TheBloke commited on
Commit
fc1bd57
1 Parent(s): 97bee93

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +478 -0
README.md ADDED
@@ -0,0 +1,478 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: NumbersStation/nsql-llama-2-7B
3
+ inference: false
4
+ license: llama2
5
+ model_creator: NumbersStation
6
+ model_name: NSQL Llama-2 7B
7
+ model_type: llama
8
+ prompt_template: '{prompt}
9
+
10
+
11
+ SELECT
12
+
13
+ '
14
+ quantized_by: TheBloke
15
+ widget:
16
+ - example_title: Number stadiums
17
+ text: "CREATE TABLE stadium (\n stadium_id number,\n location text,\n name\
18
+ \ text,\n capacity number,\n)\n\n-- Using valid SQLite, answer the following\
19
+ \ questions for the tables provided above.\n\n-- how many stadiums in total?\n\
20
+ \nSELECT"
21
+ - example_title: Open work orders
22
+ text: 'CREATE TABLE work_orders ( ID NUMBER, CREATED_AT TEXT, COST FLOAT, INVOICE_AMOUNT
23
+ FLOAT, IS_DUE BOOLEAN, IS_OPEN BOOLEAN, IS_OVERDUE BOOLEAN, COUNTRY_NAME TEXT,
24
+ )
25
+
26
+
27
+ -- Using valid SQLite, answer the following questions for the tables provided
28
+ above.
29
+
30
+
31
+ -- how many work orders are open?
32
+
33
+
34
+ SELECT'
35
+ - example_title: Stadium capacity
36
+ text: 'CREATE TABLE stadium ( stadium_id number, location text, name text, capacity
37
+ number, highest number, lowest number, average number )
38
+
39
+
40
+ CREATE TABLE singer ( singer_id number, name text, country text, song_name text,
41
+ song_release_year text, age number, is_male others )
42
+
43
+
44
+ CREATE TABLE concert ( concert_id number, concert_name text, theme text, stadium_id
45
+ text, year text )
46
+
47
+
48
+ CREATE TABLE singer_in_concert ( concert_id number, singer_id text )
49
+
50
+
51
+ -- Using valid SQLite, answer the following questions for the tables provided
52
+ above.
53
+
54
+
55
+ -- What is the maximum, the average, and the minimum capacity of stadiums ?
56
+
57
+
58
+ SELECT'
59
+ ---
60
+ <!-- markdownlint-disable MD041 -->
61
+
62
+ <!-- header start -->
63
+ <!-- 200823 -->
64
+ <div style="width: auto; margin-left: auto; margin-right: auto">
65
+ <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
66
+ </div>
67
+ <div style="display: flex; justify-content: space-between; width: 100%;">
68
+ <div style="display: flex; flex-direction: column; align-items: flex-start;">
69
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>
70
+ </div>
71
+ <div style="display: flex; flex-direction: column; align-items: flex-end;">
72
+ <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>
73
+ </div>
74
+ </div>
75
+ <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>
76
+ <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
77
+ <!-- header end -->
78
+
79
+ # NSQL Llama-2 7B - GGUF
80
+ - Model creator: [NumbersStation](https://huggingface.co/NumbersStation)
81
+ - Original model: [NSQL Llama-2 7B](https://huggingface.co/NumbersStation/nsql-llama-2-7B)
82
+
83
+ <!-- description start -->
84
+ ## Description
85
+
86
+ This repo contains GGUF format model files for [NumbersStation's NSQL Llama-2 7B](https://huggingface.co/NumbersStation/nsql-llama-2-7B).
87
+
88
+ These files were quantised using hardware kindly provided by [Massed Compute](https://massedcompute.com/).
89
+
90
+ <!-- description end -->
91
+ <!-- README_GGUF.md-about-gguf start -->
92
+ ### About GGUF
93
+
94
+ GGUF is a new format introduced by the llama.cpp team on August 21st 2023. It is a replacement for GGML, which is no longer supported by llama.cpp.
95
+
96
+ Here is an incomplete list of clients and libraries that are known to support GGUF:
97
+
98
+ * [llama.cpp](https://github.com/ggerganov/llama.cpp). The source project for GGUF. Offers a CLI and a server option.
99
+ * [text-generation-webui](https://github.com/oobabooga/text-generation-webui), the most widely used web UI, with many features and powerful extensions. Supports GPU acceleration.
100
+ * [KoboldCpp](https://github.com/LostRuins/koboldcpp), a fully featured web UI, with GPU accel across all platforms and GPU architectures. Especially good for story telling.
101
+ * [LM Studio](https://lmstudio.ai/), an easy-to-use and powerful local GUI for Windows and macOS (Silicon), with GPU acceleration.
102
+ * [LoLLMS Web UI](https://github.com/ParisNeo/lollms-webui), a great web UI with many interesting and unique features, including a full model library for easy model selection.
103
+ * [Faraday.dev](https://faraday.dev/), an attractive and easy to use character-based chat GUI for Windows and macOS (both Silicon and Intel), with GPU acceleration.
104
+ * [ctransformers](https://github.com/marella/ctransformers), a Python library with GPU accel, LangChain support, and OpenAI-compatible AI server.
105
+ * [llama-cpp-python](https://github.com/abetlen/llama-cpp-python), a Python library with GPU accel, LangChain support, and OpenAI-compatible API server.
106
+ * [candle](https://github.com/huggingface/candle), a Rust ML framework with a focus on performance, including GPU support, and ease of use.
107
+
108
+ <!-- README_GGUF.md-about-gguf end -->
109
+ <!-- repositories-available start -->
110
+ ## Repositories available
111
+
112
+ * [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/nsql-llama-2-7B-AWQ)
113
+ * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/nsql-llama-2-7B-GPTQ)
114
+ * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF)
115
+ * [NumbersStation's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/NumbersStation/nsql-llama-2-7B)
116
+ <!-- repositories-available end -->
117
+
118
+ <!-- prompt-template start -->
119
+ ## Prompt template: nsql
120
+
121
+ ```
122
+ {prompt}
123
+
124
+ SELECT
125
+
126
+ ```
127
+
128
+ <!-- prompt-template end -->
129
+
130
+
131
+ <!-- compatibility_gguf start -->
132
+ ## Compatibility
133
+
134
+ These quantised GGUFv2 files are compatible with llama.cpp from August 27th onwards, as of commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221)
135
+
136
+ They are also compatible with many third party UIs and libraries - please see the list at the top of this README.
137
+
138
+ ## Explanation of quantisation methods
139
+
140
+ <details>
141
+ <summary>Click to see details</summary>
142
+
143
+ The new methods available are:
144
+
145
+ * GGML_TYPE_Q2_K - "type-1" 2-bit quantization in super-blocks containing 16 blocks, each block having 16 weight. Block scales and mins are quantized with 4 bits. This ends up effectively using 2.5625 bits per weight (bpw)
146
+ * GGML_TYPE_Q3_K - "type-0" 3-bit quantization in super-blocks containing 16 blocks, each block having 16 weights. Scales are quantized with 6 bits. This end up using 3.4375 bpw.
147
+ * GGML_TYPE_Q4_K - "type-1" 4-bit quantization in super-blocks containing 8 blocks, each block having 32 weights. Scales and mins are quantized with 6 bits. This ends up using 4.5 bpw.
148
+ * GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw
149
+ * GGML_TYPE_Q6_K - "type-0" 6-bit quantization. Super-blocks with 16 blocks, each block having 16 weights. Scales are quantized with 8 bits. This ends up using 6.5625 bpw
150
+
151
+ Refer to the Provided Files table below to see what files use which methods, and how.
152
+ </details>
153
+ <!-- compatibility_gguf end -->
154
+
155
+ <!-- README_GGUF.md-provided-files start -->
156
+ ## Provided files
157
+
158
+ | Name | Quant method | Bits | Size | Max RAM required | Use case |
159
+ | ---- | ---- | ---- | ---- | ---- | ----- |
160
+ | [nsql-llama-2-7b.Q2_K.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q2_K.gguf) | Q2_K | 2 | 2.83 GB| 5.33 GB | smallest, significant quality loss - not recommended for most purposes |
161
+ | [nsql-llama-2-7b.Q3_K_S.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q3_K_S.gguf) | Q3_K_S | 3 | 2.95 GB| 5.45 GB | very small, high quality loss |
162
+ | [nsql-llama-2-7b.Q3_K_M.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q3_K_M.gguf) | Q3_K_M | 3 | 3.30 GB| 5.80 GB | very small, high quality loss |
163
+ | [nsql-llama-2-7b.Q3_K_L.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q3_K_L.gguf) | Q3_K_L | 3 | 3.60 GB| 6.10 GB | small, substantial quality loss |
164
+ | [nsql-llama-2-7b.Q4_0.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q4_0.gguf) | Q4_0 | 4 | 3.83 GB| 6.33 GB | legacy; small, very high quality loss - prefer using Q3_K_M |
165
+ | [nsql-llama-2-7b.Q4_K_S.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q4_K_S.gguf) | Q4_K_S | 4 | 3.86 GB| 6.36 GB | small, greater quality loss |
166
+ | [nsql-llama-2-7b.Q4_K_M.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q4_K_M.gguf) | Q4_K_M | 4 | 4.08 GB| 6.58 GB | medium, balanced quality - recommended |
167
+ | [nsql-llama-2-7b.Q5_0.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q5_0.gguf) | Q5_0 | 5 | 4.65 GB| 7.15 GB | legacy; medium, balanced quality - prefer using Q4_K_M |
168
+ | [nsql-llama-2-7b.Q5_K_S.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q5_K_S.gguf) | Q5_K_S | 5 | 4.65 GB| 7.15 GB | large, low quality loss - recommended |
169
+ | [nsql-llama-2-7b.Q5_K_M.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q5_K_M.gguf) | Q5_K_M | 5 | 4.78 GB| 7.28 GB | large, very low quality loss - recommended |
170
+ | [nsql-llama-2-7b.Q6_K.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q6_K.gguf) | Q6_K | 6 | 5.53 GB| 8.03 GB | very large, extremely low quality loss |
171
+ | [nsql-llama-2-7b.Q8_0.gguf](https://huggingface.co/TheBloke/nsql-llama-2-7B-GGUF/blob/main/nsql-llama-2-7b.Q8_0.gguf) | Q8_0 | 8 | 7.16 GB| 9.66 GB | very large, extremely low quality loss - not recommended |
172
+
173
+ **Note**: the above RAM figures assume no GPU offloading. If layers are offloaded to the GPU, this will reduce RAM usage and use VRAM instead.
174
+
175
+
176
+
177
+ <!-- README_GGUF.md-provided-files end -->
178
+
179
+ <!-- README_GGUF.md-how-to-download start -->
180
+ ## How to download GGUF files
181
+
182
+ **Note for manual downloaders:** You almost never want to clone the entire repo! Multiple different quantisation formats are provided, and most users only want to pick and download a single file.
183
+
184
+ The following clients/libraries will automatically download models for you, providing a list of available models to choose from:
185
+
186
+ * LM Studio
187
+ * LoLLMS Web UI
188
+ * Faraday.dev
189
+
190
+ ### In `text-generation-webui`
191
+
192
+ Under Download Model, you can enter the model repo: TheBloke/nsql-llama-2-7B-GGUF and below it, a specific filename to download, such as: nsql-llama-2-7b.Q4_K_M.gguf.
193
+
194
+ Then click Download.
195
+
196
+ ### On the command line, including multiple files at once
197
+
198
+ I recommend using the `huggingface-hub` Python library:
199
+
200
+ ```shell
201
+ pip3 install huggingface-hub
202
+ ```
203
+
204
+ Then you can download any individual model file to the current directory, at high speed, with a command like this:
205
+
206
+ ```shell
207
+ huggingface-cli download TheBloke/nsql-llama-2-7B-GGUF nsql-llama-2-7b.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
208
+ ```
209
+
210
+ <details>
211
+ <summary>More advanced huggingface-cli download usage</summary>
212
+
213
+ You can also download multiple files at once with a pattern:
214
+
215
+ ```shell
216
+ huggingface-cli download TheBloke/nsql-llama-2-7B-GGUF --local-dir . --local-dir-use-symlinks False --include='*Q4_K*gguf'
217
+ ```
218
+
219
+ For more documentation on downloading with `huggingface-cli`, please see: [HF -> Hub Python Library -> Download files -> Download from the CLI](https://huggingface.co/docs/huggingface_hub/guides/download#download-from-the-cli).
220
+
221
+ To accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`:
222
+
223
+ ```shell
224
+ pip3 install hf_transfer
225
+ ```
226
+
227
+ And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:
228
+
229
+ ```shell
230
+ HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/nsql-llama-2-7B-GGUF nsql-llama-2-7b.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False
231
+ ```
232
+
233
+ Windows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command.
234
+ </details>
235
+ <!-- README_GGUF.md-how-to-download end -->
236
+
237
+ <!-- README_GGUF.md-how-to-run start -->
238
+ ## Example `llama.cpp` command
239
+
240
+ Make sure you are using `llama.cpp` from commit [d0cee0d](https://github.com/ggerganov/llama.cpp/commit/d0cee0d36d5be95a0d9088b674dbb27354107221) or later.
241
+
242
+ ```shell
243
+ ./main -ngl 32 -m nsql-llama-2-7b.Q4_K_M.gguf --color -c 4096 --temp 0.7 --repeat_penalty 1.1 -n -1 -p "{prompt}\n\nSELECT"
244
+ ```
245
+
246
+ Change `-ngl 32` to the number of layers to offload to GPU. Remove it if you don't have GPU acceleration.
247
+
248
+ Change `-c 4096` to the desired sequence length. For extended sequence models - eg 8K, 16K, 32K - the necessary RoPE scaling parameters are read from the GGUF file and set by llama.cpp automatically.
249
+
250
+ If you want to have a chat-style conversation, replace the `-p <PROMPT>` argument with `-i -ins`
251
+
252
+ For other parameters and how to use them, please refer to [the llama.cpp documentation](https://github.com/ggerganov/llama.cpp/blob/master/examples/main/README.md)
253
+
254
+ ## How to run in `text-generation-webui`
255
+
256
+ Further instructions can be found in the text-generation-webui documentation, here: [text-generation-webui/docs/04 ‐ Model Tab.md](https://github.com/oobabooga/text-generation-webui/blob/main/docs/04%20%E2%80%90%20Model%20Tab.md#llamacpp).
257
+
258
+ ## How to run from Python code
259
+
260
+ You can use GGUF models from Python using the [llama-cpp-python](https://github.com/abetlen/llama-cpp-python) or [ctransformers](https://github.com/marella/ctransformers) libraries.
261
+
262
+ ### How to load this model in Python code, using ctransformers
263
+
264
+ #### First install the package
265
+
266
+ Run one of the following commands, according to your system:
267
+
268
+ ```shell
269
+ # Base ctransformers with no GPU acceleration
270
+ pip install ctransformers
271
+ # Or with CUDA GPU acceleration
272
+ pip install ctransformers[cuda]
273
+ # Or with AMD ROCm GPU acceleration (Linux only)
274
+ CT_HIPBLAS=1 pip install ctransformers --no-binary ctransformers
275
+ # Or with Metal GPU acceleration for macOS systems only
276
+ CT_METAL=1 pip install ctransformers --no-binary ctransformers
277
+ ```
278
+
279
+ #### Simple ctransformers example code
280
+
281
+ ```python
282
+ from ctransformers import AutoModelForCausalLM
283
+
284
+ # Set gpu_layers to the number of layers to offload to GPU. Set to 0 if no GPU acceleration is available on your system.
285
+ llm = AutoModelForCausalLM.from_pretrained("TheBloke/nsql-llama-2-7B-GGUF", model_file="nsql-llama-2-7b.Q4_K_M.gguf", model_type="llama", gpu_layers=50)
286
+
287
+ print(llm("AI is going to"))
288
+ ```
289
+
290
+ ## How to use with LangChain
291
+
292
+ Here are guides on using llama-cpp-python and ctransformers with LangChain:
293
+
294
+ * [LangChain + llama-cpp-python](https://python.langchain.com/docs/integrations/llms/llamacpp)
295
+ * [LangChain + ctransformers](https://python.langchain.com/docs/integrations/providers/ctransformers)
296
+
297
+ <!-- README_GGUF.md-how-to-run end -->
298
+
299
+ <!-- footer start -->
300
+ <!-- 200823 -->
301
+ ## Discord
302
+
303
+ For further support, and discussions on these models and AI in general, join us at:
304
+
305
+ [TheBloke AI's Discord server](https://discord.gg/theblokeai)
306
+
307
+ ## Thanks, and how to contribute
308
+
309
+ Thanks to the [chirper.ai](https://chirper.ai) team!
310
+
311
+ Thanks to Clay from [gpus.llm-utils.org](llm-utils)!
312
+
313
+ I've had a lot of people ask if they can contribute. I enjoy providing models and helping people, and would love to be able to spend even more time doing it, as well as expanding into new projects like fine tuning/training.
314
+
315
+ If you're able and willing to contribute it will be most gratefully received and will help me to keep providing more models, and to start work on new AI projects.
316
+
317
+ Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
318
+
319
+ * Patreon: https://patreon.com/TheBlokeAI
320
+ * Ko-Fi: https://ko-fi.com/TheBlokeAI
321
+
322
+ **Special thanks to**: Aemon Algiz.
323
+
324
+ **Patreon special mentions**: Brandon Frisco, LangChain4j, Spiking Neurons AB, transmissions 11, Joseph William Delisle, Nitin Borwankar, Willem Michiel, Michael Dempsey, vamX, Jeffrey Morgan, zynix, jjj, Omer Bin Jawed, Sean Connelly, jinyuan sun, Jeromy Smith, Shadi, Pawan Osman, Chadd, Elijah Stavena, Illia Dulskyi, Sebastain Graf, Stephen Murray, terasurfer, Edmond Seymore, Celu Ramasamy, Mandus, Alex, biorpg, Ajan Kanaga, Clay Pascal, Raven Klaugh, 阿明, K, ya boyyy, usrbinkat, Alicia Loh, John Villwock, ReadyPlayerEmma, Chris Smitley, Cap'n Zoog, fincy, GodLy, S_X, sidney chen, Cory Kujawski, OG, Mano Prime, AzureBlack, Pieter, Kalila, Spencer Kim, Tom X Nguyen, Stanislav Ovsiannikov, Michael Levine, Andrey, Trailburnt, Vadim, Enrico Ros, Talal Aujan, Brandon Phillips, Jack West, Eugene Pentland, Michael Davis, Will Dee, webtim, Jonathan Leane, Alps Aficionado, Rooh Singh, Tiffany J. Kim, theTransient, Luke @flexchar, Elle, Caitlyn Gatomon, Ari Malik, subjectnull, Johann-Peter Hartmann, Trenton Dambrowitz, Imad Khwaja, Asp the Wyvern, Emad Mostaque, Rainer Wilmers, Alexandros Triantafyllidis, Nicholas, Pedro Madruga, SuperWojo, Harry Royden McLaughlin, James Bentley, Olakabola, David Ziegler, Ai Maven, Jeff Scroggin, Nikolai Manek, Deo Leter, Matthew Berman, Fen Risland, Ken Nordquist, Manuel Alberto Morcote, Luke Pendergrass, TL, Fred von Graf, Randy H, Dan Guido, NimbleBox.ai, Vitor Caleffi, Gabriel Tamborski, knownsqashed, Lone Striker, Erik Bjäreholt, John Detwiler, Leonard Tan, Iucharbius
325
+
326
+
327
+ Thank you to all my generous patrons and donaters!
328
+
329
+ And thank you again to a16z for their generous grant.
330
+
331
+ <!-- footer end -->
332
+
333
+ <!-- original-model-card start -->
334
+ # Original model card: NumbersStation's NSQL Llama-2 7B
335
+
336
+
337
+ # NSQL-Llama-2-7B
338
+
339
+ ## Model Description
340
+
341
+ NSQL is a family of autoregressive open-source large foundation models (FMs) designed specifically for SQL generation tasks.
342
+
343
+ In this repository we are introducing a new member of NSQL, NSQL-Llama-2-7B. It's based on Meta's original [Llama-2 7B model](https://huggingface.co/meta-llama/Llama-2-7b) and further pre-trained on a dataset of general SQL queries and then fine-tuned on a dataset composed of text-to-SQL pairs.
344
+
345
+ ## Training Data
346
+
347
+ The general SQL queries are the SQL subset from [The Stack](https://huggingface.co/datasets/bigcode/the-stack), containing 1M training samples. The labeled text-to-SQL pairs come from more than 20 public sources across the web from standard datasets. We hold out Spider and GeoQuery datasets for use in evaluation.
348
+
349
+ ## Evaluation Data
350
+
351
+ We evaluate our models on two text-to-SQL benchmarks: Spider and GeoQuery.
352
+
353
+ ## Training Procedure
354
+
355
+ NSQL was trained using cross-entropy loss to maximize the likelihood of sequential inputs. For finetuning on text-to-SQL pairs, we only compute the loss over the SQL portion of the pair. The model is trained using 80GB A100s, leveraging data and model parallelism. We pre-trained for 3 epochs and fine-tuned for 10 epochs.
356
+
357
+ ## Intended Use and Limitations
358
+
359
+ The model was designed for text-to-SQL generation tasks from given table schema and natural language prompts. The model works best with the prompt format defined below and outputting `SELECT` queries.
360
+
361
+ ## How to Use
362
+
363
+ Example 1:
364
+
365
+ ```python
366
+ import torch
367
+ from transformers import AutoTokenizer, AutoModelForCausalLM
368
+ tokenizer = AutoTokenizer.from_pretrained("NumbersStation/nsql-llama-2-7B")
369
+ model = AutoModelForCausalLM.from_pretrained("NumbersStation/nsql-llama-2-7B", torch_dtype=torch.bfloat16)
370
+
371
+ text = """CREATE TABLE stadium (
372
+ stadium_id number,
373
+ location text,
374
+ name text,
375
+ capacity number,
376
+ highest number,
377
+ lowest number,
378
+ average number
379
+ )
380
+
381
+ CREATE TABLE singer (
382
+ singer_id number,
383
+ name text,
384
+ country text,
385
+ song_name text,
386
+ song_release_year text,
387
+ age number,
388
+ is_male others
389
+ )
390
+
391
+ CREATE TABLE concert (
392
+ concert_id number,
393
+ concert_name text,
394
+ theme text,
395
+ stadium_id text,
396
+ year text
397
+ )
398
+
399
+ CREATE TABLE singer_in_concert (
400
+ concert_id number,
401
+ singer_id text
402
+ )
403
+
404
+ -- Using valid SQLite, answer the following questions for the tables provided above.
405
+
406
+ -- What is the maximum, the average, and the minimum capacity of stadiums ?
407
+
408
+ SELECT"""
409
+
410
+ input_ids = tokenizer(text, return_tensors="pt").input_ids
411
+
412
+ generated_ids = model.generate(input_ids, max_length=500)
413
+ print(tokenizer.decode(generated_ids[0], skip_special_tokens=True))
414
+ ```
415
+
416
+ Example 2:
417
+
418
+ ```python
419
+ import torch
420
+ from transformers import AutoTokenizer, AutoModelForCausalLM
421
+ tokenizer = AutoTokenizer.from_pretrained("NumbersStation/nsql-llama-2-7B")
422
+ model = AutoModelForCausalLM.from_pretrained("NumbersStation/nsql-llama-2-7B", torch_dtype=torch.bfloat16)
423
+
424
+ text = """CREATE TABLE stadium (
425
+ stadium_id number,
426
+ location text,
427
+ name text,
428
+ capacity number,
429
+ )
430
+
431
+ -- Using valid SQLite, answer the following questions for the tables provided above.
432
+
433
+ -- how many stadiums in total?
434
+
435
+ SELECT"""
436
+
437
+ input_ids = tokenizer(text, return_tensors="pt").input_ids
438
+
439
+ generated_ids = model.generate(input_ids, max_length=500)
440
+ print(tokenizer.decode(generated_ids[0], skip_special_tokens=True))
441
+ ```
442
+
443
+ Example 3:
444
+
445
+ ```python
446
+ import torch
447
+ from transformers import AutoTokenizer, AutoModelForCausalLM
448
+ tokenizer = AutoTokenizer.from_pretrained("NumbersStation/nsql-llama-2-7B")
449
+ model = AutoModelForCausalLM.from_pretrained("NumbersStation/nsql-llama-2-7B", torch_dtype=torch.bfloat16)
450
+
451
+ text = """CREATE TABLE work_orders (
452
+ ID NUMBER,
453
+ CREATED_AT TEXT,
454
+ COST FLOAT,
455
+ INVOICE_AMOUNT FLOAT,
456
+ IS_DUE BOOLEAN,
457
+ IS_OPEN BOOLEAN,
458
+ IS_OVERDUE BOOLEAN,
459
+ COUNTRY_NAME TEXT,
460
+ )
461
+
462
+ -- Using valid SQLite, answer the following questions for the tables provided above.
463
+
464
+ -- how many work orders are open?
465
+
466
+ SELECT"""
467
+
468
+ input_ids = tokenizer(text, return_tensors="pt").input_ids
469
+
470
+ generated_ids = model.generate(input_ids, max_length=500)
471
+ print(tokenizer.decode(generated_ids[0], skip_special_tokens=True))
472
+ ```
473
+
474
+
475
+
476
+ For more information (e.g., run with your local database), please find examples in [this repository](https://github.com/NumbersStationAI/NSQL).
477
+
478
+ <!-- original-model-card end -->