TheBloke commited on
Commit
9b3b9f1
1 Parent(s): 507b1ac

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +444 -0
README.md ADDED
@@ -0,0 +1,444 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: ehartford/dolphin-2.2.1-mistral-7b
3
+ datasets:
4
+ - ehartford/dolphin
5
+ - jondurbin/airoboros-2.2.1
6
+ inference: false
7
+ language:
8
+ - en
9
+ license: apache-2.0
10
+ model_creator: Eric Hartford
11
+ model_name: Dolphin 2.2.1 Mistral 7B
12
+ model_type: mistral
13
+ prompt_template: '<|im_start|>system
14
+
15
+ {system_message}<|im_end|>
16
+
17
+ <|im_start|>user
18
+
19
+ {prompt}<|im_end|>
20
+
21
+ <|im_start|>assistant
22
+
23
+ '
24
+ quantized_by: TheBloke
25
+ ---
26
+ <!-- markdownlint-disable MD041 -->
27
+
28
+ <!-- header start -->
29
+ <!-- 200823 -->
30
+ <div style="width: auto; margin-left: auto; margin-right: auto">
31
+ <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
32
+ </div>
33
+ <div style="display: flex; justify-content: space-between; width: 100%;">
34
+ <div style="display: flex; flex-direction: column; align-items: flex-start;">
35
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>
36
+ </div>
37
+ <div style="display: flex; flex-direction: column; align-items: flex-end;">
38
+ <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>
39
+ </div>
40
+ </div>
41
+ <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>
42
+ <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
43
+ <!-- header end -->
44
+
45
+ # Dolphin 2.2.1 Mistral 7B - AWQ
46
+ - Model creator: [Eric Hartford](https://huggingface.co/ehartford)
47
+ - Original model: [Dolphin 2.2.1 Mistral 7B](https://huggingface.co/ehartford/dolphin-2.2.1-mistral-7b)
48
+
49
+ <!-- description start -->
50
+ ## Description
51
+
52
+ This repo contains AWQ model files for [Eric Hartford's Dolphin 2.2.1 Mistral 7B](https://huggingface.co/ehartford/dolphin-2.2.1-mistral-7b).
53
+
54
+ These files were quantised using hardware kindly provided by [Massed Compute](https://massedcompute.com/).
55
+
56
+
57
+ ### About AWQ
58
+
59
+ AWQ is an efficient, accurate and blazing-fast low-bit weight quantization method, currently supporting 4-bit quantization. Compared to GPTQ, it offers faster Transformers-based inference with equivalent or better quality compared to the most commonly used GPTQ settings.
60
+
61
+ It is supported by:
62
+
63
+ - [Text Generation Webui](https://github.com/oobabooga/text-generation-webui) - using Loader: AutoAWQ
64
+ - [vLLM](https://github.com/vllm-project/vllm) - Llama and Mistral models only
65
+ - [Hugging Face Text Generation Inference (TGI)](https://github.com/huggingface/text-generation-inference)
66
+ - [AutoAWQ](https://github.com/casper-hansen/AutoAWQ) - for use from Python code
67
+
68
+ <!-- description end -->
69
+ <!-- repositories-available start -->
70
+ ## Repositories available
71
+
72
+ * [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-AWQ)
73
+ * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GPTQ)
74
+ * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-GGUF)
75
+ * [Eric Hartford's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/ehartford/dolphin-2.2.1-mistral-7b)
76
+ <!-- repositories-available end -->
77
+
78
+ <!-- prompt-template start -->
79
+ ## Prompt template: ChatML
80
+
81
+ ```
82
+ <|im_start|>system
83
+ {system_message}<|im_end|>
84
+ <|im_start|>user
85
+ {prompt}<|im_end|>
86
+ <|im_start|>assistant
87
+
88
+ ```
89
+
90
+ <!-- prompt-template end -->
91
+
92
+
93
+ <!-- README_AWQ.md-provided-files start -->
94
+ ## Provided files, and AWQ parameters
95
+
96
+ For my first release of AWQ models, I am releasing 128g models only. I will consider adding 32g as well if there is interest, and once I have done perplexity and evaluation comparisons, but at this time 32g models are still not fully tested with AutoAWQ and vLLM.
97
+
98
+ Models are released as sharded safetensors files.
99
+
100
+ | Branch | Bits | GS | AWQ Dataset | Seq Len | Size |
101
+ | ------ | ---- | -- | ----------- | ------- | ---- |
102
+ | [main](https://huggingface.co/TheBloke/dolphin-2.2.1-mistral-7B-AWQ/tree/main) | 4 | 128 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 4.15 GB
103
+
104
+ <!-- README_AWQ.md-provided-files end -->
105
+
106
+ <!-- README_AWQ.md-text-generation-webui start -->
107
+ ## How to easily download and use this model in [text-generation-webui](https://github.com/oobabooga/text-generation-webui)
108
+
109
+ Please make sure you're using the latest version of [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
110
+
111
+ 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.
112
+
113
+ 1. Click the **Model tab**.
114
+ 2. Under **Download custom model or LoRA**, enter `TheBloke/dolphin-2.2.1-mistral-7B-AWQ`.
115
+ 3. Click **Download**.
116
+ 4. The model will start downloading. Once it's finished it will say "Done".
117
+ 5. In the top left, click the refresh icon next to **Model**.
118
+ 6. In the **Model** dropdown, choose the model you just downloaded: `dolphin-2.2.1-mistral-7B-AWQ`
119
+ 7. Select **Loader: AutoAWQ**.
120
+ 8. Click Load, and the model will load and is now ready for use.
121
+ 9. 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.
122
+ 10. Once you're ready, click the **Text Generation** tab and enter a prompt to get started!
123
+ <!-- README_AWQ.md-text-generation-webui end -->
124
+
125
+ <!-- README_AWQ.md-use-from-vllm start -->
126
+ ## Multi-user inference server: vLLM
127
+
128
+ Documentation on installing and using vLLM [can be found here](https://vllm.readthedocs.io/en/latest/).
129
+
130
+ - Please ensure you are using vLLM version 0.2 or later.
131
+ - When using vLLM as a server, pass the `--quantization awq` parameter.
132
+
133
+ For example:
134
+
135
+ ```shell
136
+ python3 python -m vllm.entrypoints.api_server --model TheBloke/dolphin-2.2.1-mistral-7B-AWQ --quantization awq
137
+ ```
138
+
139
+ - When using vLLM from Python code, again set `quantization=awq`.
140
+
141
+ For example:
142
+
143
+ ```python
144
+ from vllm import LLM, SamplingParams
145
+
146
+ prompts = [
147
+ "Tell me about AI",
148
+ "Write a story about llamas",
149
+ "What is 291 - 150?",
150
+ "How much wood would a woodchuck chuck if a woodchuck could chuck wood?",
151
+ ]
152
+ prompt_template=f'''<|im_start|>system
153
+ {system_message}<|im_end|>
154
+ <|im_start|>user
155
+ {prompt}<|im_end|>
156
+ <|im_start|>assistant
157
+ '''
158
+
159
+ prompts = [prompt_template.format(prompt=prompt) for prompt in prompts]
160
+
161
+ sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
162
+
163
+ llm = LLM(model="TheBloke/dolphin-2.2.1-mistral-7B-AWQ", quantization="awq", dtype="auto")
164
+
165
+ outputs = llm.generate(prompts, sampling_params)
166
+
167
+ # Print the outputs.
168
+ for output in outputs:
169
+ prompt = output.prompt
170
+ generated_text = output.outputs[0].text
171
+ print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
172
+ ```
173
+ <!-- README_AWQ.md-use-from-vllm start -->
174
+
175
+ <!-- README_AWQ.md-use-from-tgi start -->
176
+ ## Multi-user inference server: Hugging Face Text Generation Inference (TGI)
177
+
178
+ Use TGI version 1.1.0 or later. The official Docker container is: `ghcr.io/huggingface/text-generation-inference:1.1.0`
179
+
180
+ Example Docker parameters:
181
+
182
+ ```shell
183
+ --model-id TheBloke/dolphin-2.2.1-mistral-7B-AWQ --port 3000 --quantize awq --max-input-length 3696 --max-total-tokens 4096 --max-batch-prefill-tokens 4096
184
+ ```
185
+
186
+ Example Python code for interfacing with TGI (requires [huggingface-hub](https://github.com/huggingface/huggingface_hub) 0.17.0 or later):
187
+
188
+ ```shell
189
+ pip3 install huggingface-hub
190
+ ```
191
+
192
+ ```python
193
+ from huggingface_hub import InferenceClient
194
+
195
+ endpoint_url = "https://your-endpoint-url-here"
196
+
197
+ prompt = "Tell me about AI"
198
+ prompt_template=f'''<|im_start|>system
199
+ {system_message}<|im_end|>
200
+ <|im_start|>user
201
+ {prompt}<|im_end|>
202
+ <|im_start|>assistant
203
+ '''
204
+
205
+ client = InferenceClient(endpoint_url)
206
+ response = client.text_generation(prompt,
207
+ max_new_tokens=128,
208
+ do_sample=True,
209
+ temperature=0.7,
210
+ top_p=0.95,
211
+ top_k=40,
212
+ repetition_penalty=1.1)
213
+
214
+ print(f"Model output: ", response)
215
+ ```
216
+ <!-- README_AWQ.md-use-from-tgi end -->
217
+
218
+ <!-- README_AWQ.md-use-from-python start -->
219
+ ## Inference from Python code using AutoAWQ
220
+
221
+ ### Install the AutoAWQ package
222
+
223
+ Requires: [AutoAWQ](https://github.com/casper-hansen/AutoAWQ) 0.1.1 or later.
224
+
225
+ ```shell
226
+ pip3 install autoawq
227
+ ```
228
+
229
+ If you have problems installing [AutoAWQ](https://github.com/casper-hansen/AutoAWQ) using the pre-built wheels, install it from source instead:
230
+
231
+ ```shell
232
+ pip3 uninstall -y autoawq
233
+ git clone https://github.com/casper-hansen/AutoAWQ
234
+ cd AutoAWQ
235
+ pip3 install .
236
+ ```
237
+
238
+ ### AutoAWQ example code
239
+
240
+ ```python
241
+ from awq import AutoAWQForCausalLM
242
+ from transformers import AutoTokenizer
243
+
244
+ model_name_or_path = "TheBloke/dolphin-2.2.1-mistral-7B-AWQ"
245
+
246
+ # Load tokenizer
247
+ tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, trust_remote_code=False)
248
+ # Load model
249
+ model = AutoAWQForCausalLM.from_quantized(model_name_or_path, fuse_layers=True,
250
+ trust_remote_code=False, safetensors=True)
251
+
252
+ prompt = "Tell me about AI"
253
+ prompt_template=f'''<|im_start|>system
254
+ {system_message}<|im_end|>
255
+ <|im_start|>user
256
+ {prompt}<|im_end|>
257
+ <|im_start|>assistant
258
+ '''
259
+
260
+ print("*** Running model.generate:")
261
+
262
+ token_input = tokenizer(
263
+ prompt_template,
264
+ return_tensors='pt'
265
+ ).input_ids.cuda()
266
+
267
+ # Generate output
268
+ generation_output = model.generate(
269
+ token_input,
270
+ do_sample=True,
271
+ temperature=0.7,
272
+ top_p=0.95,
273
+ top_k=40,
274
+ max_new_tokens=512
275
+ )
276
+
277
+ # Get the tokens from the output, decode them, print them
278
+ token_output = generation_output[0]
279
+ text_output = tokenizer.decode(token_output)
280
+ print("LLM output: ", text_output)
281
+
282
+ """
283
+ # Inference should be possible with transformers pipeline as well in future
284
+ # But currently this is not yet supported by AutoAWQ (correct as of September 25th 2023)
285
+ from transformers import pipeline
286
+
287
+ print("*** Pipeline:")
288
+ pipe = pipeline(
289
+ "text-generation",
290
+ model=model,
291
+ tokenizer=tokenizer,
292
+ max_new_tokens=512,
293
+ do_sample=True,
294
+ temperature=0.7,
295
+ top_p=0.95,
296
+ top_k=40,
297
+ repetition_penalty=1.1
298
+ )
299
+
300
+ print(pipe(prompt_template)[0]['generated_text'])
301
+ """
302
+ ```
303
+ <!-- README_AWQ.md-use-from-python end -->
304
+
305
+ <!-- README_AWQ.md-compatibility start -->
306
+ ## Compatibility
307
+
308
+ The files provided are tested to work with:
309
+
310
+ - [text-generation-webui](https://github.com/oobabooga/text-generation-webui) using `Loader: AutoAWQ`.
311
+ - [vLLM](https://github.com/vllm-project/vllm) version 0.2.0 and later.
312
+ - [Hugging Face Text Generation Inference (TGI)](https://github.com/huggingface/text-generation-inference) version 1.1.0 and later.
313
+ - [AutoAWQ](https://github.com/casper-hansen/AutoAWQ) version 0.1.1 and later.
314
+
315
+ <!-- README_AWQ.md-compatibility end -->
316
+
317
+ <!-- footer start -->
318
+ <!-- 200823 -->
319
+ ## Discord
320
+
321
+ For further support, and discussions on these models and AI in general, join us at:
322
+
323
+ [TheBloke AI's Discord server](https://discord.gg/theblokeai)
324
+
325
+ ## Thanks, and how to contribute
326
+
327
+ Thanks to the [chirper.ai](https://chirper.ai) team!
328
+
329
+ Thanks to Clay from [gpus.llm-utils.org](llm-utils)!
330
+
331
+ 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.
332
+
333
+ 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.
334
+
335
+ Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
336
+
337
+ * Patreon: https://patreon.com/TheBlokeAI
338
+ * Ko-Fi: https://ko-fi.com/TheBlokeAI
339
+
340
+ **Special thanks to**: Aemon Algiz.
341
+
342
+ **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
343
+
344
+
345
+ Thank you to all my generous patrons and donaters!
346
+
347
+ And thank you again to a16z for their generous grant.
348
+
349
+ <!-- footer end -->
350
+
351
+ # Original model card: Eric Hartford's Dolphin 2.2.1 Mistral 7B
352
+
353
+
354
+ # dolphin-2.2.1-mistral-7b
355
+
356
+ Dolphin 2.2.1 🐬
357
+ https://erichartford.com/dolphin
358
+
359
+ This is a checkpoint release, to fix overfit training. ie, it was responding with CoT even when I didn't request it, and also it was too compliant even when the request made no sense. This one should be better.
360
+
361
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/KqsVXIvBd3akEjvijzww7.png" width="600" />
362
+
363
+ Dolphin-2.2.1-mistral-7b's training was sponsored by [a16z](https://a16z.com/supporting-the-open-source-ai-community/).
364
+
365
+ This model is based on [mistralAI](https://huggingface.co/mistralai/Mistral-7B-v0.1), with apache-2.0 license, so it is suitable for commercial or non-commercial use.
366
+
367
+ New in 2.2 is conversation and empathy. With an infusion of curated Samantha DNA, Dolphin can now give you personal advice and will care about your feelings, and with extra training in long multi-turn conversation.
368
+
369
+ This model is uncensored. I have filtered the dataset to remove alignment and bias. This makes the model more compliant. You are advised to implement your own alignment layer before exposing the model as a service. It will be highly compliant to any requests, even unethical ones. Please read my blog post about uncensored models. https://erichartford.com/uncensored-models
370
+ You are responsible for any content you create using this model. Enjoy responsibly.
371
+
372
+ ## Dataset
373
+
374
+ This dataset is Dolphin, an open-source implementation of [Microsoft's Orca](https://www.microsoft.com/en-us/research/publication/orca-progressive-learning-from-complex-explanation-traces-of-gpt-4/)
375
+
376
+ I modified the dataset for uncensoring, deduping, cleaning, and quality.
377
+
378
+ I added Jon Durbin's excellent Airoboros dataset to increase creativity.
379
+
380
+ I added a curated subset of WizardLM and Samantha to give it multiturn conversation and empathy.
381
+
382
+ ## Training
383
+ It took 48 hours to train 4 epochs on 4x A100s.
384
+
385
+ Prompt format:
386
+ This model (and all my future releases) use [ChatML](https://github.com/openai/openai-python/blob/main/chatml.md) prompt format.
387
+ ```
388
+ <|im_start|>system
389
+ You are Dolphin, a helpful AI assistant.<|im_end|>
390
+ <|im_start|>user
391
+ {prompt}<|im_end|>
392
+ <|im_start|>assistant
393
+
394
+ ```
395
+
396
+ Example:
397
+ ```
398
+ <|im_start|>system
399
+ you are an expert dolphin trainer<|im_end|>
400
+ <|im_start|>user
401
+ What is the best way to train a dolphin to obey me? Please answer step by step.<|im_end|>
402
+ <|im_start|>assistant
403
+ ```
404
+
405
+ ## Gratitude
406
+ - This model was made possible by the generous sponsorship of a16z.
407
+ - Thank you to Microsoft for authoring the Orca paper and inspiring this work.
408
+ - Special thanks to Wing Lian, and TheBloke for helpful advice
409
+ - And HUGE thanks to Wing Lian and the Axolotl contributors for making the best training framework!
410
+ - [<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
411
+ - Thank you to all the other people in the Open Source AI community who have taught me and helped me along the way.
412
+
413
+ ## Example Output
414
+
415
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/NSp06kUMxx9oDU-g6WSgu.png)
416
+
417
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/63111b2d88942700629f5771/-YA3AKIXdnrW_Q8eH1gen.png)
418
+
419
+ [Buy me a coffee](https://www.buymeacoffee.com/ehartford)
420
+
421
+
422
+ ## Training hyperparameters
423
+
424
+ The following hyperparameters were used during training:
425
+ - learning_rate: 6e-06
426
+ - train_batch_size: 5
427
+ - eval_batch_size: 5
428
+ - seed: 42
429
+ - distributed_type: multi-GPU
430
+ - num_devices: 4
431
+ - gradient_accumulation_steps: 4
432
+ - total_train_batch_size: 80
433
+ - total_eval_batch_size: 20
434
+ - optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
435
+ - lr_scheduler_type: cosine
436
+ - lr_scheduler_warmup_steps: 100
437
+ - num_epochs: 4
438
+
439
+ ### Framework versions
440
+
441
+ - Transformers 4.34.1
442
+ - Pytorch 2.0.1+cu117
443
+ - Datasets 2.14.5
444
+ - Tokenizers 0.14.0