TheBloke commited on
Commit
ff34cb0
1 Parent(s): 9d0e3e5

Update for Transformers GPTQ support

Browse files
README.md CHANGED
@@ -9,17 +9,20 @@ quantized_by: TheBloke
9
  ---
10
 
11
  <!-- header start -->
12
- <div style="width: 100%;">
13
- <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
 
14
  </div>
15
  <div style="display: flex; justify-content: space-between; width: 100%;">
16
  <div style="display: flex; flex-direction: column; align-items: flex-start;">
17
- <p><a href="https://discord.gg/theblokeai">Chat & support: my new Discord server</a></p>
18
  </div>
19
  <div style="display: flex; flex-direction: column; align-items: flex-end;">
20
- <p><a href="https://www.patreon.com/TheBlokeAI">Want to contribute? TheBloke's Patreon page</a></p>
21
  </div>
22
  </div>
 
 
23
  <!-- header end -->
24
 
25
  # Vicuna 13B v1.5 16K - GPTQ
@@ -70,13 +73,13 @@ All GPTQ files are made with AutoGPTQ.
70
 
71
  | Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
72
  | ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
73
- | [main](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/main) | 4 | 128 | No | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 7.26 GB | Yes | Most compatible option. Good inference speed in AutoGPTQ and GPTQ-for-LLaMa. Lower inference quality than other options. |
74
- | [gptq-4bit-32g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 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. |
75
- | [gptq-4bit-64g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/gptq-4bit-64g-actorder_True) | 4 | 64 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 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. |
76
- | [gptq-4bit-128g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/gptq-4bit-128g-actorder_True) | 4 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 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. |
77
- | [gptq-8bit--1g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/gptq-8bit--1g-actorder_True) | 8 | None | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 13.36 GB | No | 8-bit, with Act Order. No group size, to lower VRAM requirements and to improve AutoGPTQ speed. |
78
- | [gptq-8bit-64g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/gptq-8bit-64g-actorder_True) | 8 | 64 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 13.95 GB | No | 8-bit, with group size 64g and Act Order for even higher inference quality. Poor AutoGPTQ CUDA speed. |
79
- | [gptq-8bit-32g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/gptq-8bit-32g-actorder_True) | 8 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 14.54 GB | No | 8-bit, with group size 32g and Act Order for maximum inference quality. Poor AutoGPTQ CUDA speed. |
80
  | [gptq-8bit-128g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/gptq-8bit-128g-actorder_True) | 8 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 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. |
81
 
82
  ## How to download from branches
@@ -193,6 +196,7 @@ The files provided will work with AutoGPTQ (CUDA and Triton modes), GPTQ-for-LLa
193
  ExLlama works with Llama models in 4-bit. Please see the Provided Files table above for per-file compatibility.
194
 
195
  <!-- footer start -->
 
196
  ## Discord
197
 
198
  For further support, and discussions on these models and AI in general, join us at:
@@ -212,13 +216,15 @@ Donaters will get priority support on any and all AI/LLM/model questions and req
212
  * Patreon: https://patreon.com/TheBlokeAI
213
  * Ko-Fi: https://ko-fi.com/TheBlokeAI
214
 
215
- **Special thanks to**: Luke from CarbonQuill, Aemon Algiz.
216
 
217
- **Patreon special mentions**: Willem Michiel, Ajan Kanaga, Cory Kujawski, Alps Aficionado, Nikolai Manek, Jonathan Leane, Stanislav Ovsiannikov, Michael Levine, Luke Pendergrass, Sid, K, Gabriel Tamborski, Clay Pascal, Kalila, William Sang, Will Dee, Pieter, Nathan LeClaire, ya boyyy, David Flickinger, vamX, Derek Yates, Fen Risland, Jeffrey Morgan, webtim, Daniel P. Andersen, Chadd, Edmond Seymore, Pyrater, Olusegun Samson, Lone Striker, biorpg, alfie_i, Mano Prime, Chris Smitley, Dave, zynix, Trenton Dambrowitz, Johann-Peter Hartmann, Magnesian, Spencer Kim, John Detwiler, Iucharbius, Gabriel Puliatti, LangChain4j, Luke @flexchar, Vadim, Rishabh Srivastava, Preetika Verma, Ai Maven, Femi Adebogun, WelcomeToTheClub, Leonard Tan, Imad Khwaja, Steven Wood, Stefan Sabev, Sebastain Graf, usrbinkat, Dan Guido, Sam, Eugene Pentland, Mandus, transmissions 11, Slarti, Karl Bernard, Spiking Neurons AB, Artur Olbinski, Joseph William Delisle, ReadyPlayerEmma, Olakabola, Asp the Wyvern, Space Cruiser, Matthew Berman, Randy H, subjectnull, danny, John Villwock, Illia Dulskyi, Rainer Wilmers, theTransient, Pierre Kircher, Alexandros Triantafyllidis, Viktor Bowallius, terasurfer, Deep Realms, SuperWojo, senxiiz, Oscar Rangel, Alex, Stephen Murray, Talal Aujan, Raven Klaugh, Sean Connelly, Raymond Fosdick, Fred von Graf, chris gileta, Junyu Yang, Elle
218
 
219
 
220
  Thank you to all my generous patrons and donaters!
221
 
 
 
222
  <!-- footer end -->
223
 
224
  # Original model card: lmsys's Vicuna 13B v1.5 16K
@@ -232,7 +238,7 @@ Vicuna is a chat assistant trained by fine-tuning Llama 2 on user-shared convers
232
 
233
  - **Developed by:** [LMSYS](https://lmsys.org/)
234
  - **Model type:** An auto-regressive language model based on the transformer architecture
235
- - **License:** Llama 2 Community License Agreement
236
  - **Finetuned from model:** [Llama 2](https://arxiv.org/abs/2307.09288)
237
 
238
  ### Model Sources
@@ -250,7 +256,7 @@ The primary intended users of the model are researchers and hobbyists in natural
250
  ## How to Get Started with the Model
251
 
252
  - Command line interface: https://github.com/lm-sys/FastChat#vicuna-weights
253
- - APIs (OpenAI API, Huggingface API): https://github.com/lm-sys/FastChat/tree/main#api
254
 
255
  ## Training Details
256
 
 
9
  ---
10
 
11
  <!-- header start -->
12
+ <!-- 200823 -->
13
+ <div style="width: auto; margin-left: auto; margin-right: auto">
14
+ <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
15
  </div>
16
  <div style="display: flex; justify-content: space-between; width: 100%;">
17
  <div style="display: flex; flex-direction: column; align-items: flex-start;">
18
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>
19
  </div>
20
  <div style="display: flex; flex-direction: column; align-items: flex-end;">
21
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://www.patreon.com/TheBlokeAI">Want to contribute? TheBloke's Patreon page</a></p>
22
  </div>
23
  </div>
24
+ <div style="text-align:center; margin-top: 0em; margin-bottom: 0em"><p style="margin-top: 0.25em; margin-bottom: 0em;">TheBloke's LLM work is generously supported by a grant from <a href="https://a16z.com">andreessen horowitz (a16z)</a></p></div>
25
+ <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
26
  <!-- header end -->
27
 
28
  # Vicuna 13B v1.5 16K - GPTQ
 
73
 
74
  | Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
75
  | ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
76
+ | [main](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/main) | 4 | 128 | No | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 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-v1.5-16K-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 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-v1.5-16K-GPTQ/tree/gptq-4bit-64g-actorder_True) | 4 | 64 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 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-v1.5-16K-GPTQ/tree/gptq-4bit-128g-actorder_True) | 4 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 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-v1.5-16K-GPTQ/tree/gptq-8bit--1g-actorder_True) | 8 | None | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 13.36 GB | No | 8-bit, with Act Order. No group size, to lower VRAM requirements and to improve AutoGPTQ speed. |
81
+ | [gptq-8bit-64g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/gptq-8bit-64g-actorder_True) | 8 | 64 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 13.95 GB | No | 8-bit, with group size 64g and Act Order for even higher inference quality. Poor AutoGPTQ CUDA speed. |
82
+ | [gptq-8bit-32g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/gptq-8bit-32g-actorder_True) | 8 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 14.54 GB | No | 8-bit, with group size 32g and Act Order for maximum inference quality. Poor AutoGPTQ CUDA speed. |
83
  | [gptq-8bit-128g-actorder_True](https://huggingface.co/TheBloke/vicuna-13B-v1.5-16K-GPTQ/tree/gptq-8bit-128g-actorder_True) | 8 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 8192 | 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. |
84
 
85
  ## How to download from branches
 
196
  ExLlama works with Llama models in 4-bit. Please see the Provided Files table above for per-file compatibility.
197
 
198
  <!-- footer start -->
199
+ <!-- 200823 -->
200
  ## Discord
201
 
202
  For further support, and discussions on these models and AI in general, join us at:
 
216
  * Patreon: https://patreon.com/TheBlokeAI
217
  * Ko-Fi: https://ko-fi.com/TheBlokeAI
218
 
219
+ **Special thanks to**: Aemon Algiz.
220
 
221
+ **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
222
 
223
 
224
  Thank you to all my generous patrons and donaters!
225
 
226
+ And thank you again to a16z for their generous grant.
227
+
228
  <!-- footer end -->
229
 
230
  # Original model card: lmsys's Vicuna 13B v1.5 16K
 
238
 
239
  - **Developed by:** [LMSYS](https://lmsys.org/)
240
  - **Model type:** An auto-regressive language model based on the transformer architecture
241
+ - **License:** Llama 2 Community License Agreement
242
  - **Finetuned from model:** [Llama 2](https://arxiv.org/abs/2307.09288)
243
 
244
  ### Model Sources
 
256
  ## How to Get Started with the Model
257
 
258
  - Command line interface: https://github.com/lm-sys/FastChat#vicuna-weights
259
+ - APIs (OpenAI API, Huggingface API): https://github.com/lm-sys/FastChat/tree/main#api
260
 
261
  ## Training Details
262
 
config.json CHANGED
@@ -1,30 +1,41 @@
1
  {
2
- "_name_or_path": "vicuna-13b-v1.5-16k/",
3
- "architectures": [
4
- "LlamaForCausalLM"
5
- ],
6
- "bos_token_id": 1,
7
- "eos_token_id": 2,
8
- "hidden_act": "silu",
9
- "hidden_size": 5120,
10
- "initializer_range": 0.02,
11
- "intermediate_size": 13824,
12
- "max_position_embeddings": 4096,
13
- "max_sequence_length": 16384,
14
- "model_type": "llama",
15
- "num_attention_heads": 40,
16
- "num_hidden_layers": 40,
17
- "num_key_value_heads": 40,
18
- "pad_token_id": 0,
19
- "pretraining_tp": 1,
20
- "rms_norm_eps": 1e-05,
21
- "rope_scaling": {
22
- "factor": 4.0,
23
- "type": "linear"
24
- },
25
- "tie_word_embeddings": false,
26
- "torch_dtype": "float16",
27
- "transformers_version": "4.31.0",
28
- "use_cache": true,
29
- "vocab_size": 32000
 
 
 
 
 
 
 
 
 
 
 
30
  }
 
1
  {
2
+ "_name_or_path": "vicuna-13b-v1.5-16k/",
3
+ "architectures": [
4
+ "LlamaForCausalLM"
5
+ ],
6
+ "bos_token_id": 1,
7
+ "eos_token_id": 2,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 5120,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 13824,
12
+ "max_position_embeddings": 4096,
13
+ "max_sequence_length": 16384,
14
+ "model_type": "llama",
15
+ "num_attention_heads": 40,
16
+ "num_hidden_layers": 40,
17
+ "num_key_value_heads": 40,
18
+ "pad_token_id": 0,
19
+ "pretraining_tp": 1,
20
+ "rms_norm_eps": 1e-05,
21
+ "rope_scaling": {
22
+ "factor": 4.0,
23
+ "type": "linear"
24
+ },
25
+ "tie_word_embeddings": false,
26
+ "torch_dtype": "float16",
27
+ "transformers_version": "4.31.0",
28
+ "use_cache": true,
29
+ "vocab_size": 32000,
30
+ "quantization_config": {
31
+ "bits": 4,
32
+ "group_size": 128,
33
+ "damp_percent": 0.1,
34
+ "desc_act": false,
35
+ "sym": true,
36
+ "true_sequential": true,
37
+ "model_name_or_path": null,
38
+ "model_file_base_name": "model",
39
+ "quant_method": "gptq"
40
+ }
41
  }
gptq_model-4bit-128g.safetensors → model.safetensors RENAMED
File without changes
quantize_config.json CHANGED
@@ -6,5 +6,5 @@
6
  "sym": true,
7
  "true_sequential": true,
8
  "model_name_or_path": null,
9
- "model_file_base_name": null
10
  }
 
6
  "sym": true,
7
  "true_sequential": true,
8
  "model_name_or_path": null,
9
+ "model_file_base_name": "model"
10
  }