Hugging Face's logo

zhangcen456
/
baichuan2_ft

PEFT
Model card Files Community
zhangcen456 commited on
Commit
529b4c8
•
1 Parent(s): 962bf8b

Delete checkpoint-354

Browse files
Files changed (22) hide show
  1. checkpoint-354/README.md +0 -208
  2. checkpoint-354/adapter_config.json +0 -22
  3. checkpoint-354/adapter_model.bin +0 -3
  4. checkpoint-354/global_step354/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt +0 -3
  5. checkpoint-354/global_step354/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt +0 -3
  6. checkpoint-354/global_step354/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt +0 -3
  7. checkpoint-354/global_step354/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt +0 -3
  8. checkpoint-354/global_step354/zero_pp_rank_0_mp_rank_00_model_states.pt +0 -3
  9. checkpoint-354/global_step354/zero_pp_rank_1_mp_rank_00_model_states.pt +0 -3
  10. checkpoint-354/global_step354/zero_pp_rank_2_mp_rank_00_model_states.pt +0 -3
  11. checkpoint-354/global_step354/zero_pp_rank_3_mp_rank_00_model_states.pt +0 -3
  12. checkpoint-354/latest +0 -1
  13. checkpoint-354/rng_state_0.pth +0 -3
  14. checkpoint-354/rng_state_1.pth +0 -3
  15. checkpoint-354/rng_state_2.pth +0 -3
  16. checkpoint-354/rng_state_3.pth +0 -3
  17. checkpoint-354/special_tokens_map.json +0 -30
  18. checkpoint-354/tokenizer.model +0 -3
  19. checkpoint-354/tokenizer_config.json +0 -47
  20. checkpoint-354/trainer_state.json +0 -2143
  21. checkpoint-354/training_args.bin +0 -3
  22. checkpoint-354/zero_to_fp32.py +0 -587
checkpoint-354/README.md DELETED
@@ -1,208 +0,0 @@
1
- ---
2
- library_name: peft
3
- base_model: baichuan-inc/Baichuan2-7B-Base
4
- ---
5
-
6
- # Model Card for Model ID
7
-
8
- <!-- Provide a quick summary of what the model is/does. -->
9
-
10
-
11
-
12
- ## Model Details
13
-
14
- ### Model Description
15
-
16
- <!-- Provide a longer summary of what this model is. -->
17
-
18
-
19
-
20
- - **Developed by:** [More Information Needed]
21
- - **Funded by [optional]:** [More Information Needed]
22
- - **Shared by [optional]:** [More Information Needed]
23
- - **Model type:** [More Information Needed]
24
- - **Language(s) (NLP):** [More Information Needed]
25
- - **License:** [More Information Needed]
26
- - **Finetuned from model [optional]:** [More Information Needed]
27
-
28
- ### Model Sources [optional]
29
-
30
- <!-- Provide the basic links for the model. -->
31
-
32
- - **Repository:** [More Information Needed]
33
- - **Paper [optional]:** [More Information Needed]
34
- - **Demo [optional]:** [More Information Needed]
35
-
36
- ## Uses
37
-
38
- <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
39
-
40
- ### Direct Use
41
-
42
- <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
43
-
44
- [More Information Needed]
45
-
46
- ### Downstream Use [optional]
47
-
48
- <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
49
-
50
- [More Information Needed]
51
-
52
- ### Out-of-Scope Use
53
-
54
- <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
55
-
56
- [More Information Needed]
57
-
58
- ## Bias, Risks, and Limitations
59
-
60
- <!-- This section is meant to convey both technical and sociotechnical limitations. -->
61
-
62
- [More Information Needed]
63
-
64
- ### Recommendations
65
-
66
- <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
67
-
68
- Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
69
-
70
- ## How to Get Started with the Model
71
-
72
- Use the code below to get started with the model.
73
-
74
- [More Information Needed]
75
-
76
- ## Training Details
77
-
78
- ### Training Data
79
-
80
- <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
81
-
82
- [More Information Needed]
83
-
84
- ### Training Procedure
85
-
86
- <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
87
-
88
- #### Preprocessing [optional]
89
-
90
- [More Information Needed]
91
-
92
-
93
- #### Training Hyperparameters
94
-
95
- - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
96
-
97
- #### Speeds, Sizes, Times [optional]
98
-
99
- <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
100
-
101
- [More Information Needed]
102
-
103
- ## Evaluation
104
-
105
- <!-- This section describes the evaluation protocols and provides the results. -->
106
-
107
- ### Testing Data, Factors & Metrics
108
-
109
- #### Testing Data
110
-
111
- <!-- This should link to a Dataset Card if possible. -->
112
-
113
- [More Information Needed]
114
-
115
- #### Factors
116
-
117
- <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
118
-
119
- [More Information Needed]
120
-
121
- #### Metrics
122
-
123
- <!-- These are the evaluation metrics being used, ideally with a description of why. -->
124
-
125
- [More Information Needed]
126
-
127
- ### Results
128
-
129
- [More Information Needed]
130
-
131
- #### Summary
132
-
133
-
134
-
135
- ## Model Examination [optional]
136
-
137
- <!-- Relevant interpretability work for the model goes here -->
138
-
139
- [More Information Needed]
140
-
141
- ## Environmental Impact
142
-
143
- <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
144
-
145
- Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
146
-
147
- - **Hardware Type:** [More Information Needed]
148
- - **Hours used:** [More Information Needed]
149
- - **Cloud Provider:** [More Information Needed]
150
- - **Compute Region:** [More Information Needed]
151
- - **Carbon Emitted:** [More Information Needed]
152
-
153
- ## Technical Specifications [optional]
154
-
155
- ### Model Architecture and Objective
156
-
157
- [More Information Needed]
158
-
159
- ### Compute Infrastructure
160
-
161
- [More Information Needed]
162
-
163
- #### Hardware
164
-
165
- [More Information Needed]
166
-
167
- #### Software
168
-
169
- [More Information Needed]
170
-
171
- ## Citation [optional]
172
-
173
- <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
174
-
175
- **BibTeX:**
176
-
177
- [More Information Needed]
178
-
179
- **APA:**
180
-
181
- [More Information Needed]
182
-
183
- ## Glossary [optional]
184
-
185
- <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
186
-
187
- [More Information Needed]
188
-
189
- ## More Information [optional]
190
-
191
- [More Information Needed]
192
-
193
- ## Model Card Authors [optional]
194
-
195
- [More Information Needed]
196
-
197
- ## Model Card Contact
198
-
199
- [More Information Needed]
200
-
201
-
202
- ## Training procedure
203
-
204
-
205
- ### Framework versions
206
-
207
-
208
- - PEFT 0.6.2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
checkpoint-354/adapter_config.json DELETED
@@ -1,22 +0,0 @@
1
- {
2
- "alpha_pattern": {},
3
- "auto_mapping": null,
4
- "base_model_name_or_path": "baichuan-inc/Baichuan2-7B-Base",
5
- "bias": "none",
6
- "fan_in_fan_out": false,
7
- "inference_mode": true,
8
- "init_lora_weights": true,
9
- "layers_pattern": null,
10
- "layers_to_transform": null,
11
- "lora_alpha": 32,
12
- "lora_dropout": 0.1,
13
- "modules_to_save": null,
14
- "peft_type": "LORA",
15
- "r": 1,
16
- "rank_pattern": {},
17
- "revision": null,
18
- "target_modules": [
19
- "W_pack"
20
- ],
21
- "task_type": "CAUSAL_LM"
22
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
checkpoint-354/adapter_model.bin DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:da325f5cf970a401132441f46470dbc00920348b87ab0be221d1c7316322284c
3
- size 1070894
 
 
 
 
checkpoint-354/global_step354/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:4c11bab65f59c4832a75f5b58adb18de440ad320bd35e3c45ffae15c4375ee93
3
- size 1576368
 
 
 
 
checkpoint-354/global_step354/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:1deb7b48a7a3cc6a08adf4a2de757815a8ac8d3310164217b2c3b516f85a38fe
3
- size 1576368
 
 
 
 
checkpoint-354/global_step354/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:150dc2eb054ed35f09918866d318fd9192c4e48230257f8fa8c5b8064ef1aba8
3
- size 1576368
 
 
 
 
checkpoint-354/global_step354/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:81e23abba007da2f4c8e632f650ac2809b3e2c5009ba45721180edddd1b076b9
3
- size 1576368
 
 
 
 
checkpoint-354/global_step354/zero_pp_rank_0_mp_rank_00_model_states.pt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:6bed343db81975032808fdf09e85df567fa41afdd8ea68b65b4a6b9d7ebe82ba
3
- size 3753243804
 
 
 
 
checkpoint-354/global_step354/zero_pp_rank_1_mp_rank_00_model_states.pt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:db31ede9ee0adaac9a5105fc817b5618dab06c13cc1c09b0c2f846e72d651d01
3
- size 3753243804
 
 
 
 
checkpoint-354/global_step354/zero_pp_rank_2_mp_rank_00_model_states.pt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:4894083e6afab48f1b5ae4270b6f899c60d307848dfcf4834b3da634ce69ac02
3
- size 3753243804
 
 
 
 
checkpoint-354/global_step354/zero_pp_rank_3_mp_rank_00_model_states.pt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:14bbe275d438a67ed24aa7ecf86929caf351897ff2a6fbcdf62696685611c2eb
3
- size 3753243804
 
 
 
 
checkpoint-354/latest DELETED
@@ -1 +0,0 @@
1
- global_step354
 
 
checkpoint-354/rng_state_0.pth DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:caa1d6f86e7d0fe1fa14370edd5e28b0267f2e8054ba98023d05763b8e0af1aa
3
- size 15024
 
 
 
 
checkpoint-354/rng_state_1.pth DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:d37f31d4ebefb7468a8f770fcfa6fd49218aa64a6a42577fbc26df81f33d28fb
3
- size 15024
 
 
 
 
checkpoint-354/rng_state_2.pth DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:441ce8fea0362064858658b9d4735d1102e333e621c52cb8a66fc3ac175186aa
3
- size 15024
 
 
 
 
checkpoint-354/rng_state_3.pth DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:0d2afd98a129498d38110ee541a028bb643ee05bda8c9eab10f1f0491af6f650
3
- size 15024
 
 
 
 
checkpoint-354/special_tokens_map.json DELETED
@@ -1,30 +0,0 @@
1
- {
2
- "bos_token": {
3
- "content": "<s>",
4
- "lstrip": false,
5
- "normalized": true,
6
- "rstrip": false,
7
- "single_word": false
8
- },
9
- "eos_token": {
10
- "content": "</s>",
11
- "lstrip": false,
12
- "normalized": true,
13
- "rstrip": false,
14
- "single_word": false
15
- },
16
- "pad_token": {
17
- "content": "<unk>",
18
- "lstrip": false,
19
- "normalized": true,
20
- "rstrip": false,
21
- "single_word": false
22
- },
23
- "unk_token": {
24
- "content": "<unk>",
25
- "lstrip": false,
26
- "normalized": true,
27
- "rstrip": false,
28
- "single_word": false
29
- }
30
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
checkpoint-354/tokenizer.model DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:79452955be6b419a65984273a9f08af86042e1c2a75ee3ba989cbf620a133cc2
3
- size 2001107
 
 
 
 
checkpoint-354/tokenizer_config.json DELETED
@@ -1,47 +0,0 @@
1
- {
2
- "add_bos_token": false,
3
- "add_eos_token": false,
4
- "auto_map": {
5
- "AutoTokenizer": [
6
- "baichuan-inc/Baichuan2-7B-Base--tokenization_baichuan.BaichuanTokenizer",
7
- null
8
- ]
9
- },
10
- "bos_token": {
11
- "__type": "AddedToken",
12
- "content": "<s>",
13
- "lstrip": false,
14
- "normalized": true,
15
- "rstrip": false,
16
- "single_word": false
17
- },
18
- "clean_up_tokenization_spaces": false,
19
- "eos_token": {
20
- "__type": "AddedToken",
21
- "content": "</s>",
22
- "lstrip": false,
23
- "normalized": true,
24
- "rstrip": false,
25
- "single_word": true
26
- },
27
- "model_max_length": 1024,
28
- "pad_token": {
29
- "__type": "AddedToken",
30
- "content": "<unk>",
31
- "lstrip": false,
32
- "normalized": true,
33
- "rstrip": false,
34
- "single_word": true
35
- },
36
- "sp_model_kwargs": {},
37
- "tokenizer_class": "BaichuanTokenizer",
38
- "unk_token": {
39
- "__type": "AddedToken",
40
- "content": "<unk>",
41
- "lstrip": false,
42
- "normalized": true,
43
- "rstrip": false,
44
- "single_word": true
45
- },
46
- "use_fast": false
47
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
checkpoint-354/trainer_state.json DELETED
@@ -1,2143 +0,0 @@
1
- {
2
- "best_metric": null,
3
- "best_model_checkpoint": null,
4
- "epoch": 3.0,
5
- "eval_steps": 500,
6
- "global_step": 354,
7
- "is_hyper_param_search": false,
8
- "is_local_process_zero": true,
9
- "is_world_process_zero": true,
10
- "log_history": [
11
- {
12
- "epoch": 0.01,
13
- "learning_rate": 2e-05,
14
- "loss": 13.75,
15
- "step": 1
16
- },
17
- {
18
- "epoch": 0.02,
19
- "learning_rate": 2e-05,
20
- "loss": 13.375,
21
- "step": 2
22
- },
23
- {
24
- "epoch": 0.03,
25
- "learning_rate": 2e-05,
26
- "loss": 10.5938,
27
- "step": 3
28
- },
29
- {
30
- "epoch": 0.03,
31
- "learning_rate": 2e-05,
32
- "loss": 14.7656,
33
- "step": 4
34
- },
35
- {
36
- "epoch": 0.04,
37
- "learning_rate": 2e-05,
38
- "loss": 14.9062,
39
- "step": 5
40
- },
41
- {
42
- "epoch": 0.05,
43
- "learning_rate": 2e-05,
44
- "loss": 11.7109,
45
- "step": 6
46
- },
47
- {
48
- "epoch": 0.06,
49
- "learning_rate": 2e-05,
50
- "loss": 13.8438,
51
- "step": 7
52
- },
53
- {
54
- "epoch": 0.07,
55
- "learning_rate": 2e-05,
56
- "loss": 15.375,
57
- "step": 8
58
- },
59
- {
60
- "epoch": 0.08,
61
- "learning_rate": 2e-05,
62
- "loss": 13.25,
63
- "step": 9
64
- },
65
- {
66
- "epoch": 0.08,
67
- "learning_rate": 2e-05,
68
- "loss": 11.7578,
69
- "step": 10
70
- },
71
- {
72
- "epoch": 0.09,
73
- "learning_rate": 2e-05,
74
- "loss": 13.9219,
75
- "step": 11
76
- },
77
- {
78
- "epoch": 0.1,
79
- "learning_rate": 2e-05,
80
- "loss": 12.0859,
81
- "step": 12
82
- },
83
- {
84
- "epoch": 0.11,
85
- "learning_rate": 2e-05,
86
- "loss": 11.9688,
87
- "step": 13
88
- },
89
- {
90
- "epoch": 0.12,
91
- "learning_rate": 2e-05,
92
- "loss": 14.1875,
93
- "step": 14
94
- },
95
- {
96
- "epoch": 0.13,
97
- "learning_rate": 2e-05,
98
- "loss": 11.3594,
99
- "step": 15
100
- },
101
- {
102
- "epoch": 0.14,
103
- "learning_rate": 2e-05,
104
- "loss": 9.6641,
105
- "step": 16
106
- },
107
- {
108
- "epoch": 0.14,
109
- "learning_rate": 2e-05,
110
- "loss": 15.0625,
111
- "step": 17
112
- },
113
- {
114
- "epoch": 0.15,
115
- "learning_rate": 2e-05,
116
- "loss": 11.7031,
117
- "step": 18
118
- },
119
- {
120
- "epoch": 0.16,
121
- "learning_rate": 2e-05,
122
- "loss": 16.4219,
123
- "step": 19
124
- },
125
- {
126
- "epoch": 0.17,
127
- "learning_rate": 2e-05,
128
- "loss": 10.0469,
129
- "step": 20
130
- },
131
- {
132
- "epoch": 0.18,
133
- "learning_rate": 2e-05,
134
- "loss": 12.0,
135
- "step": 21
136
- },
137
- {
138
- "epoch": 0.19,
139
- "learning_rate": 2e-05,
140
- "loss": 13.4219,
141
- "step": 22
142
- },
143
- {
144
- "epoch": 0.19,
145
- "learning_rate": 2e-05,
146
- "loss": 13.0625,
147
- "step": 23
148
- },
149
- {
150
- "epoch": 0.2,
151
- "learning_rate": 2e-05,
152
- "loss": 10.9922,
153
- "step": 24
154
- },
155
- {
156
- "epoch": 0.21,
157
- "learning_rate": 2e-05,
158
- "loss": 12.3906,
159
- "step": 25
160
- },
161
- {
162
- "epoch": 0.22,
163
- "learning_rate": 2e-05,
164
- "loss": 9.3047,
165
- "step": 26
166
- },
167
- {
168
- "epoch": 0.23,
169
- "learning_rate": 2e-05,
170
- "loss": 9.6406,
171
- "step": 27
172
- },
173
- {
174
- "epoch": 0.24,
175
- "learning_rate": 2e-05,
176
- "loss": 13.4844,
177
- "step": 28
178
- },
179
- {
180
- "epoch": 0.25,
181
- "learning_rate": 2e-05,
182
- "loss": 13.0938,
183
- "step": 29
184
- },
185
- {
186
- "epoch": 0.25,
187
- "learning_rate": 2e-05,
188
- "loss": 12.7812,
189
- "step": 30
190
- },
191
- {
192
- "epoch": 0.26,
193
- "learning_rate": 2e-05,
194
- "loss": 12.0625,
195
- "step": 31
196
- },
197
- {
198
- "epoch": 0.27,
199
- "learning_rate": 2e-05,
200
- "loss": 12.4219,
201
- "step": 32
202
- },
203
- {
204
- "epoch": 0.28,
205
- "learning_rate": 2e-05,
206
- "loss": 11.3906,
207
- "step": 33
208
- },
209
- {
210
- "epoch": 0.29,
211
- "learning_rate": 2e-05,
212
- "loss": 10.6406,
213
- "step": 34
214
- },
215
- {
216
- "epoch": 0.3,
217
- "learning_rate": 2e-05,
218
- "loss": 10.4219,
219
- "step": 35
220
- },
221
- {
222
- "epoch": 0.31,
223
- "learning_rate": 2e-05,
224
- "loss": 12.0,
225
- "step": 36
226
- },
227
- {
228
- "epoch": 0.31,
229
- "learning_rate": 2e-05,
230
- "loss": 12.2031,
231
- "step": 37
232
- },
233
- {
234
- "epoch": 0.32,
235
- "learning_rate": 2e-05,
236
- "loss": 10.4297,
237
- "step": 38
238
- },
239
- {
240
- "epoch": 0.33,
241
- "learning_rate": 2e-05,
242
- "loss": 13.9844,
243
- "step": 39
244
- },
245
- {
246
- "epoch": 0.34,
247
- "learning_rate": 2e-05,
248
- "loss": 9.4062,
249
- "step": 40
250
- },
251
- {
252
- "epoch": 0.35,
253
- "learning_rate": 2e-05,
254
- "loss": 13.5781,
255
- "step": 41
256
- },
257
- {
258
- "epoch": 0.36,
259
- "learning_rate": 2e-05,
260
- "loss": 8.4688,
261
- "step": 42
262
- },
263
- {
264
- "epoch": 0.36,
265
- "learning_rate": 2e-05,
266
- "loss": 10.7656,
267
- "step": 43
268
- },
269
- {
270
- "epoch": 0.37,
271
- "learning_rate": 2e-05,
272
- "loss": 11.1719,
273
- "step": 44
274
- },
275
- {
276
- "epoch": 0.38,
277
- "learning_rate": 2e-05,
278
- "loss": 10.7656,
279
- "step": 45
280
- },
281
- {
282
- "epoch": 0.39,
283
- "learning_rate": 2e-05,
284
- "loss": 10.7891,
285
- "step": 46
286
- },
287
- {
288
- "epoch": 0.4,
289
- "learning_rate": 2e-05,
290
- "loss": 11.8438,
291
- "step": 47
292
- },
293
- {
294
- "epoch": 0.41,
295
- "learning_rate": 2e-05,
296
- "loss": 9.7812,
297
- "step": 48
298
- },
299
- {
300
- "epoch": 0.42,
301
- "learning_rate": 2e-05,
302
- "loss": 10.2188,
303
- "step": 49
304
- },
305
- {
306
- "epoch": 0.42,
307
- "learning_rate": 2e-05,
308
- "loss": 9.6719,
309
- "step": 50
310
- },
311
- {
312
- "epoch": 0.43,
313
- "learning_rate": 2e-05,
314
- "loss": 11.6719,
315
- "step": 51
316
- },
317
- {
318
- "epoch": 0.44,
319
- "learning_rate": 2e-05,
320
- "loss": 10.4375,
321
- "step": 52
322
- },
323
- {
324
- "epoch": 0.45,
325
- "learning_rate": 2e-05,
326
- "loss": 9.1406,
327
- "step": 53
328
- },
329
- {
330
- "epoch": 0.46,
331
- "learning_rate": 2e-05,
332
- "loss": 10.5625,
333
- "step": 54
334
- },
335
- {
336
- "epoch": 0.47,
337
- "learning_rate": 2e-05,
338
- "loss": 9.9688,
339
- "step": 55
340
- },
341
- {
342
- "epoch": 0.47,
343
- "learning_rate": 2e-05,
344
- "loss": 8.7891,
345
- "step": 56
346
- },
347
- {
348
- "epoch": 0.48,
349
- "learning_rate": 2e-05,
350
- "loss": 9.3125,
351
- "step": 57
352
- },
353
- {
354
- "epoch": 0.49,
355
- "learning_rate": 2e-05,
356
- "loss": 9.1797,
357
- "step": 58
358
- },
359
- {
360
- "epoch": 0.5,
361
- "learning_rate": 2e-05,
362
- "loss": 9.7812,
363
- "step": 59
364
- },
365
- {
366
- "epoch": 0.51,
367
- "learning_rate": 2e-05,
368
- "loss": 9.6094,
369
- "step": 60
370
- },
371
- {
372
- "epoch": 0.52,
373
- "learning_rate": 2e-05,
374
- "loss": 12.3281,
375
- "step": 61
376
- },
377
- {
378
- "epoch": 0.53,
379
- "learning_rate": 2e-05,
380
- "loss": 7.8828,
381
- "step": 62
382
- },
383
- {
384
- "epoch": 0.53,
385
- "learning_rate": 2e-05,
386
- "loss": 7.9844,
387
- "step": 63
388
- },
389
- {
390
- "epoch": 0.54,
391
- "learning_rate": 2e-05,
392
- "loss": 7.7578,
393
- "step": 64
394
- },
395
- {
396
- "epoch": 0.55,
397
- "learning_rate": 2e-05,
398
- "loss": 8.3906,
399
- "step": 65
400
- },
401
- {
402
- "epoch": 0.56,
403
- "learning_rate": 2e-05,
404
- "loss": 10.6328,
405
- "step": 66
406
- },
407
- {
408
- "epoch": 0.57,
409
- "learning_rate": 2e-05,
410
- "loss": 8.1562,
411
- "step": 67
412
- },
413
- {
414
- "epoch": 0.58,
415
- "learning_rate": 2e-05,
416
- "loss": 7.4688,
417
- "step": 68
418
- },
419
- {
420
- "epoch": 0.58,
421
- "learning_rate": 2e-05,
422
- "loss": 7.0078,
423
- "step": 69
424
- },
425
- {
426
- "epoch": 0.59,
427
- "learning_rate": 2e-05,
428
- "loss": 6.1641,
429
- "step": 70
430
- },
431
- {
432
- "epoch": 0.6,
433
- "learning_rate": 2e-05,
434
- "loss": 7.5703,
435
- "step": 71
436
- },
437
- {
438
- "epoch": 0.61,
439
- "learning_rate": 2e-05,
440
- "loss": 6.2109,
441
- "step": 72
442
- },
443
- {
444
- "epoch": 0.62,
445
- "learning_rate": 2e-05,
446
- "loss": 7.7656,
447
- "step": 73
448
- },
449
- {
450
- "epoch": 0.63,
451
- "learning_rate": 2e-05,
452
- "loss": 7.2266,
453
- "step": 74
454
- },
455
- {
456
- "epoch": 0.64,
457
- "learning_rate": 2e-05,
458
- "loss": 7.7266,
459
- "step": 75
460
- },
461
- {
462
- "epoch": 0.64,
463
- "learning_rate": 2e-05,
464
- "loss": 7.5391,
465
- "step": 76
466
- },
467
- {
468
- "epoch": 0.65,
469
- "learning_rate": 2e-05,
470
- "loss": 6.2031,
471
- "step": 77
472
- },
473
- {
474
- "epoch": 0.66,
475
- "learning_rate": 2e-05,
476
- "loss": 7.4922,
477
- "step": 78
478
- },
479
- {
480
- "epoch": 0.67,
481
- "learning_rate": 2e-05,
482
- "loss": 5.9062,
483
- "step": 79
484
- },
485
- {
486
- "epoch": 0.68,
487
- "learning_rate": 2e-05,
488
- "loss": 8.5469,
489
- "step": 80
490
- },
491
- {
492
- "epoch": 0.69,
493
- "learning_rate": 2e-05,
494
- "loss": 7.4844,
495
- "step": 81
496
- },
497
- {
498
- "epoch": 0.69,
499
- "learning_rate": 2e-05,
500
- "loss": 8.4297,
501
- "step": 82
502
- },
503
- {
504
- "epoch": 0.7,
505
- "learning_rate": 2e-05,
506
- "loss": 7.1641,
507
- "step": 83
508
- },
509
- {
510
- "epoch": 0.71,
511
- "learning_rate": 2e-05,
512
- "loss": 5.9844,
513
- "step": 84
514
- },
515
- {
516
- "epoch": 0.72,
517
- "learning_rate": 2e-05,
518
- "loss": 6.5078,
519
- "step": 85
520
- },
521
- {
522
- "epoch": 0.73,
523
- "learning_rate": 2e-05,
524
- "loss": 6.5312,
525
- "step": 86
526
- },
527
- {
528
- "epoch": 0.74,
529
- "learning_rate": 2e-05,
530
- "loss": 6.8984,
531
- "step": 87
532
- },
533
- {
534
- "epoch": 0.75,
535
- "learning_rate": 2e-05,
536
- "loss": 6.7188,
537
- "step": 88
538
- },
539
- {
540
- "epoch": 0.75,
541
- "learning_rate": 2e-05,
542
- "loss": 6.4453,
543
- "step": 89
544
- },
545
- {
546
- "epoch": 0.76,
547
- "learning_rate": 2e-05,
548
- "loss": 5.6562,
549
- "step": 90
550
- },
551
- {
552
- "epoch": 0.77,
553
- "learning_rate": 2e-05,
554
- "loss": 6.4766,
555
- "step": 91
556
- },
557
- {
558
- "epoch": 0.78,
559
- "learning_rate": 2e-05,
560
- "loss": 6.2031,
561
- "step": 92
562
- },
563
- {
564
- "epoch": 0.79,
565
- "learning_rate": 2e-05,
566
- "loss": 5.9531,
567
- "step": 93
568
- },
569
- {
570
- "epoch": 0.8,
571
- "learning_rate": 2e-05,
572
- "loss": 5.9453,
573
- "step": 94
574
- },
575
- {
576
- "epoch": 0.81,
577
- "learning_rate": 2e-05,
578
- "loss": 5.4297,
579
- "step": 95
580
- },
581
- {
582
- "epoch": 0.81,
583
- "learning_rate": 2e-05,
584
- "loss": 6.3125,
585
- "step": 96
586
- },
587
- {
588
- "epoch": 0.82,
589
- "learning_rate": 2e-05,
590
- "loss": 6.0703,
591
- "step": 97
592
- },
593
- {
594
- "epoch": 0.83,
595
- "learning_rate": 2e-05,
596
- "loss": 5.7891,
597
- "step": 98
598
- },
599
- {
600
- "epoch": 0.84,
601
- "learning_rate": 2e-05,
602
- "loss": 5.7969,
603
- "step": 99
604
- },
605
- {
606
- "epoch": 0.85,
607
- "learning_rate": 2e-05,
608
- "loss": 7.875,
609
- "step": 100
610
- },
611
- {
612
- "epoch": 0.86,
613
- "learning_rate": 2e-05,
614
- "loss": 6.25,
615
- "step": 101
616
- },
617
- {
618
- "epoch": 0.86,
619
- "learning_rate": 2e-05,
620
- "loss": 5.6484,
621
- "step": 102
622
- },
623
- {
624
- "epoch": 0.87,
625
- "learning_rate": 2e-05,
626
- "loss": 8.4609,
627
- "step": 103
628
- },
629
- {
630
- "epoch": 0.88,
631
- "learning_rate": 2e-05,
632
- "loss": 6.0156,
633
- "step": 104
634
- },
635
- {
636
- "epoch": 0.89,
637
- "learning_rate": 2e-05,
638
- "loss": 6.1172,
639
- "step": 105
640
- },
641
- {
642
- "epoch": 0.9,
643
- "learning_rate": 2e-05,
644
- "loss": 6.0156,
645
- "step": 106
646
- },
647
- {
648
- "epoch": 0.91,
649
- "learning_rate": 2e-05,
650
- "loss": 6.1562,
651
- "step": 107
652
- },
653
- {
654
- "epoch": 0.92,
655
- "learning_rate": 2e-05,
656
- "loss": 5.457,
657
- "step": 108
658
- },
659
- {
660
- "epoch": 0.92,
661
- "learning_rate": 2e-05,
662
- "loss": 5.6953,
663
- "step": 109
664
- },
665
- {
666
- "epoch": 0.93,
667
- "learning_rate": 2e-05,
668
- "loss": 6.3828,
669
- "step": 110
670
- },
671
- {
672
- "epoch": 0.94,
673
- "learning_rate": 2e-05,
674
- "loss": 4.7812,
675
- "step": 111
676
- },
677
- {
678
- "epoch": 0.95,
679
- "learning_rate": 2e-05,
680
- "loss": 7.5156,
681
- "step": 112
682
- },
683
- {
684
- "epoch": 0.96,
685
- "learning_rate": 2e-05,
686
- "loss": 5.7109,
687
- "step": 113
688
- },
689
- {
690
- "epoch": 0.97,
691
- "learning_rate": 2e-05,
692
- "loss": 6.1797,
693
- "step": 114
694
- },
695
- {
696
- "epoch": 0.97,
697
- "learning_rate": 2e-05,
698
- "loss": 7.5078,
699
- "step": 115
700
- },
701
- {
702
- "epoch": 0.98,
703
- "learning_rate": 2e-05,
704
- "loss": 6.7891,
705
- "step": 116
706
- },
707
- {
708
- "epoch": 0.99,
709
- "learning_rate": 2e-05,
710
- "loss": 7.0156,
711
- "step": 117
712
- },
713
- {
714
- "epoch": 1.0,
715
- "learning_rate": 2e-05,
716
- "loss": 5.8438,
717
- "step": 118
718
- },
719
- {
720
- "epoch": 1.01,
721
- "learning_rate": 2e-05,
722
- "loss": 5.4297,
723
- "step": 119
724
- },
725
- {
726
- "epoch": 1.02,
727
- "learning_rate": 2e-05,
728
- "loss": 5.5312,
729
- "step": 120
730
- },
731
- {
732
- "epoch": 1.03,
733
- "learning_rate": 2e-05,
734
- "loss": 5.0156,
735
- "step": 121
736
- },
737
- {
738
- "epoch": 1.03,
739
- "learning_rate": 2e-05,
740
- "loss": 5.8359,
741
- "step": 122
742
- },
743
- {
744
- "epoch": 1.04,
745
- "learning_rate": 2e-05,
746
- "loss": 5.75,
747
- "step": 123
748
- },
749
- {
750
- "epoch": 1.05,
751
- "learning_rate": 2e-05,
752
- "loss": 5.0273,
753
- "step": 124
754
- },
755
- {
756
- "epoch": 1.06,
757
- "learning_rate": 2e-05,
758
- "loss": 5.4688,
759
- "step": 125
760
- },
761
- {
762
- "epoch": 1.07,
763
- "learning_rate": 2e-05,
764
- "loss": 5.8906,
765
- "step": 126
766
- },
767
- {
768
- "epoch": 1.08,
769
- "learning_rate": 2e-05,
770
- "loss": 5.4609,
771
- "step": 127
772
- },
773
- {
774
- "epoch": 1.08,
775
- "learning_rate": 2e-05,
776
- "loss": 5.0508,
777
- "step": 128
778
- },
779
- {
780
- "epoch": 1.09,
781
- "learning_rate": 2e-05,
782
- "loss": 5.5547,
783
- "step": 129
784
- },
785
- {
786
- "epoch": 1.1,
787
- "learning_rate": 2e-05,
788
- "loss": 6.3984,
789
- "step": 130
790
- },
791
- {
792
- "epoch": 1.11,
793
- "learning_rate": 2e-05,
794
- "loss": 5.2578,
795
- "step": 131
796
- },
797
- {
798
- "epoch": 1.12,
799
- "learning_rate": 2e-05,
800
- "loss": 5.6562,
801
- "step": 132
802
- },
803
- {
804
- "epoch": 1.13,
805
- "learning_rate": 2e-05,
806
- "loss": 5.0664,
807
- "step": 133
808
- },
809
- {
810
- "epoch": 1.14,
811
- "learning_rate": 2e-05,
812
- "loss": 4.7344,
813
- "step": 134
814
- },
815
- {
816
- "epoch": 1.14,
817
- "learning_rate": 2e-05,
818
- "loss": 5.9219,
819
- "step": 135
820
- },
821
- {
822
- "epoch": 1.15,
823
- "learning_rate": 2e-05,
824
- "loss": 5.0469,
825
- "step": 136
826
- },
827
- {
828
- "epoch": 1.16,
829
- "learning_rate": 2e-05,
830
- "loss": 8.0,
831
- "step": 137
832
- },
833
- {
834
- "epoch": 1.17,
835
- "learning_rate": 2e-05,
836
- "loss": 4.8438,
837
- "step": 138
838
- },
839
- {
840
- "epoch": 1.18,
841
- "learning_rate": 2e-05,
842
- "loss": 5.25,
843
- "step": 139
844
- },
845
- {
846
- "epoch": 1.19,
847
- "learning_rate": 2e-05,
848
- "loss": 5.4141,
849
- "step": 140
850
- },
851
- {
852
- "epoch": 1.19,
853
- "learning_rate": 2e-05,
854
- "loss": 5.4375,
855
- "step": 141
856
- },
857
- {
858
- "epoch": 1.2,
859
- "learning_rate": 2e-05,
860
- "loss": 4.875,
861
- "step": 142
862
- },
863
- {
864
- "epoch": 1.21,
865
- "learning_rate": 2e-05,
866
- "loss": 5.4375,
867
- "step": 143
868
- },
869
- {
870
- "epoch": 1.22,
871
- "learning_rate": 2e-05,
872
- "loss": 4.5273,
873
- "step": 144
874
- },
875
- {
876
- "epoch": 1.23,
877
- "learning_rate": 2e-05,
878
- "loss": 4.6797,
879
- "step": 145
880
- },
881
- {
882
- "epoch": 1.24,
883
- "learning_rate": 2e-05,
884
- "loss": 5.7344,
885
- "step": 146
886
- },
887
- {
888
- "epoch": 1.25,
889
- "learning_rate": 2e-05,
890
- "loss": 5.6953,
891
- "step": 147
892
- },
893
- {
894
- "epoch": 1.25,
895
- "learning_rate": 2e-05,
896
- "loss": 5.5625,
897
- "step": 148
898
- },
899
- {
900
- "epoch": 1.26,
901
- "learning_rate": 2e-05,
902
- "loss": 5.3281,
903
- "step": 149
904
- },
905
- {
906
- "epoch": 1.27,
907
- "learning_rate": 2e-05,
908
- "loss": 5.3828,
909
- "step": 150
910
- },
911
- {
912
- "epoch": 1.28,
913
- "learning_rate": 2e-05,
914
- "loss": 5.2031,
915
- "step": 151
916
- },
917
- {
918
- "epoch": 1.29,
919
- "learning_rate": 2e-05,
920
- "loss": 4.9688,
921
- "step": 152
922
- },
923
- {
924
- "epoch": 1.3,
925
- "learning_rate": 2e-05,
926
- "loss": 5.0391,
927
- "step": 153
928
- },
929
- {
930
- "epoch": 1.31,
931
- "learning_rate": 2e-05,
932
- "loss": 5.5312,
933
- "step": 154
934
- },
935
- {
936
- "epoch": 1.31,
937
- "learning_rate": 2e-05,
938
- "loss": 5.5781,
939
- "step": 155
940
- },
941
- {
942
- "epoch": 1.32,
943
- "learning_rate": 2e-05,
944
- "loss": 4.8242,
945
- "step": 156
946
- },
947
- {
948
- "epoch": 1.33,
949
- "learning_rate": 2e-05,
950
- "loss": 6.0703,
951
- "step": 157
952
- },
953
- {
954
- "epoch": 1.34,
955
- "learning_rate": 2e-05,
956
- "loss": 4.7422,
957
- "step": 158
958
- },
959
- {
960
- "epoch": 1.35,
961
- "learning_rate": 2e-05,
962
- "loss": 6.5469,
963
- "step": 159
964
- },
965
- {
966
- "epoch": 1.36,
967
- "learning_rate": 2e-05,
968
- "loss": 4.5469,
969
- "step": 160
970
- },
971
- {
972
- "epoch": 1.36,
973
- "learning_rate": 2e-05,
974
- "loss": 5.375,
975
- "step": 161
976
- },
977
- {
978
- "epoch": 1.37,
979
- "learning_rate": 2e-05,
980
- "loss": 5.4922,
981
- "step": 162
982
- },
983
- {
984
- "epoch": 1.38,
985
- "learning_rate": 2e-05,
986
- "loss": 5.5547,
987
- "step": 163
988
- },
989
- {
990
- "epoch": 1.39,
991
- "learning_rate": 2e-05,
992
- "loss": 5.4375,
993
- "step": 164
994
- },
995
- {
996
- "epoch": 1.4,
997
- "learning_rate": 2e-05,
998
- "loss": 6.3984,
999
- "step": 165
1000
- },
1001
- {
1002
- "epoch": 1.41,
1003
- "learning_rate": 2e-05,
1004
- "loss": 5.2734,
1005
- "step": 166
1006
- },
1007
- {
1008
- "epoch": 1.42,
1009
- "learning_rate": 2e-05,
1010
- "loss": 5.3047,
1011
- "step": 167
1012
- },
1013
- {
1014
- "epoch": 1.42,
1015
- "learning_rate": 2e-05,
1016
- "loss": 4.9531,
1017
- "step": 168
1018
- },
1019
- {
1020
- "epoch": 1.43,
1021
- "learning_rate": 2e-05,
1022
- "loss": 6.8906,
1023
- "step": 169
1024
- },
1025
- {
1026
- "epoch": 1.44,
1027
- "learning_rate": 2e-05,
1028
- "loss": 6.5039,
1029
- "step": 170
1030
- },
1031
- {
1032
- "epoch": 1.45,
1033
- "learning_rate": 2e-05,
1034
- "loss": 5.1172,
1035
- "step": 171
1036
- },
1037
- {
1038
- "epoch": 1.46,
1039
- "learning_rate": 2e-05,
1040
- "loss": 5.6797,
1041
- "step": 172
1042
- },
1043
- {
1044
- "epoch": 1.47,
1045
- "learning_rate": 2e-05,
1046
- "loss": 5.5508,
1047
- "step": 173
1048
- },
1049
- {
1050
- "epoch": 1.47,
1051
- "learning_rate": 2e-05,
1052
- "loss": 5.2656,
1053
- "step": 174
1054
- },
1055
- {
1056
- "epoch": 1.48,
1057
- "learning_rate": 2e-05,
1058
- "loss": 5.0508,
1059
- "step": 175
1060
- },
1061
- {
1062
- "epoch": 1.49,
1063
- "learning_rate": 2e-05,
1064
- "loss": 5.2266,
1065
- "step": 176
1066
- },
1067
- {
1068
- "epoch": 1.5,
1069
- "learning_rate": 2e-05,
1070
- "loss": 5.3711,
1071
- "step": 177
1072
- },
1073
- {
1074
- "epoch": 1.51,
1075
- "learning_rate": 2e-05,
1076
- "loss": 5.0898,
1077
- "step": 178
1078
- },
1079
- {
1080
- "epoch": 1.52,
1081
- "learning_rate": 2e-05,
1082
- "loss": 6.4219,
1083
- "step": 179
1084
- },
1085
- {
1086
- "epoch": 1.53,
1087
- "learning_rate": 2e-05,
1088
- "loss": 5.0234,
1089
- "step": 180
1090
- },
1091
- {
1092
- "epoch": 1.53,
1093
- "learning_rate": 2e-05,
1094
- "loss": 5.0625,
1095
- "step": 181
1096
- },
1097
- {
1098
- "epoch": 1.54,
1099
- "learning_rate": 2e-05,
1100
- "loss": 5.1797,
1101
- "step": 182
1102
- },
1103
- {
1104
- "epoch": 1.55,
1105
- "learning_rate": 2e-05,
1106
- "loss": 5.2812,
1107
- "step": 183
1108
- },
1109
- {
1110
- "epoch": 1.56,
1111
- "learning_rate": 2e-05,
1112
- "loss": 5.5547,
1113
- "step": 184
1114
- },
1115
- {
1116
- "epoch": 1.57,
1117
- "learning_rate": 2e-05,
1118
- "loss": 4.8281,
1119
- "step": 185
1120
- },
1121
- {
1122
- "epoch": 1.58,
1123
- "learning_rate": 2e-05,
1124
- "loss": 4.8867,
1125
- "step": 186
1126
- },
1127
- {
1128
- "epoch": 1.58,
1129
- "learning_rate": 2e-05,
1130
- "loss": 4.7852,
1131
- "step": 187
1132
- },
1133
- {
1134
- "epoch": 1.59,
1135
- "learning_rate": 2e-05,
1136
- "loss": 4.4297,
1137
- "step": 188
1138
- },
1139
- {
1140
- "epoch": 1.6,
1141
- "learning_rate": 2e-05,
1142
- "loss": 5.4688,
1143
- "step": 189
1144
- },
1145
- {
1146
- "epoch": 1.61,
1147
- "learning_rate": 2e-05,
1148
- "loss": 4.668,
1149
- "step": 190
1150
- },
1151
- {
1152
- "epoch": 1.62,
1153
- "learning_rate": 2e-05,
1154
- "loss": 5.4062,
1155
- "step": 191
1156
- },
1157
- {
1158
- "epoch": 1.63,
1159
- "learning_rate": 2e-05,
1160
- "loss": 4.8828,
1161
- "step": 192
1162
- },
1163
- {
1164
- "epoch": 1.64,
1165
- "learning_rate": 2e-05,
1166
- "loss": 5.6953,
1167
- "step": 193
1168
- },
1169
- {
1170
- "epoch": 1.64,
1171
- "learning_rate": 2e-05,
1172
- "loss": 5.6094,
1173
- "step": 194
1174
- },
1175
- {
1176
- "epoch": 1.65,
1177
- "learning_rate": 2e-05,
1178
- "loss": 4.8516,
1179
- "step": 195
1180
- },
1181
- {
1182
- "epoch": 1.66,
1183
- "learning_rate": 2e-05,
1184
- "loss": 5.3359,
1185
- "step": 196
1186
- },
1187
- {
1188
- "epoch": 1.67,
1189
- "learning_rate": 2e-05,
1190
- "loss": 4.6953,
1191
- "step": 197
1192
- },
1193
- {
1194
- "epoch": 1.68,
1195
- "learning_rate": 2e-05,
1196
- "loss": 5.5859,
1197
- "step": 198
1198
- },
1199
- {
1200
- "epoch": 1.69,
1201
- "learning_rate": 2e-05,
1202
- "loss": 5.1016,
1203
- "step": 199
1204
- },
1205
- {
1206
- "epoch": 1.69,
1207
- "learning_rate": 2e-05,
1208
- "loss": 6.8984,
1209
- "step": 200
1210
- },
1211
- {
1212
- "epoch": 1.7,
1213
- "learning_rate": 2e-05,
1214
- "loss": 5.1055,
1215
- "step": 201
1216
- },
1217
- {
1218
- "epoch": 1.71,
1219
- "learning_rate": 2e-05,
1220
- "loss": 4.8164,
1221
- "step": 202
1222
- },
1223
- {
1224
- "epoch": 1.72,
1225
- "learning_rate": 2e-05,
1226
- "loss": 5.2969,
1227
- "step": 203
1228
- },
1229
- {
1230
- "epoch": 1.73,
1231
- "learning_rate": 2e-05,
1232
- "loss": 5.4297,
1233
- "step": 204
1234
- },
1235
- {
1236
- "epoch": 1.74,
1237
- "learning_rate": 2e-05,
1238
- "loss": 4.8711,
1239
- "step": 205
1240
- },
1241
- {
1242
- "epoch": 1.75,
1243
- "learning_rate": 2e-05,
1244
- "loss": 5.0742,
1245
- "step": 206
1246
- },
1247
- {
1248
- "epoch": 1.75,
1249
- "learning_rate": 2e-05,
1250
- "loss": 5.3672,
1251
- "step": 207
1252
- },
1253
- {
1254
- "epoch": 1.76,
1255
- "learning_rate": 2e-05,
1256
- "loss": 4.4961,
1257
- "step": 208
1258
- },
1259
- {
1260
- "epoch": 1.77,
1261
- "learning_rate": 2e-05,
1262
- "loss": 5.4453,
1263
- "step": 209
1264
- },
1265
- {
1266
- "epoch": 1.78,
1267
- "learning_rate": 2e-05,
1268
- "loss": 5.0547,
1269
- "step": 210
1270
- },
1271
- {
1272
- "epoch": 1.79,
1273
- "learning_rate": 2e-05,
1274
- "loss": 5.1562,
1275
- "step": 211
1276
- },
1277
- {
1278
- "epoch": 1.8,
1279
- "learning_rate": 2e-05,
1280
- "loss": 4.9414,
1281
- "step": 212
1282
- },
1283
- {
1284
- "epoch": 1.81,
1285
- "learning_rate": 2e-05,
1286
- "loss": 4.3828,
1287
- "step": 213
1288
- },
1289
- {
1290
- "epoch": 1.81,
1291
- "learning_rate": 2e-05,
1292
- "loss": 5.3828,
1293
- "step": 214
1294
- },
1295
- {
1296
- "epoch": 1.82,
1297
- "learning_rate": 2e-05,
1298
- "loss": 5.1953,
1299
- "step": 215
1300
- },
1301
- {
1302
- "epoch": 1.83,
1303
- "learning_rate": 2e-05,
1304
- "loss": 4.8984,
1305
- "step": 216
1306
- },
1307
- {
1308
- "epoch": 1.84,
1309
- "learning_rate": 2e-05,
1310
- "loss": 4.9297,
1311
- "step": 217
1312
- },
1313
- {
1314
- "epoch": 1.85,
1315
- "learning_rate": 2e-05,
1316
- "loss": 5.2695,
1317
- "step": 218
1318
- },
1319
- {
1320
- "epoch": 1.86,
1321
- "learning_rate": 2e-05,
1322
- "loss": 5.5625,
1323
- "step": 219
1324
- },
1325
- {
1326
- "epoch": 1.86,
1327
- "learning_rate": 2e-05,
1328
- "loss": 4.9102,
1329
- "step": 220
1330
- },
1331
- {
1332
- "epoch": 1.87,
1333
- "learning_rate": 2e-05,
1334
- "loss": 6.8047,
1335
- "step": 221
1336
- },
1337
- {
1338
- "epoch": 1.88,
1339
- "learning_rate": 2e-05,
1340
- "loss": 5.2422,
1341
- "step": 222
1342
- },
1343
- {
1344
- "epoch": 1.89,
1345
- "learning_rate": 2e-05,
1346
- "loss": 5.3672,
1347
- "step": 223
1348
- },
1349
- {
1350
- "epoch": 1.9,
1351
- "learning_rate": 2e-05,
1352
- "loss": 5.1719,
1353
- "step": 224
1354
- },
1355
- {
1356
- "epoch": 1.91,
1357
- "learning_rate": 2e-05,
1358
- "loss": 5.4141,
1359
- "step": 225
1360
- },
1361
- {
1362
- "epoch": 1.92,
1363
- "learning_rate": 2e-05,
1364
- "loss": 4.7305,
1365
- "step": 226
1366
- },
1367
- {
1368
- "epoch": 1.92,
1369
- "learning_rate": 2e-05,
1370
- "loss": 4.9609,
1371
- "step": 227
1372
- },
1373
- {
1374
- "epoch": 1.93,
1375
- "learning_rate": 2e-05,
1376
- "loss": 5.7422,
1377
- "step": 228
1378
- },
1379
- {
1380
- "epoch": 1.94,
1381
- "learning_rate": 2e-05,
1382
- "loss": 3.957,
1383
- "step": 229
1384
- },
1385
- {
1386
- "epoch": 1.95,
1387
- "learning_rate": 2e-05,
1388
- "loss": 5.7656,
1389
- "step": 230
1390
- },
1391
- {
1392
- "epoch": 1.96,
1393
- "learning_rate": 2e-05,
1394
- "loss": 4.9766,
1395
- "step": 231
1396
- },
1397
- {
1398
- "epoch": 1.97,
1399
- "learning_rate": 2e-05,
1400
- "loss": 5.5391,
1401
- "step": 232
1402
- },
1403
- {
1404
- "epoch": 1.97,
1405
- "learning_rate": 2e-05,
1406
- "loss": 6.2344,
1407
- "step": 233
1408
- },
1409
- {
1410
- "epoch": 1.98,
1411
- "learning_rate": 2e-05,
1412
- "loss": 5.4844,
1413
- "step": 234
1414
- },
1415
- {
1416
- "epoch": 1.99,
1417
- "learning_rate": 2e-05,
1418
- "loss": 5.3281,
1419
- "step": 235
1420
- },
1421
- {
1422
- "epoch": 2.0,
1423
- "learning_rate": 2e-05,
1424
- "loss": 5.25,
1425
- "step": 236
1426
- },
1427
- {
1428
- "epoch": 2.01,
1429
- "learning_rate": 2e-05,
1430
- "loss": 4.8867,
1431
- "step": 237
1432
- },
1433
- {
1434
- "epoch": 2.02,
1435
- "learning_rate": 2e-05,
1436
- "loss": 4.9688,
1437
- "step": 238
1438
- },
1439
- {
1440
- "epoch": 2.03,
1441
- "learning_rate": 2e-05,
1442
- "loss": 4.3711,
1443
- "step": 239
1444
- },
1445
- {
1446
- "epoch": 2.03,
1447
- "learning_rate": 2e-05,
1448
- "loss": 5.2891,
1449
- "step": 240
1450
- },
1451
- {
1452
- "epoch": 2.04,
1453
- "learning_rate": 2e-05,
1454
- "loss": 5.2578,
1455
- "step": 241
1456
- },
1457
- {
1458
- "epoch": 2.05,
1459
- "learning_rate": 2e-05,
1460
- "loss": 4.5,
1461
- "step": 242
1462
- },
1463
- {
1464
- "epoch": 2.06,
1465
- "learning_rate": 2e-05,
1466
- "loss": 4.9727,
1467
- "step": 243
1468
- },
1469
- {
1470
- "epoch": 2.07,
1471
- "learning_rate": 2e-05,
1472
- "loss": 5.4375,
1473
- "step": 244
1474
- },
1475
- {
1476
- "epoch": 2.08,
1477
- "learning_rate": 2e-05,
1478
- "loss": 5.0,
1479
- "step": 245
1480
- },
1481
- {
1482
- "epoch": 2.08,
1483
- "learning_rate": 2e-05,
1484
- "loss": 4.5664,
1485
- "step": 246
1486
- },
1487
- {
1488
- "epoch": 2.09,
1489
- "learning_rate": 2e-05,
1490
- "loss": 5.1484,
1491
- "step": 247
1492
- },
1493
- {
1494
- "epoch": 2.1,
1495
- "learning_rate": 2e-05,
1496
- "loss": 5.2109,
1497
- "step": 248
1498
- },
1499
- {
1500
- "epoch": 2.11,
1501
- "learning_rate": 2e-05,
1502
- "loss": 4.7969,
1503
- "step": 249
1504
- },
1505
- {
1506
- "epoch": 2.12,
1507
- "learning_rate": 2e-05,
1508
- "loss": 5.2109,
1509
- "step": 250
1510
- },
1511
- {
1512
- "epoch": 2.13,
1513
- "learning_rate": 2e-05,
1514
- "loss": 4.6172,
1515
- "step": 251
1516
- },
1517
- {
1518
- "epoch": 2.14,
1519
- "learning_rate": 2e-05,
1520
- "loss": 4.2891,
1521
- "step": 252
1522
- },
1523
- {
1524
- "epoch": 2.14,
1525
- "learning_rate": 2e-05,
1526
- "loss": 5.4531,
1527
- "step": 253
1528
- },
1529
- {
1530
- "epoch": 2.15,
1531
- "learning_rate": 2e-05,
1532
- "loss": 4.5898,
1533
- "step": 254
1534
- },
1535
- {
1536
- "epoch": 2.16,
1537
- "learning_rate": 2e-05,
1538
- "loss": 7.4609,
1539
- "step": 255
1540
- },
1541
- {
1542
- "epoch": 2.17,
1543
- "learning_rate": 2e-05,
1544
- "loss": 4.4453,
1545
- "step": 256
1546
- },
1547
- {
1548
- "epoch": 2.18,
1549
- "learning_rate": 2e-05,
1550
- "loss": 4.875,
1551
- "step": 257
1552
- },
1553
- {
1554
- "epoch": 2.19,
1555
- "learning_rate": 2e-05,
1556
- "loss": 5.0625,
1557
- "step": 258
1558
- },
1559
- {
1560
- "epoch": 2.19,
1561
- "learning_rate": 2e-05,
1562
- "loss": 5.0781,
1563
- "step": 259
1564
- },
1565
- {
1566
- "epoch": 2.2,
1567
- "learning_rate": 2e-05,
1568
- "loss": 4.5312,
1569
- "step": 260
1570
- },
1571
- {
1572
- "epoch": 2.21,
1573
- "learning_rate": 2e-05,
1574
- "loss": 5.0703,
1575
- "step": 261
1576
- },
1577
- {
1578
- "epoch": 2.22,
1579
- "learning_rate": 2e-05,
1580
- "loss": 4.1914,
1581
- "step": 262
1582
- },
1583
- {
1584
- "epoch": 2.23,
1585
- "learning_rate": 2e-05,
1586
- "loss": 4.3555,
1587
- "step": 263
1588
- },
1589
- {
1590
- "epoch": 2.24,
1591
- "learning_rate": 2e-05,
1592
- "loss": 5.4141,
1593
- "step": 264
1594
- },
1595
- {
1596
- "epoch": 2.25,
1597
- "learning_rate": 2e-05,
1598
- "loss": 5.3516,
1599
- "step": 265
1600
- },
1601
- {
1602
- "epoch": 2.25,
1603
- "learning_rate": 2e-05,
1604
- "loss": 5.2422,
1605
- "step": 266
1606
- },
1607
- {
1608
- "epoch": 2.26,
1609
- "learning_rate": 2e-05,
1610
- "loss": 5.0156,
1611
- "step": 267
1612
- },
1613
- {
1614
- "epoch": 2.27,
1615
- "learning_rate": 2e-05,
1616
- "loss": 5.0859,
1617
- "step": 268
1618
- },
1619
- {
1620
- "epoch": 2.28,
1621
- "learning_rate": 2e-05,
1622
- "loss": 4.8984,
1623
- "step": 269
1624
- },
1625
- {
1626
- "epoch": 2.29,
1627
- "learning_rate": 2e-05,
1628
- "loss": 4.6719,
1629
- "step": 270
1630
- },
1631
- {
1632
- "epoch": 2.3,
1633
- "learning_rate": 2e-05,
1634
- "loss": 4.7383,
1635
- "step": 271
1636
- },
1637
- {
1638
- "epoch": 2.31,
1639
- "learning_rate": 2e-05,
1640
- "loss": 5.25,
1641
- "step": 272
1642
- },
1643
- {
1644
- "epoch": 2.31,
1645
- "learning_rate": 2e-05,
1646
- "loss": 5.2891,
1647
- "step": 273
1648
- },
1649
- {
1650
- "epoch": 2.32,
1651
- "learning_rate": 2e-05,
1652
- "loss": 4.5508,
1653
- "step": 274
1654
- },
1655
- {
1656
- "epoch": 2.33,
1657
- "learning_rate": 2e-05,
1658
- "loss": 5.7422,
1659
- "step": 275
1660
- },
1661
- {
1662
- "epoch": 2.34,
1663
- "learning_rate": 2e-05,
1664
- "loss": 4.4688,
1665
- "step": 276
1666
- },
1667
- {
1668
- "epoch": 2.35,
1669
- "learning_rate": 2e-05,
1670
- "loss": 5.7969,
1671
- "step": 277
1672
- },
1673
- {
1674
- "epoch": 2.36,
1675
- "learning_rate": 2e-05,
1676
- "loss": 4.2852,
1677
- "step": 278
1678
- },
1679
- {
1680
- "epoch": 2.36,
1681
- "learning_rate": 2e-05,
1682
- "loss": 5.1172,
1683
- "step": 279
1684
- },
1685
- {
1686
- "epoch": 2.37,
1687
- "learning_rate": 2e-05,
1688
- "loss": 5.2344,
1689
- "step": 280
1690
- },
1691
- {
1692
- "epoch": 2.38,
1693
- "learning_rate": 2e-05,
1694
- "loss": 5.2891,
1695
- "step": 281
1696
- },
1697
- {
1698
- "epoch": 2.39,
1699
- "learning_rate": 2e-05,
1700
- "loss": 5.1641,
1701
- "step": 282
1702
- },
1703
- {
1704
- "epoch": 2.4,
1705
- "learning_rate": 2e-05,
1706
- "loss": 5.375,
1707
- "step": 283
1708
- },
1709
- {
1710
- "epoch": 2.41,
1711
- "learning_rate": 2e-05,
1712
- "loss": 5.0312,
1713
- "step": 284
1714
- },
1715
- {
1716
- "epoch": 2.42,
1717
- "learning_rate": 2e-05,
1718
- "loss": 5.0508,
1719
- "step": 285
1720
- },
1721
- {
1722
- "epoch": 2.42,
1723
- "learning_rate": 2e-05,
1724
- "loss": 4.6875,
1725
- "step": 286
1726
- },
1727
- {
1728
- "epoch": 2.43,
1729
- "learning_rate": 2e-05,
1730
- "loss": 6.0469,
1731
- "step": 287
1732
- },
1733
- {
1734
- "epoch": 2.44,
1735
- "learning_rate": 2e-05,
1736
- "loss": 5.1523,
1737
- "step": 288
1738
- },
1739
- {
1740
- "epoch": 2.45,
1741
- "learning_rate": 2e-05,
1742
- "loss": 4.875,
1743
- "step": 289
1744
- },
1745
- {
1746
- "epoch": 2.46,
1747
- "learning_rate": 2e-05,
1748
- "loss": 5.4102,
1749
- "step": 290
1750
- },
1751
- {
1752
- "epoch": 2.47,
1753
- "learning_rate": 2e-05,
1754
- "loss": 5.2891,
1755
- "step": 291
1756
- },
1757
- {
1758
- "epoch": 2.47,
1759
- "learning_rate": 2e-05,
1760
- "loss": 5.0234,
1761
- "step": 292
1762
- },
1763
- {
1764
- "epoch": 2.48,
1765
- "learning_rate": 2e-05,
1766
- "loss": 4.8086,
1767
- "step": 293
1768
- },
1769
- {
1770
- "epoch": 2.49,
1771
- "learning_rate": 2e-05,
1772
- "loss": 4.9766,
1773
- "step": 294
1774
- },
1775
- {
1776
- "epoch": 2.5,
1777
- "learning_rate": 2e-05,
1778
- "loss": 5.1211,
1779
- "step": 295
1780
- },
1781
- {
1782
- "epoch": 2.51,
1783
- "learning_rate": 2e-05,
1784
- "loss": 4.8594,
1785
- "step": 296
1786
- },
1787
- {
1788
- "epoch": 2.52,
1789
- "learning_rate": 2e-05,
1790
- "loss": 5.9375,
1791
- "step": 297
1792
- },
1793
- {
1794
- "epoch": 2.53,
1795
- "learning_rate": 2e-05,
1796
- "loss": 4.8281,
1797
- "step": 298
1798
- },
1799
- {
1800
- "epoch": 2.53,
1801
- "learning_rate": 2e-05,
1802
- "loss": 4.8594,
1803
- "step": 299
1804
- },
1805
- {
1806
- "epoch": 2.54,
1807
- "learning_rate": 2e-05,
1808
- "loss": 4.9688,
1809
- "step": 300
1810
- },
1811
- {
1812
- "epoch": 2.55,
1813
- "learning_rate": 2e-05,
1814
- "loss": 5.043,
1815
- "step": 301
1816
- },
1817
- {
1818
- "epoch": 2.56,
1819
- "learning_rate": 2e-05,
1820
- "loss": 5.2422,
1821
- "step": 302
1822
- },
1823
- {
1824
- "epoch": 2.57,
1825
- "learning_rate": 2e-05,
1826
- "loss": 4.5977,
1827
- "step": 303
1828
- },
1829
- {
1830
- "epoch": 2.58,
1831
- "learning_rate": 2e-05,
1832
- "loss": 4.6797,
1833
- "step": 304
1834
- },
1835
- {
1836
- "epoch": 2.58,
1837
- "learning_rate": 2e-05,
1838
- "loss": 4.5781,
1839
- "step": 305
1840
- },
1841
- {
1842
- "epoch": 2.59,
1843
- "learning_rate": 2e-05,
1844
- "loss": 4.2266,
1845
- "step": 306
1846
- },
1847
- {
1848
- "epoch": 2.6,
1849
- "learning_rate": 2e-05,
1850
- "loss": 5.2656,
1851
- "step": 307
1852
- },
1853
- {
1854
- "epoch": 2.61,
1855
- "learning_rate": 2e-05,
1856
- "loss": 4.4609,
1857
- "step": 308
1858
- },
1859
- {
1860
- "epoch": 2.62,
1861
- "learning_rate": 2e-05,
1862
- "loss": 5.1953,
1863
- "step": 309
1864
- },
1865
- {
1866
- "epoch": 2.63,
1867
- "learning_rate": 2e-05,
1868
- "loss": 4.6719,
1869
- "step": 310
1870
- },
1871
- {
1872
- "epoch": 2.64,
1873
- "learning_rate": 2e-05,
1874
- "loss": 5.4766,
1875
- "step": 311
1876
- },
1877
- {
1878
- "epoch": 2.64,
1879
- "learning_rate": 2e-05,
1880
- "loss": 5.4062,
1881
- "step": 312
1882
- },
1883
- {
1884
- "epoch": 2.65,
1885
- "learning_rate": 2e-05,
1886
- "loss": 4.6484,
1887
- "step": 313
1888
- },
1889
- {
1890
- "epoch": 2.66,
1891
- "learning_rate": 2e-05,
1892
- "loss": 5.1172,
1893
- "step": 314
1894
- },
1895
- {
1896
- "epoch": 2.67,
1897
- "learning_rate": 2e-05,
1898
- "loss": 4.5312,
1899
- "step": 315
1900
- },
1901
- {
1902
- "epoch": 2.68,
1903
- "learning_rate": 2e-05,
1904
- "loss": 5.3438,
1905
- "step": 316
1906
- },
1907
- {
1908
- "epoch": 2.69,
1909
- "learning_rate": 2e-05,
1910
- "loss": 4.8984,
1911
- "step": 317
1912
- },
1913
- {
1914
- "epoch": 2.69,
1915
- "learning_rate": 2e-05,
1916
- "loss": 6.6211,
1917
- "step": 318
1918
- },
1919
- {
1920
- "epoch": 2.7,
1921
- "learning_rate": 2e-05,
1922
- "loss": 4.8906,
1923
- "step": 319
1924
- },
1925
- {
1926
- "epoch": 2.71,
1927
- "learning_rate": 2e-05,
1928
- "loss": 4.6445,
1929
- "step": 320
1930
- },
1931
- {
1932
- "epoch": 2.72,
1933
- "learning_rate": 2e-05,
1934
- "loss": 5.0977,
1935
- "step": 321
1936
- },
1937
- {
1938
- "epoch": 2.73,
1939
- "learning_rate": 2e-05,
1940
- "loss": 5.2344,
1941
- "step": 322
1942
- },
1943
- {
1944
- "epoch": 2.74,
1945
- "learning_rate": 2e-05,
1946
- "loss": 4.6875,
1947
- "step": 323
1948
- },
1949
- {
1950
- "epoch": 2.75,
1951
- "learning_rate": 2e-05,
1952
- "loss": 4.8828,
1953
- "step": 324
1954
- },
1955
- {
1956
- "epoch": 2.75,
1957
- "learning_rate": 2e-05,
1958
- "loss": 5.1797,
1959
- "step": 325
1960
- },
1961
- {
1962
- "epoch": 2.76,
1963
- "learning_rate": 2e-05,
1964
- "loss": 4.3203,
1965
- "step": 326
1966
- },
1967
- {
1968
- "epoch": 2.77,
1969
- "learning_rate": 2e-05,
1970
- "loss": 5.2656,
1971
- "step": 327
1972
- },
1973
- {
1974
- "epoch": 2.78,
1975
- "learning_rate": 2e-05,
1976
- "loss": 4.875,
1977
- "step": 328
1978
- },
1979
- {
1980
- "epoch": 2.79,
1981
- "learning_rate": 2e-05,
1982
- "loss": 4.9922,
1983
- "step": 329
1984
- },
1985
- {
1986
- "epoch": 2.8,
1987
- "learning_rate": 2e-05,
1988
- "loss": 4.7852,
1989
- "step": 330
1990
- },
1991
- {
1992
- "epoch": 2.81,
1993
- "learning_rate": 2e-05,
1994
- "loss": 4.207,
1995
- "step": 331
1996
- },
1997
- {
1998
- "epoch": 2.81,
1999
- "learning_rate": 2e-05,
2000
- "loss": 5.1953,
2001
- "step": 332
2002
- },
2003
- {
2004
- "epoch": 2.82,
2005
- "learning_rate": 2e-05,
2006
- "loss": 5.0312,
2007
- "step": 333
2008
- },
2009
- {
2010
- "epoch": 2.83,
2011
- "learning_rate": 2e-05,
2012
- "loss": 4.7344,
2013
- "step": 334
2014
- },
2015
- {
2016
- "epoch": 2.84,
2017
- "learning_rate": 2e-05,
2018
- "loss": 4.7656,
2019
- "step": 335
2020
- },
2021
- {
2022
- "epoch": 2.85,
2023
- "learning_rate": 2e-05,
2024
- "loss": 5.0898,
2025
- "step": 336
2026
- },
2027
- {
2028
- "epoch": 2.86,
2029
- "learning_rate": 2e-05,
2030
- "loss": 5.3867,
2031
- "step": 337
2032
- },
2033
- {
2034
- "epoch": 2.86,
2035
- "learning_rate": 2e-05,
2036
- "loss": 4.7578,
2037
- "step": 338
2038
- },
2039
- {
2040
- "epoch": 2.87,
2041
- "learning_rate": 2e-05,
2042
- "loss": 6.0547,
2043
- "step": 339
2044
- },
2045
- {
2046
- "epoch": 2.88,
2047
- "learning_rate": 2e-05,
2048
- "loss": 5.0781,
2049
- "step": 340
2050
- },
2051
- {
2052
- "epoch": 2.89,
2053
- "learning_rate": 2e-05,
2054
- "loss": 5.1914,
2055
- "step": 341
2056
- },
2057
- {
2058
- "epoch": 2.9,
2059
- "learning_rate": 2e-05,
2060
- "loss": 5.0078,
2061
- "step": 342
2062
- },
2063
- {
2064
- "epoch": 2.91,
2065
- "learning_rate": 2e-05,
2066
- "loss": 5.2344,
2067
- "step": 343
2068
- },
2069
- {
2070
- "epoch": 2.92,
2071
- "learning_rate": 2e-05,
2072
- "loss": 4.5703,
2073
- "step": 344
2074
- },
2075
- {
2076
- "epoch": 2.92,
2077
- "learning_rate": 2e-05,
2078
- "loss": 4.8125,
2079
- "step": 345
2080
- },
2081
- {
2082
- "epoch": 2.93,
2083
- "learning_rate": 2e-05,
2084
- "loss": 5.5781,
2085
- "step": 346
2086
- },
2087
- {
2088
- "epoch": 2.94,
2089
- "learning_rate": 2e-05,
2090
- "loss": 3.8281,
2091
- "step": 347
2092
- },
2093
- {
2094
- "epoch": 2.95,
2095
- "learning_rate": 2e-05,
2096
- "loss": 5.4375,
2097
- "step": 348
2098
- },
2099
- {
2100
- "epoch": 2.96,
2101
- "learning_rate": 2e-05,
2102
- "loss": 4.832,
2103
- "step": 349
2104
- },
2105
- {
2106
- "epoch": 2.97,
2107
- "learning_rate": 2e-05,
2108
- "loss": 5.375,
2109
- "step": 350
2110
- },
2111
- {
2112
- "epoch": 2.97,
2113
- "learning_rate": 2e-05,
2114
- "loss": 5.0977,
2115
- "step": 351
2116
- },
2117
- {
2118
- "epoch": 2.98,
2119
- "learning_rate": 2e-05,
2120
- "loss": 5.3125,
2121
- "step": 352
2122
- },
2123
- {
2124
- "epoch": 2.99,
2125
- "learning_rate": 2e-05,
2126
- "loss": 4.9531,
2127
- "step": 353
2128
- },
2129
- {
2130
- "epoch": 3.0,
2131
- "learning_rate": 2e-05,
2132
- "loss": 5.1016,
2133
- "step": 354
2134
- }
2135
- ],
2136
- "logging_steps": 1.0,
2137
- "max_steps": 472,
2138
- "num_train_epochs": 4,
2139
- "save_steps": 500,
2140
- "total_flos": 13755035418624.0,
2141
- "trial_name": null,
2142
- "trial_params": null
2143
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
checkpoint-354/training_args.bin DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:ae681d6d6369bd19a423e7d631a266a0c8a18a4b8705cf9f2f00a80e2dbc0e83
3
- size 5816
 
 
 
 
checkpoint-354/zero_to_fp32.py DELETED
@@ -1,587 +0,0 @@
1
- #!/usr/bin/env python
2
-
3
- # Copyright (c) Microsoft Corporation.
4
- # SPDX-License-Identifier: Apache-2.0
5
-
6
- # DeepSpeed Team
7
-
8
- # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
- # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
- # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
- # application.
12
- #
13
- # example: python zero_to_fp32.py . pytorch_model.bin
14
-
15
- import argparse
16
- import torch
17
- import glob
18
- import math
19
- import os
20
- import re
21
- from collections import OrderedDict
22
- from dataclasses import dataclass
23
-
24
- # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
25
- # DeepSpeed data structures it has to be available in the current python environment.
26
- from deepspeed.utils import logger
27
- from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
28
- FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
29
- FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
30
-
31
-
32
- @dataclass
33
- class zero_model_state:
34
- buffers: dict()
35
- param_shapes: dict()
36
- shared_params: list
37
- ds_version: int
38
- frozen_param_shapes: dict()
39
- frozen_param_fragments: dict()
40
-
41
-
42
- debug = 0
43
-
44
- # load to cpu
45
- device = torch.device('cpu')
46
-
47
-
48
- def atoi(text):
49
- return int(text) if text.isdigit() else text
50
-
51
-
52
- def natural_keys(text):
53
- '''
54
- alist.sort(key=natural_keys) sorts in human order
55
- http://nedbatchelder.com/blog/200712/human_sorting.html
56
- (See Toothy's implementation in the comments)
57
- '''
58
- return [atoi(c) for c in re.split(r'(\d+)', text)]
59
-
60
-
61
- def get_model_state_file(checkpoint_dir, zero_stage):
62
- if not os.path.isdir(checkpoint_dir):
63
- raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
64
-
65
- # there should be only one file
66
- if zero_stage <= 2:
67
- file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
68
- elif zero_stage == 3:
69
- file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
70
-
71
- if not os.path.exists(file):
72
- raise FileNotFoundError(f"can't find model states file at '{file}'")
73
-
74
- return file
75
-
76
-
77
- def get_checkpoint_files(checkpoint_dir, glob_pattern):
78
- # XXX: need to test that this simple glob rule works for multi-node setup too
79
- ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
80
-
81
- if len(ckpt_files) == 0:
82
- raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
83
-
84
- return ckpt_files
85
-
86
-
87
- def get_optim_files(checkpoint_dir):
88
- return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
89
-
90
-
91
- def get_model_state_files(checkpoint_dir):
92
- return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
93
-
94
-
95
- def parse_model_states(files):
96
- zero_model_states = []
97
- for file in files:
98
- state_dict = torch.load(file, map_location=device)
99
-
100
- if BUFFER_NAMES not in state_dict:
101
- raise ValueError(f"{file} is not a model state checkpoint")
102
- buffer_names = state_dict[BUFFER_NAMES]
103
- if debug:
104
- print("Found buffers:", buffer_names)
105
-
106
- # recover just the buffers while restoring them to fp32 if they were saved in fp16
107
- buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
108
- param_shapes = state_dict[PARAM_SHAPES]
109
-
110
- # collect parameters that are included in param_shapes
111
- param_names = []
112
- for s in param_shapes:
113
- for name in s.keys():
114
- param_names.append(name)
115
-
116
- # update with frozen parameters
117
- frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
118
- if frozen_param_shapes is not None:
119
- if debug:
120
- print(f"Found frozen_param_shapes: {frozen_param_shapes}")
121
- param_names += list(frozen_param_shapes.keys())
122
-
123
- # handle shared params
124
- shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
125
-
126
- ds_version = state_dict.get(DS_VERSION, None)
127
-
128
- frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
129
-
130
- z_model_state = zero_model_state(buffers=buffers,
131
- param_shapes=param_shapes,
132
- shared_params=shared_params,
133
- ds_version=ds_version,
134
- frozen_param_shapes=frozen_param_shapes,
135
- frozen_param_fragments=frozen_param_fragments)
136
- zero_model_states.append(z_model_state)
137
-
138
- return zero_model_states
139
-
140
-
141
- def parse_optim_states(files, ds_checkpoint_dir):
142
-
143
- total_files = len(files)
144
- state_dicts = []
145
- for f in files:
146
- state_dict = torch.load(f, map_location=device)
147
- # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
148
- # and also handle the case where it was already removed by another helper script
149
- state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
150
- state_dicts.append(state_dict)
151
-
152
- if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
153
- raise ValueError(f"{files[0]} is not a zero checkpoint")
154
- zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
155
- world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
156
-
157
- # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
158
- # parameters can be different from data parallelism for non-expert parameters. So we can just
159
- # use the max of the partition_count to get the dp world_size.
160
-
161
- if type(world_size) is list:
162
- world_size = max(world_size)
163
-
164
- if world_size != total_files:
165
- raise ValueError(
166
- f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
167
- "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
168
- )
169
-
170
- # the groups are named differently in each stage
171
- if zero_stage <= 2:
172
- fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
173
- elif zero_stage == 3:
174
- fp32_groups_key = FP32_FLAT_GROUPS
175
- else:
176
- raise ValueError(f"unknown zero stage {zero_stage}")
177
-
178
- if zero_stage <= 2:
179
- fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
180
- elif zero_stage == 3:
181
- # if there is more than one param group, there will be multiple flattened tensors - one
182
- # flattened tensor per group - for simplicity merge them into a single tensor
183
- #
184
- # XXX: could make the script more memory efficient for when there are multiple groups - it
185
- # will require matching the sub-lists of param_shapes for each param group flattened tensor
186
-
187
- fp32_flat_groups = [
188
- torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
189
- ]
190
-
191
- return zero_stage, world_size, fp32_flat_groups
192
-
193
-
194
- def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
195
- """
196
- Returns fp32 state_dict reconstructed from ds checkpoint
197
-
198
- Args:
199
- - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
200
-
201
- """
202
- print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
203
-
204
- optim_files = get_optim_files(ds_checkpoint_dir)
205
- zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
206
- print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
207
-
208
- model_files = get_model_state_files(ds_checkpoint_dir)
209
-
210
- zero_model_states = parse_model_states(model_files)
211
- print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
212
-
213
- if zero_stage <= 2:
214
- return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
215
- elif zero_stage == 3:
216
- return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
217
-
218
-
219
- def _zero2_merge_frozen_params(state_dict, zero_model_states):
220
- if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
221
- return
222
-
223
- frozen_param_shapes = zero_model_states[0].frozen_param_shapes
224
- frozen_param_fragments = zero_model_states[0].frozen_param_fragments
225
-
226
- if debug:
227
- num_elem = sum(s.numel() for s in frozen_param_shapes.values())
228
- print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
229
-
230
- wanted_params = len(frozen_param_shapes)
231
- wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
232
- avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
233
- print(f'Frozen params: Have {avail_numel} numels to process.')
234
- print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
235
-
236
- total_params = 0
237
- total_numel = 0
238
- for name, shape in frozen_param_shapes.items():
239
- total_params += 1
240
- unpartitioned_numel = shape.numel()
241
- total_numel += unpartitioned_numel
242
-
243
- state_dict[name] = frozen_param_fragments[name]
244
-
245
- if debug:
246
- print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
247
-
248
- print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
249
-
250
-
251
- def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
252
- param_shapes = zero_model_states[0].param_shapes
253
-
254
- # Reconstruction protocol:
255
- #
256
- # XXX: document this
257
-
258
- if debug:
259
- for i in range(world_size):
260
- for j in range(len(fp32_flat_groups[0])):
261
- print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
262
-
263
- # XXX: memory usage doubles here (zero2)
264
- num_param_groups = len(fp32_flat_groups[0])
265
- merged_single_partition_of_fp32_groups = []
266
- for i in range(num_param_groups):
267
- merged_partitions = [sd[i] for sd in fp32_flat_groups]
268
- full_single_fp32_vector = torch.cat(merged_partitions, 0)
269
- merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
270
- avail_numel = sum(
271
- [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
272
-
273
- if debug:
274
- wanted_params = sum([len(shapes) for shapes in param_shapes])
275
- wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
276
- # not asserting if there is a mismatch due to possible padding
277
- print(f"Have {avail_numel} numels to process.")
278
- print(f"Need {wanted_numel} numels in {wanted_params} params.")
279
-
280
- # params
281
- # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
282
- # out-of-core computing solution
283
- total_numel = 0
284
- total_params = 0
285
- for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
286
- offset = 0
287
- avail_numel = full_single_fp32_vector.numel()
288
- for name, shape in shapes.items():
289
-
290
- unpartitioned_numel = shape.numel()
291
- total_numel += unpartitioned_numel
292
- total_params += 1
293
-
294
- if debug:
295
- print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
296
- state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
297
- offset += unpartitioned_numel
298
-
299
- # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
300
- # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
301
- # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
302
- # live optimizer object, so we are checking that the numbers are within the right range
303
- align_to = 2 * world_size
304
-
305
- def zero2_align(x):
306
- return align_to * math.ceil(x / align_to)
307
-
308
- if debug:
309
- print(f"original offset={offset}, avail_numel={avail_numel}")
310
-
311
- offset = zero2_align(offset)
312
- avail_numel = zero2_align(avail_numel)
313
-
314
- if debug:
315
- print(f"aligned offset={offset}, avail_numel={avail_numel}")
316
-
317
- # Sanity check
318
- if offset != avail_numel:
319
- raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
320
-
321
- print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
322
-
323
-
324
- def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
325
- state_dict = OrderedDict()
326
-
327
- # buffers
328
- buffers = zero_model_states[0].buffers
329
- state_dict.update(buffers)
330
- if debug:
331
- print(f"added {len(buffers)} buffers")
332
-
333
- _zero2_merge_frozen_params(state_dict, zero_model_states)
334
-
335
- _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
336
-
337
- # recover shared parameters
338
- for pair in zero_model_states[0].shared_params:
339
- if pair[1] in state_dict:
340
- state_dict[pair[0]] = state_dict[pair[1]]
341
-
342
- return state_dict
343
-
344
-
345
- def zero3_partitioned_param_info(unpartitioned_numel, world_size):
346
- remainder = unpartitioned_numel % world_size
347
- padding_numel = (world_size - remainder) if remainder else 0
348
- partitioned_numel = math.ceil(unpartitioned_numel / world_size)
349
- return partitioned_numel, padding_numel
350
-
351
-
352
- def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
353
- if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
354
- return
355
-
356
- if debug:
357
- for i in range(world_size):
358
- num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
359
- print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
360
-
361
- frozen_param_shapes = zero_model_states[0].frozen_param_shapes
362
- wanted_params = len(frozen_param_shapes)
363
- wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
364
- avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
365
- print(f'Frozen params: Have {avail_numel} numels to process.')
366
- print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
367
-
368
- total_params = 0
369
- total_numel = 0
370
- for name, shape in zero_model_states[0].frozen_param_shapes.items():
371
- total_params += 1
372
- unpartitioned_numel = shape.numel()
373
- total_numel += unpartitioned_numel
374
-
375
- param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
376
- state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
377
-
378
- partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
379
-
380
- if debug:
381
- print(
382
- f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
383
- )
384
-
385
- print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
386
-
387
-
388
- def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
389
- param_shapes = zero_model_states[0].param_shapes
390
- avail_numel = fp32_flat_groups[0].numel() * world_size
391
- # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
392
- # param, re-consolidating each param, while dealing with padding if any
393
-
394
- # merge list of dicts, preserving order
395
- param_shapes = {k: v for d in param_shapes for k, v in d.items()}
396
-
397
- if debug:
398
- for i in range(world_size):
399
- print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
400
-
401
- wanted_params = len(param_shapes)
402
- wanted_numel = sum(shape.numel() for shape in param_shapes.values())
403
- # not asserting if there is a mismatch due to possible padding
404
- avail_numel = fp32_flat_groups[0].numel() * world_size
405
- print(f"Trainable params: Have {avail_numel} numels to process.")
406
- print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
407
-
408
- # params
409
- # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
410
- # out-of-core computing solution
411
- offset = 0
412
- total_numel = 0
413
- total_params = 0
414
- for name, shape in param_shapes.items():
415
-
416
- unpartitioned_numel = shape.numel()
417
- total_numel += unpartitioned_numel
418
- total_params += 1
419
-
420
- partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
421
-
422
- if debug:
423
- print(
424
- f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
425
- )
426
-
427
- # XXX: memory usage doubles here
428
- state_dict[name] = torch.cat(
429
- tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
430
- 0).narrow(0, 0, unpartitioned_numel).view(shape)
431
- offset += partitioned_numel
432
-
433
- offset *= world_size
434
-
435
- # Sanity check
436
- if offset != avail_numel:
437
- raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
438
-
439
- print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
440
-
441
-
442
- def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
443
- state_dict = OrderedDict()
444
-
445
- # buffers
446
- buffers = zero_model_states[0].buffers
447
- state_dict.update(buffers)
448
- if debug:
449
- print(f"added {len(buffers)} buffers")
450
-
451
- _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
452
-
453
- _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
454
-
455
- # recover shared parameters
456
- for pair in zero_model_states[0].shared_params:
457
- if pair[1] in state_dict:
458
- state_dict[pair[0]] = state_dict[pair[1]]
459
-
460
- return state_dict
461
-
462
-
463
- def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
464
- """
465
- Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
466
- ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
467
- via a model hub.
468
-
469
- Args:
470
- - ``checkpoint_dir``: path to the desired checkpoint folder
471
- - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
472
-
473
- Returns:
474
- - pytorch ``state_dict``
475
-
476
- Note: this approach may not work if your application doesn't have sufficient free CPU memory and
477
- you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
478
- the checkpoint.
479
-
480
- A typical usage might be ::
481
-
482
- from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
483
- # do the training and checkpoint saving
484
- state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
485
- model = model.cpu() # move to cpu
486
- model.load_state_dict(state_dict)
487
- # submit to model hub or save the model to share with others
488
-
489
- In this example the ``model`` will no longer be usable in the deepspeed context of the same
490
- application. i.e. you will need to re-initialize the deepspeed engine, since
491
- ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
492
-
493
- If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
494
-
495
- """
496
- if tag is None:
497
- latest_path = os.path.join(checkpoint_dir, 'latest')
498
- if os.path.isfile(latest_path):
499
- with open(latest_path, 'r') as fd:
500
- tag = fd.read().strip()
501
- else:
502
- raise ValueError(f"Unable to find 'latest' file at {latest_path}")
503
-
504
- ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
505
-
506
- if not os.path.isdir(ds_checkpoint_dir):
507
- raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
508
-
509
- return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
510
-
511
-
512
- def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
513
- """
514
- Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
515
- loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
516
-
517
- Args:
518
- - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
519
- - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
520
- - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
521
- """
522
-
523
- state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
524
- print(f"Saving fp32 state dict to {output_file}")
525
- torch.save(state_dict, output_file)
526
-
527
-
528
- def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
529
- """
530
- 1. Put the provided model to cpu
531
- 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
532
- 3. Load it into the provided model
533
-
534
- Args:
535
- - ``model``: the model object to update
536
- - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
537
- - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
538
-
539
- Returns:
540
- - ``model`: modified model
541
-
542
- Make sure you have plenty of CPU memory available before you call this function. If you don't
543
- have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
544
- conveniently placed for you in the checkpoint folder.
545
-
546
- A typical usage might be ::
547
-
548
- from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
549
- model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
550
- # submit to model hub or save the model to share with others
551
-
552
- Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
553
- of the same application. i.e. you will need to re-initialize the deepspeed engine, since
554
- ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
555
-
556
- """
557
- logger.info(f"Extracting fp32 weights")
558
- state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
559
-
560
- logger.info(f"Overwriting model with fp32 weights")
561
- model = model.cpu()
562
- model.load_state_dict(state_dict, strict=False)
563
-
564
- return model
565
-
566
-
567
- if __name__ == "__main__":
568
-
569
- parser = argparse.ArgumentParser()
570
- parser.add_argument("checkpoint_dir",
571
- type=str,
572
- help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
573
- parser.add_argument(
574
- "output_file",
575
- type=str,
576
- help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
577
- parser.add_argument("-t",
578
- "--tag",
579
- type=str,
580
- default=None,
581
- help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
582
- parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
583
- args = parser.parse_args()
584
-
585
- debug = args.debug
586
-
587
- convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file, tag=args.tag)