TheBloke commited on
Commit
aecefb3
1 Parent(s): 911b6f1

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +67 -58
README.md CHANGED
@@ -1,6 +1,6 @@
1
  ---
2
  inference: false
3
- license: other
4
  model_creator: lmsys
5
  model_link: https://huggingface.co/lmsys/vicuna-13b-v1.5
6
  model_name: Vicuna 13B v1.5
@@ -29,57 +29,67 @@ quantized_by: TheBloke
29
  - Model creator: [lmsys](https://huggingface.co/lmsys)
30
  - Original model: [Vicuna 13B v1.5](https://huggingface.co/lmsys/vicuna-13b-v1.5)
31
 
 
32
  ## Description
33
 
34
  This repo contains GPTQ model files for [lmsys's Vicuna 13B v1.5](https://huggingface.co/lmsys/vicuna-13b-v1.5).
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/vicuna-13B-v1.5-GPTQ)
41
- * [2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GGML)
 
42
  * [lmsys's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/lmsys/vicuna-13b-v1.5)
 
43
 
 
44
  ## Prompt template: Vicuna
45
 
46
  ```
47
- A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.
48
 
49
- USER: {prompt}
50
- ASSISTANT:
51
  ```
52
 
 
 
 
53
  ## Provided files and GPTQ parameters
54
 
55
  Multiple quantisation parameters are provided, to allow you to choose the best one for your hardware and requirements.
56
 
57
  Each separate quant is in a different branch. See below for instructions on fetching from different branches.
58
 
59
- All GPTQ files are made with AutoGPTQ.
60
 
61
  <details>
62
  <summary>Explanation of GPTQ parameters</summary>
63
 
64
  - Bits: The bit size of the quantised model.
65
  - GS: GPTQ group size. Higher numbers use less VRAM, but have lower quantisation accuracy. "None" is the lowest possible value.
66
- - 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.
67
  - Damp %: A GPTQ parameter that affects how samples are processed for quantisation. 0.01 is default, but 0.1 results in slightly better accuracy.
68
- - GPTQ dataset: The dataset used for quantisation. The dataset used for quantisation can affect the quantisation accuracy. The dataset used for quantisation is not the same as the dataset used to train the model.
69
- - 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 affects the quantisation accuracy on longer inference sequences.
70
  - ExLlama Compatibility: Whether this file can be loaded with ExLlama, which currently only supports Llama models in 4-bit.
71
 
72
  </details>
73
 
74
  | Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
75
  | ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
76
- | [main](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/main) | 4 | 128 | No | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.26 GB | Yes | Most compatible option. Good inference speed in AutoGPTQ and GPTQ-for-LLaMa. Lower inference quality than other options. |
77
- | [gptq-4bit-32g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 8.00 GB | Yes | 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage. Poor AutoGPTQ CUDA speed. |
78
- | [gptq-4bit-64g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/gptq-4bit-64g-actorder_True) | 4 | 64 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.51 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. |
79
- | [gptq-4bit-128g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/gptq-4bit-128g-actorder_True) | 4 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.26 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. |
80
- | [gptq-8bit--1g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/gptq-8bit--1g-actorder_True) | 8 | None | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 13.36 GB | No | 8-bit, with Act Order. No group size, to lower VRAM requirements and to improve AutoGPTQ speed. |
81
  | [gptq-8bit-128g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/gptq-8bit-128g-actorder_True) | 8 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 13.65 GB | No | 8-bit, with group size 128g for higher inference quality and with Act Order for even higher accuracy. Poor AutoGPTQ CUDA speed. |
82
 
 
 
 
83
  ## How to download from branches
84
 
85
  - In text-generation-webui, you can add `:branch` to the end of the download name, eg `TheBloke/vicuna-13B-v1.5-GPTQ:gptq-4bit-32g-actorder_True`
@@ -88,78 +98,75 @@ All GPTQ files are made with AutoGPTQ.
88
  git clone --single-branch --branch gptq-4bit-32g-actorder_True https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ
89
  ```
90
  - In Python Transformers code, the branch is the `revision` parameter; see below.
91
-
 
92
  ## How to easily download and use this model in [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
93
 
94
  Please make sure you're using the latest version of [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
95
 
96
- It is strongly recommended to use the text-generation-webui one-click-installers unless you know how to make a manual install.
97
 
98
  1. Click the **Model tab**.
99
  2. Under **Download custom model or LoRA**, enter `TheBloke/vicuna-13B-v1.5-GPTQ`.
100
  - To download from a specific branch, enter for example `TheBloke/vicuna-13B-v1.5-GPTQ:gptq-4bit-32g-actorder_True`
101
  - see Provided Files above for the list of branches for each option.
102
  3. Click **Download**.
103
- 4. The model will start downloading. Once it's finished it will say "Done"
104
  5. In the top left, click the refresh icon next to **Model**.
105
  6. In the **Model** dropdown, choose the model you just downloaded: `vicuna-13B-v1.5-GPTQ`
106
  7. The model will automatically load, and is now ready for use!
107
  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.
108
- * Note that you do not need to set GPTQ parameters any more. These are set automatically from the file `quantize_config.json`.
109
  9. Once you're ready, click the **Text Generation tab** and enter a prompt to get started!
 
110
 
 
111
  ## How to use this GPTQ model from Python code
112
 
113
- First make sure you have [AutoGPTQ](https://github.com/PanQiWei/AutoGPTQ) 0.3.1 or later installed:
114
 
115
- ```
116
- pip3 install auto-gptq
117
- ```
118
 
119
- If you have problems installing AutoGPTQ, please build from source instead:
 
 
120
  ```
 
 
 
 
121
  pip3 uninstall -y auto-gptq
122
  git clone https://github.com/PanQiWei/AutoGPTQ
123
  cd AutoGPTQ
124
  pip3 install .
125
  ```
126
 
127
- Then try the following example code:
 
 
 
 
 
 
 
 
128
 
129
  ```python
130
- from transformers import AutoTokenizer, pipeline, logging
131
- from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
132
 
133
  model_name_or_path = "TheBloke/vicuna-13B-v1.5-GPTQ"
134
-
135
- use_triton = False
 
 
 
 
136
 
137
  tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
138
 
139
- model = AutoGPTQForCausalLM.from_quantized(model_name_or_path,
140
- use_safetensors=True,
141
- trust_remote_code=False,
142
- device="cuda:0",
143
- use_triton=use_triton,
144
- quantize_config=None)
145
-
146
- """
147
- # To download from a specific branch, use the revision parameter, as in this example:
148
- # Note that `revision` requires AutoGPTQ 0.3.1 or later!
149
-
150
- model = AutoGPTQForCausalLM.from_quantized(model_name_or_path,
151
- revision="gptq-4bit-32g-actorder_True",
152
- use_safetensors=True,
153
- trust_remote_code=False,
154
- device="cuda:0",
155
- quantize_config=None)
156
- """
157
-
158
  prompt = "Tell me about AI"
159
- prompt_template=f'''A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.
160
 
161
- USER: {prompt}
162
- ASSISTANT:
163
  '''
164
 
165
  print("\n\n*** Generate:")
@@ -170,9 +177,6 @@ print(tokenizer.decode(output[0]))
170
 
171
  # Inference can also be done using transformers' pipeline
172
 
173
- # Prevent printing spurious transformers error when using pipeline with AutoGPTQ
174
- logging.set_verbosity(logging.CRITICAL)
175
-
176
  print("*** Pipeline:")
177
  pipe = pipeline(
178
  "text-generation",
@@ -186,12 +190,17 @@ pipe = pipeline(
186
 
187
  print(pipe(prompt_template)[0]['generated_text'])
188
  ```
 
189
 
 
190
  ## Compatibility
191
 
192
- 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.
 
 
193
 
194
- ExLlama works with Llama models in 4-bit. Please see the Provided Files table above for per-file compatibility.
 
195
 
196
  <!-- footer start -->
197
  <!-- 200823 -->
@@ -216,7 +225,7 @@ Donaters will get priority support on any and all AI/LLM/model questions and req
216
 
217
  **Special thanks to**: Aemon Algiz.
218
 
219
- **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 Wikieł, Rishabh Srivastava, David Ziegler, Luke Pendergrass, Andrey, Gabriel Puliatti, Lone Striker, Sebastain Graf, Pierre Kircher, Randy H, NimbleBox.ai, Vadim, danny, Deo Leter
220
 
221
 
222
  Thank you to all my generous patrons and donaters!
@@ -236,7 +245,7 @@ Vicuna is a chat assistant trained by fine-tuning Llama 2 on user-shared convers
236
 
237
  - **Developed by:** [LMSYS](https://lmsys.org/)
238
  - **Model type:** An auto-regressive language model based on the transformer architecture
239
- - **License:** Llama 2 Community License Agreement
240
  - **Finetuned from model:** [Llama 2](https://arxiv.org/abs/2307.09288)
241
 
242
  ### Model Sources
@@ -254,7 +263,7 @@ The primary intended users of the model are researchers and hobbyists in natural
254
  ## How to Get Started with the Model
255
 
256
  - Command line interface: https://github.com/lm-sys/FastChat#vicuna-weights
257
- - APIs (OpenAI API, Huggingface API): https://github.com/lm-sys/FastChat/tree/main#api
258
 
259
  ## Training Details
260
 
 
1
  ---
2
  inference: false
3
+ license: llama2
4
  model_creator: lmsys
5
  model_link: https://huggingface.co/lmsys/vicuna-13b-v1.5
6
  model_name: Vicuna 13B v1.5
 
29
  - Model creator: [lmsys](https://huggingface.co/lmsys)
30
  - Original model: [Vicuna 13B v1.5](https://huggingface.co/lmsys/vicuna-13b-v1.5)
31
 
32
+ <!-- description start -->
33
  ## Description
34
 
35
  This repo contains GPTQ model files for [lmsys's Vicuna 13B v1.5](https://huggingface.co/lmsys/vicuna-13b-v1.5).
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/vicuna-13B-v1.5-GPTQ)
44
+ * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GGUF)
45
+ * [2, 3, 4, 5, 6 and 8-bit GGML models for CPU+GPU inference (deprecated)](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GGML)
46
  * [lmsys's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/lmsys/vicuna-13b-v1.5)
47
+ <!-- repositories-available end -->
48
 
49
+ <!-- prompt-template start -->
50
  ## Prompt template: Vicuna
51
 
52
  ```
53
+ A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions. USER: {prompt} ASSISTANT:
54
 
 
 
55
  ```
56
 
57
+ <!-- prompt-template end -->
58
+
59
+ <!-- README_GPTQ.md-provided-files start -->
60
  ## Provided files and GPTQ parameters
61
 
62
  Multiple quantisation parameters are provided, to allow you to choose the best one for your hardware and requirements.
63
 
64
  Each separate quant is in a different branch. See below for instructions on fetching from different branches.
65
 
66
+ 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.
67
 
68
  <details>
69
  <summary>Explanation of GPTQ parameters</summary>
70
 
71
  - Bits: The bit size of the quantised model.
72
  - GS: GPTQ group size. Higher numbers use less VRAM, but have lower quantisation accuracy. "None" is the lowest possible value.
73
+ - 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.
74
  - Damp %: A GPTQ parameter that affects how samples are processed for quantisation. 0.01 is default, but 0.1 results in slightly better accuracy.
75
+ - 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).
76
+ - 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
  - ExLlama Compatibility: Whether this file can be loaded with ExLlama, which currently only supports Llama models in 4-bit.
78
 
79
  </details>
80
 
81
  | Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
82
  | ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
83
+ | [main](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/main) | 4 | 128 | No | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.26 GB | Yes | Most compatible option. Good inference speed in AutoGPTQ and GPTQ-for-LLaMa. Lower inference quality than other options. |
84
+ | [gptq-4bit-32g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 8.00 GB | Yes | 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage. Poor AutoGPTQ CUDA speed. |
85
+ | [gptq-4bit-64g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/gptq-4bit-64g-actorder_True) | 4 | 64 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.51 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. |
86
+ | [gptq-4bit-128g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/gptq-4bit-128g-actorder_True) | 4 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.26 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. |
87
+ | [gptq-8bit--1g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/gptq-8bit--1g-actorder_True) | 8 | None | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 13.36 GB | No | 8-bit, with Act Order. No group size, to lower VRAM requirements and to improve AutoGPTQ speed. |
88
  | [gptq-8bit-128g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ/tree/gptq-8bit-128g-actorder_True) | 8 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 13.65 GB | No | 8-bit, with group size 128g for higher inference quality and with Act Order for even higher accuracy. Poor AutoGPTQ CUDA speed. |
89
 
90
+ <!-- README_GPTQ.md-provided-files end -->
91
+
92
+ <!-- README_GPTQ.md-download-from-branches start -->
93
  ## How to download from branches
94
 
95
  - In text-generation-webui, you can add `:branch` to the end of the download name, eg `TheBloke/vicuna-13B-v1.5-GPTQ:gptq-4bit-32g-actorder_True`
 
98
  git clone --single-branch --branch gptq-4bit-32g-actorder_True https://huggingface.co/TheBloke/vicuna-13B-v1.5-GPTQ
99
  ```
100
  - In Python Transformers code, the branch is the `revision` parameter; see below.
101
+ <!-- README_GPTQ.md-download-from-branches end -->
102
+ <!-- README_GPTQ.md-text-generation-webui start -->
103
  ## How to easily download and use this model in [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
104
 
105
  Please make sure you're using the latest version of [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
106
 
107
+ 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.
108
 
109
  1. Click the **Model tab**.
110
  2. Under **Download custom model or LoRA**, enter `TheBloke/vicuna-13B-v1.5-GPTQ`.
111
  - To download from a specific branch, enter for example `TheBloke/vicuna-13B-v1.5-GPTQ:gptq-4bit-32g-actorder_True`
112
  - see Provided Files above for the list of branches for each option.
113
  3. Click **Download**.
114
+ 4. The model will start downloading. Once it's finished it will say "Done".
115
  5. In the top left, click the refresh icon next to **Model**.
116
  6. In the **Model** dropdown, choose the model you just downloaded: `vicuna-13B-v1.5-GPTQ`
117
  7. The model will automatically load, and is now ready for use!
118
  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.
119
+ * 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`.
120
  9. Once you're ready, click the **Text Generation tab** and enter a prompt to get started!
121
+ <!-- README_GPTQ.md-text-generation-webui end -->
122
 
123
+ <!-- README_GPTQ.md-use-from-python start -->
124
  ## How to use this GPTQ model from Python code
125
 
126
+ ### Install the necessary packages
127
 
128
+ Requires: Transformers 4.32.0 or later, Optimum 1.12.0 or later, and AutoGPTQ 0.4.2 or later.
 
 
129
 
130
+ ```shell
131
+ pip3 install transformers>=4.32.0 optimum>=1.12.0
132
+ pip3 install auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/ # Use cu117 if on CUDA 11.7
133
  ```
134
+
135
+ If you have problems installing AutoGPTQ using the pre-built wheels, install it from source instead:
136
+
137
+ ```shell
138
  pip3 uninstall -y auto-gptq
139
  git clone https://github.com/PanQiWei/AutoGPTQ
140
  cd AutoGPTQ
141
  pip3 install .
142
  ```
143
 
144
+ ### For CodeLlama models only: you must use Transformers 4.33.0 or later.
145
+
146
+ If 4.33.0 is not yet released when you read this, you will need to install Transformers from source:
147
+ ```shell
148
+ pip3 uninstall -y transformers
149
+ pip3 install git+https://github.com/huggingface/transformers.git
150
+ ```
151
+
152
+ ### You can then use the following code
153
 
154
  ```python
155
+ from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 
156
 
157
  model_name_or_path = "TheBloke/vicuna-13B-v1.5-GPTQ"
158
+ # To use a different branch, change revision
159
+ # For example: revision="gptq-4bit-32g-actorder_True"
160
+ model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
161
+ torch_dtype=torch.float16,
162
+ device_map="auto",
163
+ revision="main")
164
 
165
  tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
166
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
167
  prompt = "Tell me about AI"
168
+ prompt_template=f'''A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions. USER: {prompt} ASSISTANT:
169
 
 
 
170
  '''
171
 
172
  print("\n\n*** Generate:")
 
177
 
178
  # Inference can also be done using transformers' pipeline
179
 
 
 
 
180
  print("*** Pipeline:")
181
  pipe = pipeline(
182
  "text-generation",
 
190
 
191
  print(pipe(prompt_template)[0]['generated_text'])
192
  ```
193
+ <!-- README_GPTQ.md-use-from-python end -->
194
 
195
+ <!-- README_GPTQ.md-compatibility start -->
196
  ## Compatibility
197
 
198
+ 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).
199
+
200
+ [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.
201
 
202
+ [Huggingface Text Generation Inference (TGI)](https://github.com/huggingface/text-generation-inference) is compatible with all GPTQ models.
203
+ <!-- README_GPTQ.md-compatibility end -->
204
 
205
  <!-- footer start -->
206
  <!-- 200823 -->
 
225
 
226
  **Special thanks to**: Aemon Algiz.
227
 
228
+ **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 Wikieł, 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 Bjäreholt, 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
229
 
230
 
231
  Thank you to all my generous patrons and donaters!
 
245
 
246
  - **Developed by:** [LMSYS](https://lmsys.org/)
247
  - **Model type:** An auto-regressive language model based on the transformer architecture
248
+ - **License:** Llama 2 Community License Agreement
249
  - **Finetuned from model:** [Llama 2](https://arxiv.org/abs/2307.09288)
250
 
251
  ### Model Sources
 
263
  ## How to Get Started with the Model
264
 
265
  - Command line interface: https://github.com/lm-sys/FastChat#vicuna-weights
266
+ - APIs (OpenAI API, Huggingface API): https://github.com/lm-sys/FastChat/tree/main#api
267
 
268
  ## Training Details
269