File size: 17,884 Bytes
9bd1fec
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
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
/home/dat/pino/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.
  warnings.warn(
/home/dat/pino/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.
  warnings.warn(
Epoch ... (1/5):   0%|                                    | 0/5 [00:00<?, ?it/s]
Traceback (most recent call last):                    | 0/46383 [00:00<?, ?it/s]
  File "/usr/lib/python3.8/threading.py", line 932, in _bootstrap_inner
    self.run()
  File "/usr/lib/python3.8/threading.py", line 870, in run
    self._target(*self._args, **self._kwargs)
  File "/home/dat/pino/lib/python3.8/site-packages/wandb/sdk/wandb_run.py", line 183, in check_network_status
    status_response = self._interface.communicate_network_status()
  File "/home/dat/pino/lib/python3.8/site-packages/wandb/sdk/interface/interface.py", line 755, in communicate_network_status
    resp = self._communicate(req, timeout=timeout, local=True)
  File "/home/dat/pino/lib/python3.8/site-packages/wandb/sdk/interface/interface.py", line 545, in _communicate
    return self._communicate_async(rec, local=local).get(timeout=timeout)
  File "/home/dat/pino/lib/python3.8/site-packages/wandb/sdk/interface/interface.py", line 550, in _communicate_async
    raise Exception("The wandb backend process has shutdown")
Exception: The wandb backend process has shutdown
Training...:   0%|                                    | 0/46383 [01:25<?, ?it/s]
Epoch ... (1/5):   0%|                                    | 0/5 [02:13<?, ?it/s]
Traceback (most recent call last):
  File "/home/dat/pino/lib/python3.8/site-packages/jax/_src/traceback_util.py", line 183, in reraise_with_filtered_traceback
    return fun(*args, **kwargs)
  File "/home/dat/pino/lib/python3.8/site-packages/jax/_src/api.py", line 1647, in f_pmapped
    out = pxla.xla_pmap(
  File "/home/dat/pino/lib/python3.8/site-packages/jax/core.py", line 1620, in bind
    return call_bind(self, fun, *args, **params)
  File "/home/dat/pino/lib/python3.8/site-packages/jax/core.py", line 1551, in call_bind
    outs = primitive.process(top_trace, fun, tracers, params)
  File "/home/dat/pino/lib/python3.8/site-packages/jax/core.py", line 1623, in process
    return trace.process_map(self, fun, tracers, params)
  File "/home/dat/pino/lib/python3.8/site-packages/jax/core.py", line 606, in process_call
    return primitive.impl(f, *tracers, **params)
  File "/home/dat/pino/lib/python3.8/site-packages/jax/interpreters/pxla.py", line 624, in xla_pmap_impl
    compiled_fun, fingerprint = parallel_callable(fun, backend, axis_name, axis_size,
  File "/home/dat/pino/lib/python3.8/site-packages/jax/linear_util.py", line 262, in memoized_fun
    ans = call(fun, *args)
  File "/home/dat/pino/lib/python3.8/site-packages/jax/interpreters/pxla.py", line 899, in parallel_callable
    compiled = xla.backend_compile(backend, built, compile_options)
  File "/home/dat/pino/lib/python3.8/site-packages/jax/interpreters/xla.py", line 360, in backend_compile
    return backend.compile(built_c, compile_options=options)
RuntimeError: Resource exhausted: Ran out of memory in memory space hbm. Used 20.30G of 15.48G hbm. Exceeded hbm capacity by 4.82G.
Total hbm usage >= 20.82G:
    reserved        530.00M
    program          20.30G
    arguments            0B
Output size 0B; shares 0B with arguments.
Program hbm requirement 20.30G:
    global           660.0K
    scoped           125.0K
    HLO temp         20.30G (63.5% utilization: Unpadded (12.44G) Padded (19.60G), 3.5% fragmentation (717.54M))
  Largest program allocations in hbm:
  1. Size: 1.54G
     Operator: op_type="dot_general" op_name="pmap(train_step)/dot_general[ dimension_numbers=(((2,), (0,)), ((), ()))\n                              precision=None\n                              preferred_element_type=None ]" source_file="/home/dat/pino/lib/python3.8/site-packages/flax/linen/linear.py" source_line=175
     Shape: bf16[4,4096,50358]{1,2,0:T(8,128)(2,1)}
     Unpadded size: 1.54G
     Extra memory due to padding: 64.0K (1.0x expansion)
     XLA label: %fusion.1304.remat4 = bf16[4,4096,50358]{1,2,0:T(8,128)(2,1)} fusion(bf16[50358,768]{1,0:T(8,128)(2,1)} %copy.16213, f32[768]{0:T(1024)} %fusion.8859, f32[768]{0:T(1024)} %fusion.8860, f32[4,4096]{1,0:T(4,128)} %get-tuple-element.16597, f32[4,4096]{1,0:T(4...
     Allocation type: HLO temp
     ==========================
  2. Size: 360.00M
     Operator: op_type="div" op_name="pmap(train_step)/div" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=619
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.135 = bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)} fusion(f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.485, f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.5710, f32[4,12,60,64,64]{3,4,2,1,0:T(8,128)} %get-tuple-element.16812, f32[4,12,60,64,192]{3,4,2,1,0...
     Allocation type: HLO temp
     ==========================
  3. Size: 360.00M
     Operator: op_type="div" op_name="pmap(train_step)/div" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=619
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.144.remat = (bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}, bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}) fusion(f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.494, f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.5719, f32[4,12,60,64,64]{3,4,2,1,0:T(8,128)} ...
     Allocation type: HLO temp
     ==========================
  4. Size: 360.00M
     Operator: op_type="div" op_name="pmap(train_step)/div" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=619
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.143.remat = (bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}, bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}) fusion(f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.493, f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.5718, f32[4,12,60,64,64]{3,4,2,1,0:T(8,128)} ...
     Allocation type: HLO temp
     ==========================
  5. Size: 360.00M
     Operator: op_type="div" op_name="pmap(train_step)/div" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=619
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.142.remat = (bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}, bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}) fusion(f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.492, f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.5717, f32[4,12,60,64,64]{3,4,2,1,0:T(8,128)} ...
     Allocation type: HLO temp
     ==========================
  6. Size: 360.00M
     Operator: op_type="div" op_name="pmap(train_step)/div" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=619
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.141.remat = (bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}, bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}) fusion(f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.491, f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.5716, f32[4,12,60,64,64]{3,4,2,1,0:T(8,128)} ...
     Allocation type: HLO temp
     ==========================
  7. Size: 360.00M
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.134.remat_uncompressed = bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)} copy(bf16[4,12,60,64,512]{4,3,2,1,0:T(8,128)(2,1)} %fusion.134.remat_compressed)
     Allocation type: HLO temp
     ==========================
  8. Size: 360.00M
     Operator: op_type="div" op_name="pmap(train_step)/div" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=619
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.140.remat = (bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}, bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}) fusion(f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.490, f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.5715, f32[4,12,60,64,64]{3,4,2,1,0:T(8,128)} ...
     Allocation type: HLO temp
     ==========================
  9. Size: 360.00M
     Operator: op_type="div" op_name="pmap(train_step)/div" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=619
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.139.remat = (bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}, bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}) fusion(f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.489, f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.5714, f32[4,12,60,64,64]{3,4,2,1,0:T(8,128)} ...
     Allocation type: HLO temp
     ==========================
  10. Size: 360.00M
     Operator: op_type="div" op_name="pmap(train_step)/div" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=619
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.138.remat = (bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}, bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}) fusion(f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.488, f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.5713, f32[4,12,60,64,64]{3,4,2,1,0:T(8,128)} ...
     Allocation type: HLO temp
     ==========================
  11. Size: 360.00M
     Operator: op_type="div" op_name="pmap(train_step)/div" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=619
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.137.remat = (bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}, bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}) fusion(f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.487, f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.5712, f32[4,12,60,64,64]{3,4,2,1,0:T(8,128)} ...
     Allocation type: HLO temp
     ==========================
  12. Size: 360.00M
     Operator: op_type="div" op_name="pmap(train_step)/div" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=619
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.136.remat = (bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}, bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}) fusion(f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.486, f32[4,12,60,64]{3,2,1,0:T(8,128)} %fusion.5711, f32[4,12,60,64,64]{3,4,2,1,0:T(8,128)} ...
     Allocation type: HLO temp
     ==========================
  13. Size: 360.00M
     Shape: bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)}
     Unpadded size: 180.00M
     Extra memory due to padding: 180.00M (2.0x expansion)
     XLA label: %fusion.133.remat_uncompressed = bf16[4,12,60,64,512]{3,4,2,1,0:T(8,128)(2,1)} copy(bf16[4,12,60,64,512]{4,3,2,1,0:T(8,128)(2,1)} %fusion.133.remat_compressed)
     Allocation type: HLO temp
     ==========================
  14. Size: 270.00M
     Operator: op_type="dot_general" op_name="pmap(train_step)/jit(jvp(_einsum))/dot_general[ dimension_numbers=(((4,), (4,)), ((0, 1, 2), (0, 1, 2)))\n                                                precision=None\n                                                preferred_element_type=None ]" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=584
     Shape: f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)}
     Unpadded size: 135.00M
     Extra memory due to padding: 135.00M (2.0x expansion)
     XLA label: %fusion.378.remat5 = f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)} fusion(f32[4,60,64,192]{2,3,1,0:T(8,128)} %get-tuple-element.17038, bf16[4,12,64,64,64]{4,3,2,1,0:T(8,128)(2,1)} %copy.14428, bf16[4,12,60,192,64]{3,2,4,1,0:T(8,128)(2,1)} %fusion.655), kind=kOut...
     Allocation type: HLO temp
     ==========================
  15. Size: 270.00M
     Shape: f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)}
     Unpadded size: 135.00M
     Extra memory due to padding: 135.00M (2.0x expansion)
     XLA label: %fusion.310.remat_uncompressed = f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)} copy(f32[4,12,60,64,192]{4,3,2,1,0:T(8,128)} %fusion.310.remat_compressed)
     Allocation type: HLO temp
     ==========================
  16. Size: 270.00M
     Operator: op_type="dot_general" op_name="pmap(train_step)/jit(jvp(_einsum))/dot_general[ dimension_numbers=(((4,), (4,)), ((0, 1, 2), (0, 1, 2)))\n                                                precision=None\n                                                preferred_element_type=None ]" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=584
     Shape: f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)}
     Unpadded size: 135.00M
     Extra memory due to padding: 135.00M (2.0x expansion)
     XLA label: %fusion.386.remat6 = f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)} fusion(f32[4,60,64,192]{2,3,1,0:T(8,128)} %get-tuple-element.17038, bf16[4,12,64,64,64]{4,3,2,1,0:T(8,128)(2,1)} %copy.13900, bf16[4,12,60,192,64]{3,2,4,1,0:T(8,128)(2,1)} %fusion.639), kind=kOut...
     Allocation type: HLO temp
     ==========================
  17. Size: 270.00M
     Shape: f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)}
     Unpadded size: 135.00M
     Extra memory due to padding: 135.00M (2.0x expansion)
     XLA label: %fusion.326.remat_uncompressed.remat2 = f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)} copy(f32[4,12,60,64,192]{4,3,2,1,0:T(8,128)} %fusion.326.remat_compressed)
     Allocation type: HLO temp
     ==========================
  18. Size: 270.00M
     Operator: op_type="dot_general" op_name="pmap(train_step)/jit(jvp(_einsum))/dot_general[ dimension_numbers=(((4,), (4,)), ((0, 1, 2), (0, 1, 2)))\n                                                precision=None\n                                                preferred_element_type=None ]" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=591
     Shape: f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)}
     Unpadded size: 135.00M
     Extra memory due to padding: 135.00M (2.0x expansion)
     XLA label: %fusion.10361 = (f32[4,12,60,64]{3,2,1,0:T(8,128)}, f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)}) fusion(s32[4,12,62,64,192]{3,4,2,1,0:T(8,128)} %get-tuple-element.18295, bf16[4,12,64,64,64]{4,3,2,1,0:T(8,128)(2,1)} %copy.14494, bf16[4,12,60,192,64]{3,2,1,0,4:T...
     Allocation type: HLO temp
     ==========================
  19. Size: 270.00M
     Operator: op_type="dot_general" op_name="pmap(train_step)/jit(jvp(_einsum))/dot_general[ dimension_numbers=(((4,), (4,)), ((0, 1, 2), (0, 1, 2)))\n                                                precision=None\n                                                preferred_element_type=None ]" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=584
     Shape: f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)}
     Unpadded size: 135.00M
     Extra memory due to padding: 135.00M (2.0x expansion)
     XLA label: %fusion.380.remat5 = f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)} fusion(f32[4,60,64,192]{2,3,1,0:T(8,128)} %get-tuple-element.17038, bf16[4,12,64,64,64]{4,3,2,1,0:T(8,128)(2,1)} %copy.14296, bf16[4,12,60,192,64]{3,2,4,1,0:T(8,128)(2,1)} %fusion.651), kind=kOut...
     Allocation type: HLO temp
     ==========================
  20. Size: 270.00M
     Operator: op_type="dot_general" op_name="pmap(train_step)/jit(jvp(_einsum))/dot_general[ dimension_numbers=(((4,), (4,)), ((0, 1, 2), (0, 1, 2)))\n                                                precision=None\n                                                preferred_element_type=None ]" source_file="/home/dat/transformers/src/transformers/models/big_bird/modeling_flax_big_bird.py" source_line=584
     Shape: f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)}
     Unpadded size: 135.00M
     Extra memory due to padding: 135.00M (2.0x expansion)
     XLA label: %fusion.379.remat3 = f32[4,12,60,64,192]{3,4,2,1,0:T(8,128)} fusion(f32[4,60,64,192]{2,3,1,0:T(8,128)} %get-tuple-element.17038, bf16[4,12,64,64,64]{4,3,2,1,0:T(8,128)(2,1)} %copy.14362, bf16[4,12,60,192,64]{3,2,4,1,0:T(8,128)(2,1)} %fusion.653), kind=kOut...
     Allocation type: HLO temp
     ==========================
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
  File "./run_mlm_flax.py", line 709, in <module>
    state, train_metric, dropout_rngs = p_train_step(state, model_inputs, dropout_rngs)
  File "/home/dat/pino/lib/python3.8/site-packages/jax/_src/traceback_util.py", line 183, in reraise_with_filtered_traceback
    return fun(*args, **kwargs)
KeyboardInterrupt