Hugging Face's logo

PrunaAI
/
distributed-optimized-gpt2-2b-HQQ-4bit-smashed

gpt_optimized
pruna-ai
custom_code
Model card Files Community
15
sharpenb commited on
Commit
8f7e8a1
·
verified ·
1 Parent(s): 0d71107

Upload folder using huggingface_hub (#1)

Browse files

- d408d4a901ea32b540ac407e50a25c828e2abf673635cfd5578b09d2727cc78f (01dbb13d47bce9b10c247685227d51b876814f99)

Files changed (5) hide show
  1. README.md +89 -0
  2. config.json +373 -0
  3. configuration_gpt_optimized.py +22 -0
  4. qmodel.pt +3 -0
  5. smash_config.json +34 -0
README.md ADDED
@@ -0,0 +1,89 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ thumbnail: "https://assets-global.website-files.com/646b351987a8d8ce158d1940/64ec9e96b4334c0e1ac41504_Logo%20with%20white%20text.svg"
3
+ base_model: ORIGINAL_REPO_NAME
4
+ metrics:
5
+ - memory_disk
6
+ - memory_inference
7
+ - inference_latency
8
+ - inference_throughput
9
+ - inference_CO2_emissions
10
+ - inference_energy_consumption
11
+ tags:
12
+ - pruna-ai
13
+ ---
14
+ <!-- header start -->
15
+ <!-- 200823 -->
16
+ <div style="width: auto; margin-left: auto; margin-right: auto">
17
+ <a href="https://www.pruna.ai/" target="_blank" rel="noopener noreferrer">
18
+ <img src="https://i.imgur.com/eDAlcgk.png" alt="PrunaAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
19
+ </a>
20
+ </div>
21
+ <!-- header end -->
22
+
23
+ [![Twitter](https://img.shields.io/twitter/follow/PrunaAI?style=social)](https://twitter.com/PrunaAI)
24
+ [![GitHub](https://img.shields.io/github/followers/PrunaAI?label=Follow%20%40PrunaAI&style=social)](https://github.com/PrunaAI)
25
+ [![LinkedIn](https://img.shields.io/badge/LinkedIn-Connect-blue)](https://www.linkedin.com/company/93832878/admin/feed/posts/?feedType=following)
26
+ [![Discord](https://img.shields.io/badge/Discord-Join%20Us-blue?style=social&logo=discord)](https://discord.gg/rskEr4BZJx)
27
+
28
+ # Simply make AI models cheaper, smaller, faster, and greener!
29
+
30
+ - Give a thumbs up if you like this model!
31
+ - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
32
+ - Request access to easily compress your *own* AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
33
+ - Read the documentations to know more [here](https://pruna-ai-pruna.readthedocs-hosted.com/en/latest/)
34
+ - Join Pruna AI community on Discord [here](https://discord.gg/CP4VSgck) to share feedback/suggestions or get help.
35
+
36
+ ## Results
37
+
38
+ ![image info](./plots.png)
39
+
40
+ **Frequently Asked Questions**
41
+ - ***How does the compression work?*** The model is compressed with hqq.
42
+ - ***How does the model quality change?*** The quality of the model output might vary compared to the base model.
43
+ - ***How is the model efficiency evaluated?*** These results were obtained with configuration described in `model/smash_config.json` and are obtained after a hardware warmup. The smashed model is directly compared to the original base model. Efficiency results may vary in other settings (e.g. other hardware, image size, batch size, ...). We recommend to directly run them in the use-case conditions to know if the smashed model can benefit you.
44
+ - ***What is the model format?*** We use safetensors.
45
+ - ***What calibration data has been used?*** If needed by the compression method, we used WikiText as the calibration data.
46
+ - ***What is the naming convention for Pruna Huggingface models?*** We take the original model name and append "turbo", "tiny", or "green" if the smashed model has a measured inference speed, inference memory, or inference energy consumption which is less than 90% of the original base model.
47
+ - ***How to compress my own models?*** You can request premium access to more compression methods and tech support for your specific use-cases [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
48
+ - ***What are "first" metrics?*** Results mentioning "first" are obtained after the first run of the model. The first run might take more memory or be slower than the subsequent runs due cuda overheads.
49
+ - ***What are "Sync" and "Async" metrics?*** "Sync" metrics are obtained by syncing all GPU processes and stop measurement when all of them are executed. "Async" metrics are obtained without syncing all GPU processes and stop when the model output can be used by the CPU. We provide both metrics since both could be relevant depending on the use-case. We recommend to test the efficiency gains directly in your use-cases.
50
+
51
+ ## Setup
52
+
53
+ You can run the smashed model with these steps:
54
+
55
+ 0. Check requirements from the original repo ORIGINAL_REPO_NAME installed. In particular, check python, cuda, and transformers versions.
56
+ 1. Make sure that you have installed quantization related packages.
57
+ ```bash
58
+ pip install hqq
59
+ ```
60
+ 2. Load & run the model.
61
+ ```python
62
+ from transformers import AutoModelForCausalLM, AutoTokenizer
63
+ from hqq.engine.hf import HQQModelForCausalLM
64
+ from hqq.models.hf.base import AutoHQQHFModel
65
+
66
+ try:
67
+ model = HQQModelForCausalLM.from_quantized("PrunaAI/distributed-optimized-gpt2-2b-HQQ-4bit-smashed", device_map='auto')
68
+ except:
69
+ model = AutoHQQHFModel.from_quantized("PrunaAI/distributed-optimized-gpt2-2b-HQQ-4bit-smashed")
70
+ tokenizer = AutoTokenizer.from_pretrained("ORIGINAL_REPO_NAME")
71
+
72
+ input_ids = tokenizer("What is the color of prunes?,", return_tensors='pt').to(model.device)["input_ids"]
73
+
74
+ outputs = model.generate(input_ids, max_new_tokens=216)
75
+ tokenizer.decode(outputs[0])
76
+ ```
77
+
78
+ ## Configurations
79
+
80
+ The configuration info are in `smash_config.json`.
81
+
82
+ ## Credits & License
83
+
84
+ The license of the smashed model follows the license of the original model. Please check the license of the original model ORIGINAL_REPO_NAME before using this model which provided the base model. The license of the `pruna-engine` is [here](https://pypi.org/project/pruna-engine/) on Pypi.
85
+
86
+ ## Want to compress other models?
87
+
88
+ - Contact us and tell us which model to compress next [here](https://www.pruna.ai/contact).
89
+ - Request access to easily compress your own AI models [here](https://z0halsaff74.typeform.com/pruna-access?typeform-source=www.pruna.ai).
config.json ADDED
@@ -0,0 +1,373 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/tmp/models/tmp1cq_dazl/tmpz1oy8_gx",
3
+ "activation_function": "gelu_new",
4
+ "all_reduce_scores": {
5
+ "0": "NON_PARTICIPATING",
6
+ "1": "SUCCESS",
7
+ "10": "SUCCESS",
8
+ "100": "NON_PARTICIPATING",
9
+ "101": "NON_PARTICIPATING",
10
+ "102": "NON_PARTICIPATING",
11
+ "103": "NON_PARTICIPATING",
12
+ "104": "NON_PARTICIPATING",
13
+ "105": "NON_PARTICIPATING",
14
+ "106": "SUCCESS",
15
+ "107": "NON_PARTICIPATING",
16
+ "108": "NON_PARTICIPATING",
17
+ "109": "NON_PARTICIPATING",
18
+ "11": "SUCCESS",
19
+ "110": "NON_PARTICIPATING",
20
+ "111": "NON_PARTICIPATING",
21
+ "112": "NON_PARTICIPATING",
22
+ "113": "SUCCESS",
23
+ "114": "NON_PARTICIPATING",
24
+ "115": "NON_PARTICIPATING",
25
+ "116": "SUCCESS",
26
+ "117": "NON_PARTICIPATING",
27
+ "118": "NON_PARTICIPATING",
28
+ "119": "SUCCESS",
29
+ "12": "SUCCESS",
30
+ "120": "NON_PARTICIPATING",
31
+ "121": "NON_PARTICIPATING",
32
+ "122": "NON_PARTICIPATING",
33
+ "123": "SUCCESS",
34
+ "124": "NON_PARTICIPATING",
35
+ "125": "SUCCESS",
36
+ "126": "NON_PARTICIPATING",
37
+ "127": "SUCCESS",
38
+ "128": "NON_PARTICIPATING",
39
+ "129": "NON_PARTICIPATING",
40
+ "13": "SUCCESS",
41
+ "130": "NON_PARTICIPATING",
42
+ "131": "NON_PARTICIPATING",
43
+ "132": "SUCCESS",
44
+ "133": "NON_PARTICIPATING",
45
+ "134": "NON_PARTICIPATING",
46
+ "135": "NON_PARTICIPATING",
47
+ "136": "NON_PARTICIPATING",
48
+ "137": "NON_PARTICIPATING",
49
+ "138": "SUCCESS",
50
+ "139": "SUCCESS",
51
+ "14": "NON_PARTICIPATING",
52
+ "140": "SUCCESS",
53
+ "141": "NON_PARTICIPATING",
54
+ "142": "NON_PARTICIPATING",
55
+ "143": "NON_PARTICIPATING",
56
+ "144": "NON_PARTICIPATING",
57
+ "145": "NON_PARTICIPATING",
58
+ "146": "NON_PARTICIPATING",
59
+ "147": "SUCCESS",
60
+ "148": "SUCCESS",
61
+ "149": "NON_PARTICIPATING",
62
+ "15": "SUCCESS",
63
+ "150": "NON_PARTICIPATING",
64
+ "151": "SUCCESS",
65
+ "152": "SUCCESS",
66
+ "153": "NON_PARTICIPATING",
67
+ "154": "NON_PARTICIPATING",
68
+ "155": "NON_PARTICIPATING",
69
+ "156": "NON_PARTICIPATING",
70
+ "157": "SUCCESS",
71
+ "158": "NON_PARTICIPATING",
72
+ "159": "SUCCESS",
73
+ "16": "NON_PARTICIPATING",
74
+ "160": "SUCCESS",
75
+ "161": "NON_PARTICIPATING",
76
+ "162": "NON_PARTICIPATING",
77
+ "163": "NON_PARTICIPATING",
78
+ "164": "NON_PARTICIPATING",
79
+ "165": "NON_PARTICIPATING",
80
+ "166": "SUCCESS",
81
+ "167": "SUCCESS",
82
+ "168": "NON_PARTICIPATING",
83
+ "169": "NON_PARTICIPATING",
84
+ "17": "NON_PARTICIPATING",
85
+ "170": "NON_PARTICIPATING",
86
+ "171": "NON_PARTICIPATING",
87
+ "172": "NON_PARTICIPATING",
88
+ "173": "SUCCESS",
89
+ "174": "SUCCESS",
90
+ "175": "SUCCESS",
91
+ "176": "SUCCESS",
92
+ "177": "NON_PARTICIPATING",
93
+ "178": "SUCCESS",
94
+ "179": "SUCCESS",
95
+ "18": "SUCCESS",
96
+ "180": "SUCCESS",
97
+ "181": "NON_PARTICIPATING",
98
+ "182": "NON_PARTICIPATING",
99
+ "183": "NON_PARTICIPATING",
100
+ "184": "SUCCESS",
101
+ "185": "NON_PARTICIPATING",
102
+ "186": "NON_PARTICIPATING",
103
+ "187": "SUCCESS",
104
+ "188": "SUCCESS",
105
+ "189": "SUCCESS",
106
+ "19": "NON_PARTICIPATING",
107
+ "190": "NON_PARTICIPATING",
108
+ "191": "SUCCESS",
109
+ "192": "SUCCESS",
110
+ "193": "SUCCESS",
111
+ "194": "SUCCESS",
112
+ "195": "SUCCESS",
113
+ "196": "SUCCESS",
114
+ "197": "NON_PARTICIPATING",
115
+ "198": "NON_PARTICIPATING",
116
+ "199": "SUCCESS",
117
+ "2": "NON_PARTICIPATING",
118
+ "20": "SUCCESS",
119
+ "200": "NON_PARTICIPATING",
120
+ "201": "NON_PARTICIPATING",
121
+ "202": "NON_PARTICIPATING",
122
+ "203": "NON_PARTICIPATING",
123
+ "204": "SUCCESS",
124
+ "205": "SUCCESS",
125
+ "206": "NON_PARTICIPATING",
126
+ "207": "NON_PARTICIPATING",
127
+ "208": "NON_PARTICIPATING",
128
+ "209": "SUCCESS",
129
+ "21": "NON_PARTICIPATING",
130
+ "210": "NON_PARTICIPATING",
131
+ "211": "NON_PARTICIPATING",
132
+ "212": "SUCCESS",
133
+ "213": "SUCCESS",
134
+ "214": "SUCCESS",
135
+ "215": "NON_PARTICIPATING",
136
+ "216": "SUCCESS",
137
+ "217": "NON_PARTICIPATING",
138
+ "218": "NON_PARTICIPATING",
139
+ "219": "NON_PARTICIPATING",
140
+ "22": "SUCCESS",
141
+ "220": "NON_PARTICIPATING",
142
+ "221": "SUCCESS",
143
+ "222": "NON_PARTICIPATING",
144
+ "223": "NON_PARTICIPATING",
145
+ "224": "NON_PARTICIPATING",
146
+ "225": "NON_PARTICIPATING",
147
+ "226": "NON_PARTICIPATING",
148
+ "227": "SUCCESS",
149
+ "228": "SUCCESS",
150
+ "229": "NON_PARTICIPATING",
151
+ "23": "NON_PARTICIPATING",
152
+ "230": "NON_PARTICIPATING",
153
+ "231": "NON_PARTICIPATING",
154
+ "232": "NON_PARTICIPATING",
155
+ "233": "NON_PARTICIPATING",
156
+ "234": "NON_PARTICIPATING",
157
+ "235": "NON_PARTICIPATING",
158
+ "236": "NON_PARTICIPATING",
159
+ "237": "SUCCESS",
160
+ "238": "NON_PARTICIPATING",
161
+ "239": "SUCCESS",
162
+ "24": "NON_PARTICIPATING",
163
+ "240": "SUCCESS",
164
+ "241": "SUCCESS",
165
+ "242": "NON_PARTICIPATING",
166
+ "243": "NON_PARTICIPATING",
167
+ "244": "NON_PARTICIPATING",
168
+ "245": "NON_PARTICIPATING",
169
+ "246": "NON_PARTICIPATING",
170
+ "247": "NON_PARTICIPATING",
171
+ "248": "NON_PARTICIPATING",
172
+ "249": "NON_PARTICIPATING",
173
+ "25": "SUCCESS",
174
+ "250": "SUCCESS",
175
+ "251": "SUCCESS",
176
+ "252": "NON_PARTICIPATING",
177
+ "253": "NON_PARTICIPATING",
178
+ "254": "NON_PARTICIPATING",
179
+ "255": "NON_PARTICIPATING",
180
+ "26": "SUCCESS",
181
+ "27": "NON_PARTICIPATING",
182
+ "28": "NON_PARTICIPATING",
183
+ "29": "NON_PARTICIPATING",
184
+ "3": "SUCCESS",
185
+ "30": "NON_PARTICIPATING",
186
+ "31": "NON_PARTICIPATING",
187
+ "32": "SUCCESS",
188
+ "33": "NON_PARTICIPATING",
189
+ "34": "NON_PARTICIPATING",
190
+ "35": "SUCCESS",
191
+ "36": "SUCCESS",
192
+ "37": "NON_PARTICIPATING",
193
+ "38": "NON_PARTICIPATING",
194
+ "39": "NON_PARTICIPATING",
195
+ "4": "NON_PARTICIPATING",
196
+ "40": "SUCCESS",
197
+ "41": "NON_PARTICIPATING",
198
+ "42": "NON_PARTICIPATING",
199
+ "43": "NON_PARTICIPATING",
200
+ "44": "NON_PARTICIPATING",
201
+ "45": "NON_PARTICIPATING",
202
+ "46": "NON_PARTICIPATING",
203
+ "47": "NON_PARTICIPATING",
204
+ "48": "NON_PARTICIPATING",
205
+ "49": "NON_PARTICIPATING",
206
+ "5": "NON_PARTICIPATING",
207
+ "50": "NON_PARTICIPATING",
208
+ "51": "NON_PARTICIPATING",
209
+ "52": "NON_PARTICIPATING",
210
+ "53": "NON_PARTICIPATING",
211
+ "54": "NON_PARTICIPATING",
212
+ "55": "NON_PARTICIPATING",
213
+ "56": "NON_PARTICIPATING",
214
+ "57": "SUCCESS",
215
+ "58": "NON_PARTICIPATING",
216
+ "59": "NON_PARTICIPATING",
217
+ "6": "SUCCESS",
218
+ "60": "SUCCESS",
219
+ "61": "SUCCESS",
220
+ "62": "SUCCESS",
221
+ "63": "NON_PARTICIPATING",
222
+ "64": "NON_PARTICIPATING",
223
+ "65": "NON_PARTICIPATING",
224
+ "66": "SUCCESS",
225
+ "67": "NON_PARTICIPATING",
226
+ "68": "NON_PARTICIPATING",
227
+ "69": "NON_PARTICIPATING",
228
+ "7": "SUCCESS",
229
+ "70": "SUCCESS",
230
+ "71": "NON_PARTICIPATING",
231
+ "72": "SUCCESS",
232
+ "73": "SUCCESS",
233
+ "74": "NON_PARTICIPATING",
234
+ "75": "SUCCESS",
235
+ "76": "SUCCESS",
236
+ "77": "NON_PARTICIPATING",
237
+ "78": "NON_PARTICIPATING",
238
+ "79": "NON_PARTICIPATING",
239
+ "8": "NON_PARTICIPATING",
240
+ "80": "NON_PARTICIPATING",
241
+ "81": "NON_PARTICIPATING",
242
+ "82": "NON_PARTICIPATING",
243
+ "83": "NON_PARTICIPATING",
244
+ "84": "SUCCESS",
245
+ "85": "SUCCESS",
246
+ "86": "NON_PARTICIPATING",
247
+ "87": "NON_PARTICIPATING",
248
+ "88": "NON_PARTICIPATING",
249
+ "89": "NON_PARTICIPATING",
250
+ "9": "SUCCESS",
251
+ "90": "NON_PARTICIPATING",
252
+ "91": "SUCCESS",
253
+ "92": "NON_PARTICIPATING",
254
+ "93": "NON_PARTICIPATING",
255
+ "94": "SUCCESS",
256
+ "95": "NON_PARTICIPATING",
257
+ "96": "NON_PARTICIPATING",
258
+ "97": "NON_PARTICIPATING",
259
+ "98": "SUCCESS",
260
+ "99": "NON_PARTICIPATING"
261
+ },
262
+ "architectures": [
263
+ "GPTOptim"
264
+ ],
265
+ "attn_pdrop": 0.1,
266
+ "auto_map": {
267
+ "AutoConfig": "configuration_gpt_optimized.GPTOptimConfig",
268
+ "AutoModelForCausalLM": "distributed/optimized-gpt2-2b--modeling_gpt_optimized.GPTOptim"
269
+ },
270
+ "block_size": 1024,
271
+ "bos_token_id": 50256,
272
+ "embd_pdrop": 0.1,
273
+ "eos_token_id": 50256,
274
+ "initializer_range": 0.02,
275
+ "layer_norm_epsilon": 1e-05,
276
+ "model_type": "gpt_optimized",
277
+ "n_embd": 1904,
278
+ "n_head": 34,
279
+ "n_inner": null,
280
+ "n_layer": 44,
281
+ "n_positions": 1024,
282
+ "quantization_config": {
283
+ "transformer.h.attn.c_attn": {
284
+ "offload_meta": false,
285
+ "scale_quant_params": null,
286
+ "weight_quant_params": {
287
+ "axis": 0,
288
+ "channel_wise": true,
289
+ "group_size": 64,
290
+ "nbits": 4,
291
+ "optimize": true,
292
+ "round_zero": true,
293
+ "view_as_float": false
294
+ },
295
+ "zero_quant_params": {
296
+ "channel_wise": false,
297
+ "group_size": null,
298
+ "nbits": 8,
299
+ "optimize": false
300
+ }
301
+ },
302
+ "transformer.h.attn.c_proj": {
303
+ "offload_meta": false,
304
+ "scale_quant_params": null,
305
+ "weight_quant_params": {
306
+ "axis": 0,
307
+ "channel_wise": true,
308
+ "group_size": 64,
309
+ "nbits": 4,
310
+ "optimize": true,
311
+ "round_zero": true,
312
+ "view_as_float": false
313
+ },
314
+ "zero_quant_params": {
315
+ "channel_wise": false,
316
+ "group_size": null,
317
+ "nbits": 8,
318
+ "optimize": false
319
+ }
320
+ },
321
+ "transformer.h.mlp.c_fc": {
322
+ "offload_meta": false,
323
+ "scale_quant_params": null,
324
+ "weight_quant_params": {
325
+ "axis": 0,
326
+ "channel_wise": true,
327
+ "group_size": 64,
328
+ "nbits": 4,
329
+ "optimize": true,
330
+ "round_zero": true,
331
+ "view_as_float": false
332
+ },
333
+ "zero_quant_params": {
334
+ "channel_wise": false,
335
+ "group_size": null,
336
+ "nbits": 8,
337
+ "optimize": false
338
+ }
339
+ },
340
+ "transformer.h.mlp.c_proj": {
341
+ "offload_meta": false,
342
+ "scale_quant_params": null,
343
+ "weight_quant_params": {
344
+ "axis": 0,
345
+ "channel_wise": true,
346
+ "group_size": 64,
347
+ "nbits": 4,
348
+ "optimize": true,
349
+ "round_zero": true,
350
+ "view_as_float": false
351
+ },
352
+ "zero_quant_params": {
353
+ "channel_wise": false,
354
+ "group_size": null,
355
+ "nbits": 8,
356
+ "optimize": false
357
+ }
358
+ }
359
+ },
360
+ "reorder_and_upcast_attn": false,
361
+ "resid_pdrop": 0.1,
362
+ "scale_attn_by_inverse_layer_idx": false,
363
+ "scale_attn_weights": true,
364
+ "summary_activation": null,
365
+ "summary_first_dropout": 0.1,
366
+ "summary_proj_to_labels": true,
367
+ "summary_type": "cls_index",
368
+ "summary_use_proj": true,
369
+ "torch_dtype": "float32",
370
+ "transformers_version": "4.44.0",
371
+ "use_cache": true,
372
+ "vocab_size": 50257
373
+ }
configuration_gpt_optimized.py ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from transformers import PretrainedConfig, GPT2Config
2
+ from typing import List
3
+
4
+
5
+ class GPTOptimConfig(GPT2Config):
6
+ model_type = "gpt_optimized"
7
+
8
+ def __init__(
9
+ self,
10
+ block_size: int = 1024, # max sequence length
11
+ vocab_size: int = 50257, # number of tokens: 50,000 BPE merges + 256 bytes tokens + 1 <|endoftext|> token
12
+ n_layer: int = 16, # number of layers
13
+ n_head: int = 16, # number of heads
14
+ n_embd: int = 1024, # embedding dimension
15
+ **kwargs,
16
+ ):
17
+ super().__init__(**kwargs)
18
+ self.block_size = block_size
19
+ self.vocab_size = vocab_size
20
+ self.n_layer = n_layer
21
+ self.n_head = n_head
22
+ self.n_embd = n_embd
qmodel.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53249ea6dc44533391f12b3d049f434029249e9d6174965151938c3fa0e97c13
3
+ size 1583744137
smash_config.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "comp_cgenerate_active": false,
3
+ "comp_ctranslate_active": false,
4
+ "comp_cwhisper_active": false,
5
+ "comp_diffusers2_active": false,
6
+ "comp_flux_caching_active": false,
7
+ "comp_ifw_active": false,
8
+ "comp_onediff_active": false,
9
+ "comp_step_caching_active": false,
10
+ "comp_torch_compile_active": false,
11
+ "comp_ws2t_active": false,
12
+ "comp_x-fast_active": false,
13
+ "prune_torch-structured_active": false,
14
+ "prune_torch-unstructured_active": false,
15
+ "quant_aqlm_active": false,
16
+ "quant_awq_active": false,
17
+ "quant_gptq_active": false,
18
+ "quant_half_active": false,
19
+ "quant_hqq_active": true,
20
+ "quant_llm-int8_active": false,
21
+ "quant_quanto_active": false,
22
+ "quant_torch_dynamic_active": false,
23
+ "quant_torch_static_active": false,
24
+ "quant_hqq_backend": "torchao_int4",
25
+ "quant_hqq_group_size": 64,
26
+ "quant_hqq_weight_bits": 4,
27
+ "max_batch_size": 1,
28
+ "device": "cuda",
29
+ "cache_dir": "/tmp/models/tmp1cq_dazl",
30
+ "task": "",
31
+ "save_load_fn": "hqq",
32
+ "save_load_fn_args": {},
33
+ "api_key": null
34
+ }