nipunsadvilkar
commited on
Commit
•
c3a9018
1
Parent(s):
47ebced
Saving weights and logs of step 500
Browse files
run.log
ADDED
@@ -0,0 +1,273 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
0 |
0%| | 0/5 [00:00<?, ?ba/s]
|
1 |
20%|██ | 1/5 [00:00<00:00, 8.80ba/s]
|
2 |
40%|████ | 2/5 [00:00<00:00, 9.16ba/s]
|
3 |
60%|██████ | 3/5 [00:00<00:00, 9.52ba/s]
|
4 |
80%|████████ | 4/5 [00:00<00:00, 7.13ba/s]
|
|
|
5 |
0%| | 0/1 [00:00<?, ?ba/s]
|
|
|
6 |
0%| | 0/5 [00:00<?, ?ba/s]
|
7 |
20%|██ | 1/5 [00:00<00:03, 1.32ba/s]
|
8 |
40%|████ | 2/5 [00:01<00:02, 1.24ba/s]
|
9 |
60%|██████ | 3/5 [00:02<00:01, 1.21ba/s]
|
10 |
80%|████████ | 4/5 [00:03<00:00, 1.13ba/s]
|
|
|
11 |
0%| | 0/1 [00:00<?, ?ba/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
|
|
13 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
[16:04:24] - INFO - __main__ - Training/evaluation parameters TrainingArguments(
|
2 |
+
_n_gpu=-1,
|
3 |
+
adafactor=False,
|
4 |
+
adam_beta1=0.9,
|
5 |
+
adam_beta2=0.999,
|
6 |
+
adam_epsilon=1e-08,
|
7 |
+
dataloader_drop_last=False,
|
8 |
+
dataloader_num_workers=0,
|
9 |
+
dataloader_pin_memory=True,
|
10 |
+
ddp_find_unused_parameters=None,
|
11 |
+
debug=[],
|
12 |
+
deepspeed=None,
|
13 |
+
disable_tqdm=False,
|
14 |
+
do_eval=False,
|
15 |
+
do_predict=False,
|
16 |
+
do_train=False,
|
17 |
+
eval_accumulation_steps=None,
|
18 |
+
eval_steps=500,
|
19 |
+
evaluation_strategy=IntervalStrategy.NO,
|
20 |
+
fp16=False,
|
21 |
+
fp16_backend=auto,
|
22 |
+
fp16_full_eval=False,
|
23 |
+
fp16_opt_level=O1,
|
24 |
+
gradient_accumulation_steps=1,
|
25 |
+
greater_is_better=None,
|
26 |
+
group_by_length=False,
|
27 |
+
ignore_data_skip=False,
|
28 |
+
label_names=None,
|
29 |
+
label_smoothing_factor=0.0,
|
30 |
+
learning_rate=0.0003,
|
31 |
+
length_column_name=length,
|
32 |
+
load_best_model_at_end=False,
|
33 |
+
local_rank=-1,
|
34 |
+
log_level=-1,
|
35 |
+
log_level_replica=-1,
|
36 |
+
log_on_each_node=True,
|
37 |
+
logging_dir=./runs/Jul08_16-04-24_t1v-n-112df4a9-w-0,
|
38 |
+
logging_first_step=False,
|
39 |
+
logging_steps=500,
|
40 |
+
logging_strategy=IntervalStrategy.STEPS,
|
41 |
+
lr_scheduler_type=SchedulerType.LINEAR,
|
42 |
+
max_grad_norm=1.0,
|
43 |
+
max_steps=-1,
|
44 |
+
metric_for_best_model=None,
|
45 |
+
mp_parameters=,
|
46 |
+
no_cuda=False,
|
47 |
+
num_train_epochs=8.0,
|
48 |
+
output_dir=./,
|
49 |
+
overwrite_output_dir=True,
|
50 |
+
past_index=-1,
|
51 |
+
per_device_eval_batch_size=4,
|
52 |
+
per_device_train_batch_size=4,
|
53 |
+
prediction_loss_only=False,
|
54 |
+
push_to_hub=True,
|
55 |
+
push_to_hub_model_id=flax-community/roberta-base-mr,
|
56 |
+
push_to_hub_organization=None,
|
57 |
+
push_to_hub_token=vdIAyRvCACJNslYtyLHufmNDnUIyknPzUgVDMFiXqJoulvMqjoubonLJzXOJQJczWfRMJumVaMFjGSFVnQAMdswvZkzNIthKrxBeARBXfqnIwjABkKpCbjGEgnkjpjKi,
|
58 |
+
remove_unused_columns=True,
|
59 |
+
report_to=[],
|
60 |
+
resume_from_checkpoint=None,
|
61 |
+
run_name=./,
|
62 |
+
save_on_each_node=False,
|
63 |
+
save_steps=500,
|
64 |
+
save_strategy=IntervalStrategy.STEPS,
|
65 |
+
save_total_limit=None,
|
66 |
+
seed=42,
|
67 |
+
sharded_ddp=[],
|
68 |
+
skip_memory_metrics=True,
|
69 |
+
tpu_metrics_debug=False,
|
70 |
+
tpu_num_cores=None,
|
71 |
+
use_legacy_prediction_loop=False,
|
72 |
+
warmup_ratio=0.0,
|
73 |
+
warmup_steps=1000,
|
74 |
+
weight_decay=0.0,
|
75 |
+
)
|
76 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - Starting new HTTPS connection (1): s3.amazonaws.com:443
|
77 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - https://s3.amazonaws.com:443 "HEAD /datasets.huggingface.co/datasets/datasets/oscar/oscar.py HTTP/1.1" 404 0
|
78 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - Starting new HTTPS connection (1): raw.githubusercontent.com:443
|
79 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - https://raw.githubusercontent.com:443 "HEAD /huggingface/datasets/master/datasets/oscar/oscar.py HTTP/1.1" 200 0
|
80 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - Starting new HTTPS connection (1): raw.githubusercontent.com:443
|
81 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - https://raw.githubusercontent.com:443 "HEAD /huggingface/datasets/master/datasets/oscar/dataset_infos.json HTTP/1.1" 200 0
|
82 |
+
[16:04:24] - WARNING - datasets.builder - Reusing dataset oscar (/home/nipunsadvilkar/.cache/huggingface/datasets/oscar/unshuffled_deduplicated_als/1.0.0/84838bd49d2295f62008383b05620571535451d84545037bb94d6f3501651df2)
|
83 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - Starting new HTTPS connection (1): s3.amazonaws.com:443
|
84 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - https://s3.amazonaws.com:443 "HEAD /datasets.huggingface.co/datasets/datasets/oscar/oscar.py HTTP/1.1" 404 0
|
85 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - Starting new HTTPS connection (1): raw.githubusercontent.com:443
|
86 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - https://raw.githubusercontent.com:443 "HEAD /huggingface/datasets/master/datasets/oscar/oscar.py HTTP/1.1" 200 0
|
87 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - Starting new HTTPS connection (1): raw.githubusercontent.com:443
|
88 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - https://raw.githubusercontent.com:443 "HEAD /huggingface/datasets/master/datasets/oscar/dataset_infos.json HTTP/1.1" 200 0
|
89 |
+
[16:04:24] - WARNING - datasets.builder - Reusing dataset oscar (/home/nipunsadvilkar/.cache/huggingface/datasets/oscar/unshuffled_deduplicated_als/1.0.0/84838bd49d2295f62008383b05620571535451d84545037bb94d6f3501651df2)
|
90 |
+
[16:04:24] - DEBUG - urllib3.connectionpool - Starting new HTTPS connection (1): s3.amazonaws.com:443
|
91 |
+
[16:04:25] - DEBUG - urllib3.connectionpool - https://s3.amazonaws.com:443 "HEAD /datasets.huggingface.co/datasets/datasets/oscar/oscar.py HTTP/1.1" 404 0
|
92 |
+
[16:04:25] - DEBUG - urllib3.connectionpool - Starting new HTTPS connection (1): raw.githubusercontent.com:443
|
93 |
+
[16:04:25] - DEBUG - urllib3.connectionpool - https://raw.githubusercontent.com:443 "HEAD /huggingface/datasets/master/datasets/oscar/oscar.py HTTP/1.1" 200 0
|
94 |
+
[16:04:25] - DEBUG - urllib3.connectionpool - Starting new HTTPS connection (1): raw.githubusercontent.com:443
|
95 |
+
[16:04:25] - DEBUG - urllib3.connectionpool - https://raw.githubusercontent.com:443 "HEAD /huggingface/datasets/master/datasets/oscar/dataset_infos.json HTTP/1.1" 200 0
|
96 |
+
[16:04:25] - WARNING - datasets.builder - Reusing dataset oscar (/home/nipunsadvilkar/.cache/huggingface/datasets/oscar/unshuffled_deduplicated_als/1.0.0/84838bd49d2295f62008383b05620571535451d84545037bb94d6f3501651df2)
|
97 |
+
|
98 |
0%| | 0/5 [00:00<?, ?ba/s]
|
99 |
20%|██ | 1/5 [00:00<00:00, 8.80ba/s]
|
100 |
40%|████ | 2/5 [00:00<00:00, 9.16ba/s]
|
101 |
60%|██████ | 3/5 [00:00<00:00, 9.52ba/s]
|
102 |
80%|████████ | 4/5 [00:00<00:00, 7.13ba/s]
|
103 |
+
|
104 |
0%| | 0/1 [00:00<?, ?ba/s]
|
105 |
+
|
106 |
0%| | 0/5 [00:00<?, ?ba/s]
|
107 |
20%|██ | 1/5 [00:00<00:03, 1.32ba/s]
|
108 |
40%|████ | 2/5 [00:01<00:02, 1.24ba/s]
|
109 |
60%|██████ | 3/5 [00:02<00:01, 1.21ba/s]
|
110 |
80%|████████ | 4/5 [00:03<00:00, 1.13ba/s]
|
111 |
+
|
112 |
0%| | 0/1 [00:00<?, ?ba/s]
|
113 |
+
[16:04:29] - WARNING - __main__ - Unable to display metrics through TensorBoard because the package is not installed: Please run pip install tensorboard to enable.
|
114 |
+
[16:04:29] - INFO - absl - Starting the local TPU driver.
|
115 |
+
[16:04:29] - INFO - absl - Unable to initialize backend 'tpu_driver': Not found: Unable to find driver in registry given worker: local://
|
116 |
+
[16:04:29] - INFO - absl - Unable to initialize backend 'gpu': Not found: Could not find registered platform with name: "cuda". Available platform names are: Interpreter TPU Host
|
117 |
+
/home/nipunsadvilkar/roberta_mr_env/lib/python3.8/site-packages/jax/lib/xla_bridge.py:382: UserWarning: jax.host_count has been renamed to jax.process_count. This alias will eventually be removed; please update your code.
|
118 |
+
warnings.warn(
|
119 |
+
/home/nipunsadvilkar/roberta_mr_env/lib/python3.8/site-packages/jax/lib/xla_bridge.py:369: UserWarning: jax.host_id has been renamed to jax.process_index. This alias will eventually be removed; please update your code.
|
120 |
+
warnings.warn(
|
121 |
+
|
122 |
+
|
123 |
+
|
124 |
+
|
125 |
+
|
126 |
+
|
127 |
+
|
128 |
+
|
129 |
+
|
130 |
+
|
131 |
+
|
132 |
+
|
133 |
+
|
134 |
+
|
135 |
+
|
136 |
+
|
137 |
+
|
138 |
+
|
139 |
+
|
140 |
+
|
141 |
+
|
142 |
+
|
143 |
+
|
144 |
+
|
145 |
+
|
146 |
+
|
147 |
+
|
148 |
+
|
149 |
+
|
150 |
+
|
151 |
+
|
152 |
+
|
153 |
+
|
154 |
+
|
155 |
+
|
156 |
+
|
157 |
+
|
158 |
+
|
159 |
+
|
160 |
+
|
161 |
+
|
162 |
+
|
163 |
+
|
164 |
+
|
165 |
+
|
166 |
+
|
167 |
+
|
168 |
+
|
169 |
+
|
170 |
+
|
171 |
+
|
172 |
+
|
173 |
+
|
174 |
+
|
175 |
+
|
176 |
+
|
177 |
+
|
178 |
+
|
179 |
+
|
180 |
+
|
181 |
+
|
182 |
+
|
183 |
+
|
184 |
+
|
185 |
+
|
186 |
+
|
187 |
+
|
188 |
+
|
189 |
+
|
190 |
+
|
191 |
+
|
192 |
+
|
193 |
+
|
194 |
+
|
195 |
+
|
196 |
+
|
197 |
+
|
198 |
+
|
199 |
+
|
200 |
+
|
201 |
+
|
202 |
+
|
203 |
+
|
204 |
+
|
205 |
+
|
206 |
+
|
207 |
+
|
208 |
+
|
209 |
+
|
210 |
+
|
211 |
+
|
212 |
+
|
213 |
+
|
214 |
+
|
215 |
+
|
216 |
+
|
217 |
+
|
218 |
+
|
219 |
+
|
220 |
+
|
221 |
+
|
222 |
+
|
223 |
+
|
224 |
+
|
225 |
+
|
226 |
+
|
227 |
+
|
228 |
+
|
229 |
+
|
230 |
+
|
231 |
+
|
232 |
+
|
233 |
+
|
234 |
+
|
235 |
+
|
236 |
+
|
237 |
+
|
238 |
+
|
239 |
+
|
240 |
+
|
241 |
+
|
242 |
+
|
243 |
+
|
244 |
+
|
245 |
+
|
246 |
+
|
247 |
+
|
248 |
+
|
249 |
+
|
250 |
+
|
251 |
+
|
252 |
+
|
253 |
+
|
254 |
|
255 |
+
|
256 |
|
257 |
+
|
258 |
+
|
259 |
+
|
260 |
+
|
261 |
+
|
262 |
+
|
263 |
+
|
264 |
+
|
265 |
+
To disable this warning, you can either:
|
266 |
+
- Avoid using `tokenizers` before the fork if possible
|
267 |
+
- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
|
268 |
+
huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
|
269 |
+
To disable this warning, you can either:
|
270 |
+
- Avoid using `tokenizers` before the fork if possible
|
271 |
+
- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
|
272 |
+
|
273 |
+
[16:06:12] - INFO - huggingface_hub.repository - git version 2.25.1
|
274 |
+
git-lfs/2.9.2 (GitHub; linux amd64; go 1.13.5)
|
275 |
+
huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
|
276 |
+
To disable this warning, you can either:
|
277 |
+
- Avoid using `tokenizers` before the fork if possible
|
278 |
+
- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
|
279 |
+
[16:06:12] - DEBUG - huggingface_hub.repository - [Repository] is a valid git repo
|
280 |
+
huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
|
281 |
+
To disable this warning, you can either:
|
282 |
+
- Avoid using `tokenizers` before the fork if possible
|
283 |
+
- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
|
284 |
+
huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...
|
285 |
+
To disable this warning, you can either:
|
286 |
+
- Avoid using `tokenizers` before the fork if possible
|
287 |
+
- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)
|
run.sh
ADDED
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
HUB_TOKEN=`cat $HOME/.huggingface/token`
|
2 |
+
python run_mlm_flax.py \
|
3 |
+
--output_dir="./" \
|
4 |
+
--model_type="roberta" \
|
5 |
+
--config_name="./" \
|
6 |
+
--tokenizer_name="./" \
|
7 |
+
--dataset_name="oscar" \
|
8 |
+
--dataset_config_name="unshuffled_deduplicated_als" \
|
9 |
+
--max_seq_length="128" \
|
10 |
+
--per_device_train_batch_size="4" \
|
11 |
+
--per_device_eval_batch_size="4" \
|
12 |
+
--learning_rate="3e-4" \
|
13 |
+
--warmup_steps="1000" \
|
14 |
+
--overwrite_output_dir \
|
15 |
+
--num_train_epochs="8" \
|
16 |
+
--push_to_hub_model_id="flax-community/roberta-base-mr" \
|
17 |
+
--push_to_hub_token="$HUB_TOKEN" \
|
18 |
+
--push_to_hub 2>&1 | tee run.log
|
tokens.py
ADDED
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python3
|
2 |
+
from datasets import load_dataset
|
3 |
+
from tokenizers import ByteLevelBPETokenizer
|
4 |
+
# Load dataset
|
5 |
+
dataset = load_dataset("oscar", "unshuffled_deduplicated_als", split="train")
|
6 |
+
# Instantiate tokenizer
|
7 |
+
tokenizer = ByteLevelBPETokenizer()
|
8 |
+
def batch_iterator(batch_size=100_000):
|
9 |
+
for i in range(0, len(dataset), batch_size):
|
10 |
+
yield dataset["text"][i: i + batch_size]
|
11 |
+
# Customized training
|
12 |
+
tokenizer.train_from_iterator(batch_iterator(), vocab_size=50265, min_frequency=2, special_tokens=[
|
13 |
+
"<s>",
|
14 |
+
"<pad>",
|
15 |
+
"</s>",
|
16 |
+
"<unk>",
|
17 |
+
"<mask>",
|
18 |
+
])
|
19 |
+
# Save files to disk
|
20 |
+
tokenizer.save("./tokenizer.json")
|