+ </p>
308
 
309
  ❗<b>Note for model system prompts usage:</b>
310
 
311
+ Please use **the same systems prompts strictly** with us, and we do not guarantee the accuracy of the **quantified versions**.
312
+
313
+ **Default version:**
314
 
315
  ```
316
+ "Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n### Instruction:\n{instruction}\n\n### Response:"
317
+ ```
318
 
319
 
320
+ **CoT Version:** οΌˆβ—For the **simple** math questions, we do NOT recommend to use the CoT prompt.οΌ‰
 
321
 
322
 
323
+ ```
324
+ "Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n### Instruction:\n{instruction}\n\n### Response: Let's think step by step."
325
  ```
326
 
327
+ ## Inference WizardMath Demo Script
328
+
329
+ We provide the WizardMath inference demo code [here](https://github.com/nlpxucan/WizardLM/tree/main/demo).
330
+
331
+
332
  ❗<b>To commen concern about dataset:</b>
333
 
334
+ Recently, there have been clear changes in the open-source policy and regulations of our overall organization's code, data, and models.
335
  Despite this, we have still worked hard to obtain opening the weights of the model first, but the data involves stricter auditing and is in review with our legal team .
336
  Our researchers have no authority to publicly release them without authorization.
337
  Thank you for your understanding.
338
+
339
+
340
+ ## Citation
341
+
342
+ Please cite the repo if you use the data, method or code in this repo.
343
+
344
+ ```
345
+ @article{luo2023wizardmath,
346
+ title={WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct},
347
+ author={Luo, Haipeng and Sun, Qingfeng and Xu, Can and Zhao, Pu and Lou, Jianguang and Tao, Chongyang and Geng, Xiubo and Lin, Qingwei and Chen, Shifeng and Zhang, Dongmei},
348
+ journal={arXiv preprint arXiv:2308.09583},
349
+ year={2023}
350
+ }
351
+ ```