abdoelsayed commited on
Commit
7eebc09
·
verified ·
1 Parent(s): 77eec7c

Upload folder using huggingface_hub

Browse files
Files changed (8) hide show
  1. README.md +202 -0
  2. adapter_config.json +36 -0
  3. adapter_model.bin +3 -0
  4. latest +1 -0
  5. rng_state_0.pth +3 -0
  6. rng_state_1.pth +3 -0
  7. trainer_state.json +2133 -0
  8. zero_to_fp32.py +674 -0
README.md ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: meta-llama/Llama-3.1-8B
3
+ library_name: peft
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
+ ### Framework versions
201
+
202
+ - PEFT 0.13.2
adapter_config.json ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "meta-llama/Llama-3.1-8B",
5
+ "bias": "none",
6
+ "fan_in_fan_out": false,
7
+ "inference_mode": true,
8
+ "init_lora_weights": true,
9
+ "layer_replication": null,
10
+ "layers_pattern": null,
11
+ "layers_to_transform": null,
12
+ "loftq_config": {},
13
+ "lora_alpha": 64,
14
+ "lora_dropout": 0.1,
15
+ "megatron_config": null,
16
+ "megatron_core": "megatron.core",
17
+ "modules_to_save": [
18
+ "classifier",
19
+ "score"
20
+ ],
21
+ "peft_type": "LORA",
22
+ "r": 32,
23
+ "rank_pattern": {},
24
+ "revision": null,
25
+ "target_modules": [
26
+ "o_proj",
27
+ "q_proj",
28
+ "down_proj",
29
+ "up_proj",
30
+ "gate_proj",
31
+ "v_proj"
32
+ ],
33
+ "task_type": "SEQ_CLS",
34
+ "use_dora": false,
35
+ "use_rslora": false
36
+ }
adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:275e634fb50f6c12e49354f2bf2d1ff9aa62711506bae1bec3471abd4951f5b4
3
+ size 157428238
latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step3000
rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f26e041487167f77da21a2ac72855409f196531836f793417f7ca7d8d5ab9fe8
3
+ size 14512
rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60d5c6269a50366efa7d074b57043dcb5eacf969ecd38235288b84c61e372e87
3
+ size 14512
trainer_state.json ADDED
@@ -0,0 +1,2133 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.11976526009022316,
5
+ "eval_steps": 500,
6
+ "global_step": 3000,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0003992175336340772,
13
+ "grad_norm": 19.85781045879126,
14
+ "learning_rate": 2.7028212426954458e-05,
15
+ "loss": 0.8612,
16
+ "step": 10
17
+ },
18
+ {
19
+ "epoch": 0.0007984350672681544,
20
+ "grad_norm": 19.882645890181408,
21
+ "learning_rate": 3.516451509664572e-05,
22
+ "loss": 0.7978,
23
+ "step": 20
24
+ },
25
+ {
26
+ "epoch": 0.0011976526009022316,
27
+ "grad_norm": 9.094846045432918,
28
+ "learning_rate": 3.992394705293254e-05,
29
+ "loss": 0.7084,
30
+ "step": 30
31
+ },
32
+ {
33
+ "epoch": 0.0015968701345363088,
34
+ "grad_norm": 8.935490690487004,
35
+ "learning_rate": 4.330081776633698e-05,
36
+ "loss": 0.4791,
37
+ "step": 40
38
+ },
39
+ {
40
+ "epoch": 0.001996087668170386,
41
+ "grad_norm": 10.34774048559344,
42
+ "learning_rate": 4.592012218421765e-05,
43
+ "loss": 0.3076,
44
+ "step": 50
45
+ },
46
+ {
47
+ "epoch": 0.0023953052018044632,
48
+ "grad_norm": 5.330934607110759,
49
+ "learning_rate": 4.80602497226238e-05,
50
+ "loss": 0.292,
51
+ "step": 60
52
+ },
53
+ {
54
+ "epoch": 0.0027945227354385406,
55
+ "grad_norm": 23.54140808901255,
56
+ "learning_rate": 4.9869701774062655e-05,
57
+ "loss": 0.2312,
58
+ "step": 70
59
+ },
60
+ {
61
+ "epoch": 0.0031937402690726175,
62
+ "grad_norm": 17.28505020343765,
63
+ "learning_rate": 5.143712043602824e-05,
64
+ "loss": 0.2473,
65
+ "step": 80
66
+ },
67
+ {
68
+ "epoch": 0.003592957802706695,
69
+ "grad_norm": 24.762507669029908,
70
+ "learning_rate": 5.281968167891063e-05,
71
+ "loss": 0.2292,
72
+ "step": 90
73
+ },
74
+ {
75
+ "epoch": 0.003992175336340772,
76
+ "grad_norm": 10.968261437457826,
77
+ "learning_rate": 5.4056424853908915e-05,
78
+ "loss": 0.1577,
79
+ "step": 100
80
+ },
81
+ {
82
+ "epoch": 0.004391392869974849,
83
+ "grad_norm": 0.7358009926776751,
84
+ "learning_rate": 5.5175195141287474e-05,
85
+ "loss": 0.1377,
86
+ "step": 110
87
+ },
88
+ {
89
+ "epoch": 0.0047906104036089265,
90
+ "grad_norm": 2.6196353846553566,
91
+ "learning_rate": 5.619655239231506e-05,
92
+ "loss": 0.096,
93
+ "step": 120
94
+ },
95
+ {
96
+ "epoch": 0.005189827937243004,
97
+ "grad_norm": 4.457945145184635,
98
+ "learning_rate": 5.7136109984697405e-05,
99
+ "loss": 0.1624,
100
+ "step": 130
101
+ },
102
+ {
103
+ "epoch": 0.005589045470877081,
104
+ "grad_norm": 5.730548083937804,
105
+ "learning_rate": 5.8006004443753904e-05,
106
+ "loss": 0.1062,
107
+ "step": 140
108
+ },
109
+ {
110
+ "epoch": 0.0059882630045111586,
111
+ "grad_norm": 11.523201996978061,
112
+ "learning_rate": 5.8815856810195735e-05,
113
+ "loss": 0.2052,
114
+ "step": 150
115
+ },
116
+ {
117
+ "epoch": 0.006387480538145235,
118
+ "grad_norm": 0.8091674882450949,
119
+ "learning_rate": 5.957342310571951e-05,
120
+ "loss": 0.1584,
121
+ "step": 160
122
+ },
123
+ {
124
+ "epoch": 0.006786698071779312,
125
+ "grad_norm": 1.0227188199759876,
126
+ "learning_rate": 6.028504725448343e-05,
127
+ "loss": 0.131,
128
+ "step": 170
129
+ },
130
+ {
131
+ "epoch": 0.00718591560541339,
132
+ "grad_norm": 8.25751680884371,
133
+ "learning_rate": 6.095598434860189e-05,
134
+ "loss": 0.1698,
135
+ "step": 180
136
+ },
137
+ {
138
+ "epoch": 0.007585133139047467,
139
+ "grad_norm": 0.587472316727206,
140
+ "learning_rate": 6.159063639524046e-05,
141
+ "loss": 0.0801,
142
+ "step": 190
143
+ },
144
+ {
145
+ "epoch": 0.007984350672681544,
146
+ "grad_norm": 1.91099848144843,
147
+ "learning_rate": 6.219272752360016e-05,
148
+ "loss": 0.2088,
149
+ "step": 200
150
+ },
151
+ {
152
+ "epoch": 0.008383568206315622,
153
+ "grad_norm": 4.47604298858692,
154
+ "learning_rate": 6.276543640004074e-05,
155
+ "loss": 0.214,
156
+ "step": 210
157
+ },
158
+ {
159
+ "epoch": 0.008782785739949698,
160
+ "grad_norm": 4.502483070458412,
161
+ "learning_rate": 6.331149781097874e-05,
162
+ "loss": 0.101,
163
+ "step": 220
164
+ },
165
+ {
166
+ "epoch": 0.009182003273583777,
167
+ "grad_norm": 1.079585989368236,
168
+ "learning_rate": 6.383328164653496e-05,
169
+ "loss": 0.0536,
170
+ "step": 230
171
+ },
172
+ {
173
+ "epoch": 0.009581220807217853,
174
+ "grad_norm": 2.160781559557002,
175
+ "learning_rate": 6.433285506200633e-05,
176
+ "loss": 0.1073,
177
+ "step": 240
178
+ },
179
+ {
180
+ "epoch": 0.00998043834085193,
181
+ "grad_norm": 1.2451244100698013,
182
+ "learning_rate": 6.481203194148085e-05,
183
+ "loss": 0.0587,
184
+ "step": 250
185
+ },
186
+ {
187
+ "epoch": 0.010379655874486008,
188
+ "grad_norm": 1.2141713778196777,
189
+ "learning_rate": 6.527241265438867e-05,
190
+ "loss": 0.0979,
191
+ "step": 260
192
+ },
193
+ {
194
+ "epoch": 0.010778873408120084,
195
+ "grad_norm": 0.4009094506871956,
196
+ "learning_rate": 6.57154163048887e-05,
197
+ "loss": 0.0874,
198
+ "step": 270
199
+ },
200
+ {
201
+ "epoch": 0.011178090941754162,
202
+ "grad_norm": 0.8911240034837602,
203
+ "learning_rate": 6.614230711344517e-05,
204
+ "loss": 0.1043,
205
+ "step": 280
206
+ },
207
+ {
208
+ "epoch": 0.011577308475388239,
209
+ "grad_norm": 0.807314642103107,
210
+ "learning_rate": 6.655421616692034e-05,
211
+ "loss": 0.0834,
212
+ "step": 290
213
+ },
214
+ {
215
+ "epoch": 0.011976526009022317,
216
+ "grad_norm": 0.5194314428720426,
217
+ "learning_rate": 6.6952159479887e-05,
218
+ "loss": 0.161,
219
+ "step": 300
220
+ },
221
+ {
222
+ "epoch": 0.012375743542656394,
223
+ "grad_norm": 4.705567502700896,
224
+ "learning_rate": 6.733705309332979e-05,
225
+ "loss": 0.1208,
226
+ "step": 310
227
+ },
228
+ {
229
+ "epoch": 0.01277496107629047,
230
+ "grad_norm": 1.4163866939154937,
231
+ "learning_rate": 6.770972577541077e-05,
232
+ "loss": 0.0435,
233
+ "step": 320
234
+ },
235
+ {
236
+ "epoch": 0.013174178609924548,
237
+ "grad_norm": 2.9342139394642035,
238
+ "learning_rate": 6.807092976726554e-05,
239
+ "loss": 0.0686,
240
+ "step": 330
241
+ },
242
+ {
243
+ "epoch": 0.013573396143558625,
244
+ "grad_norm": 4.76683147201656,
245
+ "learning_rate": 6.842134992417469e-05,
246
+ "loss": 0.0791,
247
+ "step": 340
248
+ },
249
+ {
250
+ "epoch": 0.013972613677192703,
251
+ "grad_norm": 2.8527837428746294,
252
+ "learning_rate": 6.876161153132584e-05,
253
+ "loss": 0.086,
254
+ "step": 350
255
+ },
256
+ {
257
+ "epoch": 0.01437183121082678,
258
+ "grad_norm": 0.3845524480091741,
259
+ "learning_rate": 6.909228701829316e-05,
260
+ "loss": 0.1051,
261
+ "step": 360
262
+ },
263
+ {
264
+ "epoch": 0.014771048744460856,
265
+ "grad_norm": 4.410564236686768,
266
+ "learning_rate": 6.941390175335342e-05,
267
+ "loss": 0.0703,
268
+ "step": 370
269
+ },
270
+ {
271
+ "epoch": 0.015170266278094934,
272
+ "grad_norm": 0.9740892631024635,
273
+ "learning_rate": 6.972693906493173e-05,
274
+ "loss": 0.0624,
275
+ "step": 380
276
+ },
277
+ {
278
+ "epoch": 0.01556948381172901,
279
+ "grad_norm": 3.0544470063015607,
280
+ "learning_rate": 7.003184461067549e-05,
281
+ "loss": 0.0736,
282
+ "step": 390
283
+ },
284
+ {
285
+ "epoch": 0.015968701345363087,
286
+ "grad_norm": 0.41639345834318137,
287
+ "learning_rate": 7.032903019329144e-05,
288
+ "loss": 0.1053,
289
+ "step": 400
290
+ },
291
+ {
292
+ "epoch": 0.016367918878997165,
293
+ "grad_norm": 6.743718794059417,
294
+ "learning_rate": 7.061887710513835e-05,
295
+ "loss": 0.0689,
296
+ "step": 410
297
+ },
298
+ {
299
+ "epoch": 0.016767136412631244,
300
+ "grad_norm": 1.00073139371301,
301
+ "learning_rate": 7.0901739069732e-05,
302
+ "loss": 0.0474,
303
+ "step": 420
304
+ },
305
+ {
306
+ "epoch": 0.01716635394626532,
307
+ "grad_norm": 2.799355054285757,
308
+ "learning_rate": 7.117794483708874e-05,
309
+ "loss": 0.0747,
310
+ "step": 430
311
+ },
312
+ {
313
+ "epoch": 0.017565571479899397,
314
+ "grad_norm": 3.603772769563174,
315
+ "learning_rate": 7.144780048066999e-05,
316
+ "loss": 0.0913,
317
+ "step": 440
318
+ },
319
+ {
320
+ "epoch": 0.017964789013533475,
321
+ "grad_norm": 0.49569502418002975,
322
+ "learning_rate": 7.171159143617382e-05,
323
+ "loss": 0.055,
324
+ "step": 450
325
+ },
326
+ {
327
+ "epoch": 0.018364006547167553,
328
+ "grad_norm": 0.2783746184212658,
329
+ "learning_rate": 7.196958431622622e-05,
330
+ "loss": 0.072,
331
+ "step": 460
332
+ },
333
+ {
334
+ "epoch": 0.018763224080801628,
335
+ "grad_norm": 0.42607556365268057,
336
+ "learning_rate": 7.222202852989654e-05,
337
+ "loss": 0.1137,
338
+ "step": 470
339
+ },
340
+ {
341
+ "epoch": 0.019162441614435706,
342
+ "grad_norm": 0.4064736743213967,
343
+ "learning_rate": 7.246915773169759e-05,
344
+ "loss": 0.0695,
345
+ "step": 480
346
+ },
347
+ {
348
+ "epoch": 0.019561659148069784,
349
+ "grad_norm": 0.9397768981809936,
350
+ "learning_rate": 7.271119112117084e-05,
351
+ "loss": 0.0932,
352
+ "step": 490
353
+ },
354
+ {
355
+ "epoch": 0.01996087668170386,
356
+ "grad_norm": 3.6579870512864274,
357
+ "learning_rate": 7.294833461117211e-05,
358
+ "loss": 0.0689,
359
+ "step": 500
360
+ },
361
+ {
362
+ "epoch": 0.020360094215337937,
363
+ "grad_norm": 2.096579140979539,
364
+ "learning_rate": 7.318078188046151e-05,
365
+ "loss": 0.0848,
366
+ "step": 510
367
+ },
368
+ {
369
+ "epoch": 0.020759311748972015,
370
+ "grad_norm": 0.928087775020366,
371
+ "learning_rate": 7.340871532407993e-05,
372
+ "loss": 0.0572,
373
+ "step": 520
374
+ },
375
+ {
376
+ "epoch": 0.021158529282606094,
377
+ "grad_norm": 4.446858720966493,
378
+ "learning_rate": 7.363230691319621e-05,
379
+ "loss": 0.0872,
380
+ "step": 530
381
+ },
382
+ {
383
+ "epoch": 0.02155774681624017,
384
+ "grad_norm": 1.328344056531029,
385
+ "learning_rate": 7.385171897457998e-05,
386
+ "loss": 0.1129,
387
+ "step": 540
388
+ },
389
+ {
390
+ "epoch": 0.021956964349874247,
391
+ "grad_norm": 0.35804452540240167,
392
+ "learning_rate": 7.406710489855066e-05,
393
+ "loss": 0.1162,
394
+ "step": 550
395
+ },
396
+ {
397
+ "epoch": 0.022356181883508325,
398
+ "grad_norm": 0.6051847155052836,
399
+ "learning_rate": 7.427860978313643e-05,
400
+ "loss": 0.0588,
401
+ "step": 560
402
+ },
403
+ {
404
+ "epoch": 0.0227553994171424,
405
+ "grad_norm": 2.29458938702116,
406
+ "learning_rate": 7.448637102121855e-05,
407
+ "loss": 0.1602,
408
+ "step": 570
409
+ },
410
+ {
411
+ "epoch": 0.023154616950776478,
412
+ "grad_norm": 1.2097365397485473,
413
+ "learning_rate": 7.46905188366116e-05,
414
+ "loss": 0.1042,
415
+ "step": 580
416
+ },
417
+ {
418
+ "epoch": 0.023553834484410556,
419
+ "grad_norm": 0.7262308596218504,
420
+ "learning_rate": 7.489117677431795e-05,
421
+ "loss": 0.0738,
422
+ "step": 590
423
+ },
424
+ {
425
+ "epoch": 0.023953052018044634,
426
+ "grad_norm": 0.9117785252438455,
427
+ "learning_rate": 7.508846214957826e-05,
428
+ "loss": 0.0791,
429
+ "step": 600
430
+ },
431
+ {
432
+ "epoch": 0.02435226955167871,
433
+ "grad_norm": 1.6484189852015583,
434
+ "learning_rate": 7.528248645980503e-05,
435
+ "loss": 0.0642,
436
+ "step": 610
437
+ },
438
+ {
439
+ "epoch": 0.024751487085312787,
440
+ "grad_norm": 2.6615882255939907,
441
+ "learning_rate": 7.547335576302105e-05,
442
+ "loss": 0.1194,
443
+ "step": 620
444
+ },
445
+ {
446
+ "epoch": 0.025150704618946865,
447
+ "grad_norm": 0.8346103246891559,
448
+ "learning_rate": 7.566117102601882e-05,
449
+ "loss": 0.0638,
450
+ "step": 630
451
+ },
452
+ {
453
+ "epoch": 0.02554992215258094,
454
+ "grad_norm": 3.174551708543591,
455
+ "learning_rate": 7.584602844510203e-05,
456
+ "loss": 0.0926,
457
+ "step": 640
458
+ },
459
+ {
460
+ "epoch": 0.02594913968621502,
461
+ "grad_norm": 1.879296095806724,
462
+ "learning_rate": 7.60280197419606e-05,
463
+ "loss": 0.0787,
464
+ "step": 650
465
+ },
466
+ {
467
+ "epoch": 0.026348357219849097,
468
+ "grad_norm": 0.39642282227270165,
469
+ "learning_rate": 7.62072324369568e-05,
470
+ "loss": 0.0602,
471
+ "step": 660
472
+ },
473
+ {
474
+ "epoch": 0.02674757475348317,
475
+ "grad_norm": 1.0281650863378988,
476
+ "learning_rate": 7.638375010186134e-05,
477
+ "loss": 0.1164,
478
+ "step": 670
479
+ },
480
+ {
481
+ "epoch": 0.02714679228711725,
482
+ "grad_norm": 0.23324571822799267,
483
+ "learning_rate": 7.655765259386594e-05,
484
+ "loss": 0.0585,
485
+ "step": 680
486
+ },
487
+ {
488
+ "epoch": 0.027546009820751328,
489
+ "grad_norm": 2.602214827786925,
490
+ "learning_rate": 7.672901627251304e-05,
491
+ "loss": 0.0614,
492
+ "step": 690
493
+ },
494
+ {
495
+ "epoch": 0.027945227354385406,
496
+ "grad_norm": 1.1814042832377207,
497
+ "learning_rate": 7.689791420101711e-05,
498
+ "loss": 0.0889,
499
+ "step": 700
500
+ },
501
+ {
502
+ "epoch": 0.02834444488801948,
503
+ "grad_norm": 1.1260525545149151,
504
+ "learning_rate": 7.706441633330655e-05,
505
+ "loss": 0.0493,
506
+ "step": 710
507
+ },
508
+ {
509
+ "epoch": 0.02874366242165356,
510
+ "grad_norm": 0.9772468975665439,
511
+ "learning_rate": 7.722858968798441e-05,
512
+ "loss": 0.0817,
513
+ "step": 720
514
+ },
515
+ {
516
+ "epoch": 0.029142879955287637,
517
+ "grad_norm": 0.7893326781926734,
518
+ "learning_rate": 7.739049851029049e-05,
519
+ "loss": 0.0594,
520
+ "step": 730
521
+ },
522
+ {
523
+ "epoch": 0.029542097488921712,
524
+ "grad_norm": 1.17044118759658,
525
+ "learning_rate": 7.755020442304467e-05,
526
+ "loss": 0.1009,
527
+ "step": 740
528
+ },
529
+ {
530
+ "epoch": 0.02994131502255579,
531
+ "grad_norm": 1.2944195265483465,
532
+ "learning_rate": 7.770776656745893e-05,
533
+ "loss": 0.0716,
534
+ "step": 750
535
+ },
536
+ {
537
+ "epoch": 0.03034053255618987,
538
+ "grad_norm": 0.25418619189738,
539
+ "learning_rate": 7.7863241734623e-05,
540
+ "loss": 0.0571,
541
+ "step": 760
542
+ },
543
+ {
544
+ "epoch": 0.030739750089823947,
545
+ "grad_norm": 0.20365098448653787,
546
+ "learning_rate": 7.801668448839566e-05,
547
+ "loss": 0.0633,
548
+ "step": 770
549
+ },
550
+ {
551
+ "epoch": 0.03113896762345802,
552
+ "grad_norm": 0.45663176305105035,
553
+ "learning_rate": 7.816814728036675e-05,
554
+ "loss": 0.1258,
555
+ "step": 780
556
+ },
557
+ {
558
+ "epoch": 0.0315381851570921,
559
+ "grad_norm": 2.1824315856202383,
560
+ "learning_rate": 7.831768055749621e-05,
561
+ "loss": 0.0752,
562
+ "step": 790
563
+ },
564
+ {
565
+ "epoch": 0.031937402690726174,
566
+ "grad_norm": 0.8260546225743964,
567
+ "learning_rate": 7.84653328629827e-05,
568
+ "loss": 0.0813,
569
+ "step": 800
570
+ },
571
+ {
572
+ "epoch": 0.032336620224360256,
573
+ "grad_norm": 2.0810842739820714,
574
+ "learning_rate": 7.861115093086679e-05,
575
+ "loss": 0.0565,
576
+ "step": 810
577
+ },
578
+ {
579
+ "epoch": 0.03273583775799433,
580
+ "grad_norm": 1.2546576105110292,
581
+ "learning_rate": 7.87551797748296e-05,
582
+ "loss": 0.1007,
583
+ "step": 820
584
+ },
585
+ {
586
+ "epoch": 0.033135055291628406,
587
+ "grad_norm": 0.6627056918627381,
588
+ "learning_rate": 7.889746277160951e-05,
589
+ "loss": 0.0553,
590
+ "step": 830
591
+ },
592
+ {
593
+ "epoch": 0.03353427282526249,
594
+ "grad_norm": 2.540953740571159,
595
+ "learning_rate": 7.903804173942325e-05,
596
+ "loss": 0.0796,
597
+ "step": 840
598
+ },
599
+ {
600
+ "epoch": 0.03393349035889656,
601
+ "grad_norm": 2.50083933812989,
602
+ "learning_rate": 7.917695701174663e-05,
603
+ "loss": 0.1192,
604
+ "step": 850
605
+ },
606
+ {
607
+ "epoch": 0.03433270789253064,
608
+ "grad_norm": 0.4101172295177441,
609
+ "learning_rate": 7.931424750678e-05,
610
+ "loss": 0.0846,
611
+ "step": 860
612
+ },
613
+ {
614
+ "epoch": 0.03473192542616472,
615
+ "grad_norm": 3.3827204370144104,
616
+ "learning_rate": 7.944995079289842e-05,
617
+ "loss": 0.0982,
618
+ "step": 870
619
+ },
620
+ {
621
+ "epoch": 0.03513114295979879,
622
+ "grad_norm": 0.852766983972299,
623
+ "learning_rate": 7.958410315036126e-05,
624
+ "loss": 0.0654,
625
+ "step": 880
626
+ },
627
+ {
628
+ "epoch": 0.035530360493432875,
629
+ "grad_norm": 0.28616682842501945,
630
+ "learning_rate": 7.971673962953532e-05,
631
+ "loss": 0.0729,
632
+ "step": 890
633
+ },
634
+ {
635
+ "epoch": 0.03592957802706695,
636
+ "grad_norm": 1.5774662264412462,
637
+ "learning_rate": 7.984789410586508e-05,
638
+ "loss": 0.0732,
639
+ "step": 900
640
+ },
641
+ {
642
+ "epoch": 0.036328795560701024,
643
+ "grad_norm": 0.38963613236731504,
644
+ "learning_rate": 7.99775993318056e-05,
645
+ "loss": 0.0998,
646
+ "step": 910
647
+ },
648
+ {
649
+ "epoch": 0.036728013094335106,
650
+ "grad_norm": 0.41983097281121634,
651
+ "learning_rate": 8.010588698591749e-05,
652
+ "loss": 0.0805,
653
+ "step": 920
654
+ },
655
+ {
656
+ "epoch": 0.03712723062796918,
657
+ "grad_norm": 0.338747268427734,
658
+ "learning_rate": 8.023278771930787e-05,
659
+ "loss": 0.0686,
660
+ "step": 930
661
+ },
662
+ {
663
+ "epoch": 0.037526448161603256,
664
+ "grad_norm": 2.285289873559961,
665
+ "learning_rate": 8.035833119958781e-05,
666
+ "loss": 0.0756,
667
+ "step": 940
668
+ },
669
+ {
670
+ "epoch": 0.03792566569523734,
671
+ "grad_norm": 0.7484525000680983,
672
+ "learning_rate": 8.048254615250367e-05,
673
+ "loss": 0.0824,
674
+ "step": 950
675
+ },
676
+ {
677
+ "epoch": 0.03832488322887141,
678
+ "grad_norm": 0.6877391029510214,
679
+ "learning_rate": 8.060546040138885e-05,
680
+ "loss": 0.1115,
681
+ "step": 960
682
+ },
683
+ {
684
+ "epoch": 0.03872410076250549,
685
+ "grad_norm": 3.34429646318092,
686
+ "learning_rate": 8.072710090457124e-05,
687
+ "loss": 0.114,
688
+ "step": 970
689
+ },
690
+ {
691
+ "epoch": 0.03912331829613957,
692
+ "grad_norm": 0.5091173852862666,
693
+ "learning_rate": 8.084749379086209e-05,
694
+ "loss": 0.0751,
695
+ "step": 980
696
+ },
697
+ {
698
+ "epoch": 0.03952253582977364,
699
+ "grad_norm": 0.7224047967736694,
700
+ "learning_rate": 8.096666439324362e-05,
701
+ "loss": 0.0717,
702
+ "step": 990
703
+ },
704
+ {
705
+ "epoch": 0.03992175336340772,
706
+ "grad_norm": 0.4146459011309395,
707
+ "learning_rate": 8.108463728086336e-05,
708
+ "loss": 0.0975,
709
+ "step": 1000
710
+ },
711
+ {
712
+ "epoch": 0.0403209708970418,
713
+ "grad_norm": 0.3482393432492753,
714
+ "learning_rate": 8.120143628943691e-05,
715
+ "loss": 0.057,
716
+ "step": 1010
717
+ },
718
+ {
719
+ "epoch": 0.040720188430675874,
720
+ "grad_norm": 2.9920581278472267,
721
+ "learning_rate": 8.131708455015276e-05,
722
+ "loss": 0.0762,
723
+ "step": 1020
724
+ },
725
+ {
726
+ "epoch": 0.04111940596430995,
727
+ "grad_norm": 1.0871148234026047,
728
+ "learning_rate": 8.143160451716731e-05,
729
+ "loss": 0.0711,
730
+ "step": 1030
731
+ },
732
+ {
733
+ "epoch": 0.04151862349794403,
734
+ "grad_norm": 0.7656470221828197,
735
+ "learning_rate": 8.15450179937712e-05,
736
+ "loss": 0.07,
737
+ "step": 1040
738
+ },
739
+ {
740
+ "epoch": 0.041917841031578106,
741
+ "grad_norm": 1.134395887081975,
742
+ "learning_rate": 8.165734615730393e-05,
743
+ "loss": 0.0574,
744
+ "step": 1050
745
+ },
746
+ {
747
+ "epoch": 0.04231705856521219,
748
+ "grad_norm": 0.28513287710915014,
749
+ "learning_rate": 8.176860958288747e-05,
750
+ "loss": 0.0714,
751
+ "step": 1060
752
+ },
753
+ {
754
+ "epoch": 0.04271627609884626,
755
+ "grad_norm": 0.4929994620467896,
756
+ "learning_rate": 8.187882826604561e-05,
757
+ "loss": 0.0744,
758
+ "step": 1070
759
+ },
760
+ {
761
+ "epoch": 0.04311549363248034,
762
+ "grad_norm": 2.26376869173837,
763
+ "learning_rate": 8.198802164427124e-05,
764
+ "loss": 0.0753,
765
+ "step": 1080
766
+ },
767
+ {
768
+ "epoch": 0.04351471116611442,
769
+ "grad_norm": 1.0769524498013112,
770
+ "learning_rate": 8.209620861759952e-05,
771
+ "loss": 0.0825,
772
+ "step": 1090
773
+ },
774
+ {
775
+ "epoch": 0.04391392869974849,
776
+ "grad_norm": 0.19856366308280235,
777
+ "learning_rate": 8.220340756824192e-05,
778
+ "loss": 0.0575,
779
+ "step": 1100
780
+ },
781
+ {
782
+ "epoch": 0.04431314623338257,
783
+ "grad_norm": 1.5972492667518492,
784
+ "learning_rate": 8.23096363793315e-05,
785
+ "loss": 0.0686,
786
+ "step": 1110
787
+ },
788
+ {
789
+ "epoch": 0.04471236376701665,
790
+ "grad_norm": 0.21452916203736186,
791
+ "learning_rate": 8.241491245282769e-05,
792
+ "loss": 0.0508,
793
+ "step": 1120
794
+ },
795
+ {
796
+ "epoch": 0.045111581300650724,
797
+ "grad_norm": 0.3305103751193663,
798
+ "learning_rate": 8.251925272662533e-05,
799
+ "loss": 0.0525,
800
+ "step": 1130
801
+ },
802
+ {
803
+ "epoch": 0.0455107988342848,
804
+ "grad_norm": 2.2298184302133053,
805
+ "learning_rate": 8.262267369090983e-05,
806
+ "loss": 0.061,
807
+ "step": 1140
808
+ },
809
+ {
810
+ "epoch": 0.04591001636791888,
811
+ "grad_norm": 0.93817128366632,
812
+ "learning_rate": 8.272519140379816e-05,
813
+ "loss": 0.0901,
814
+ "step": 1150
815
+ },
816
+ {
817
+ "epoch": 0.046309233901552956,
818
+ "grad_norm": 0.9468187201583408,
819
+ "learning_rate": 8.282682150630287e-05,
820
+ "loss": 0.0643,
821
+ "step": 1160
822
+ },
823
+ {
824
+ "epoch": 0.04670845143518703,
825
+ "grad_norm": 0.3289089088828509,
826
+ "learning_rate": 8.292757923665358e-05,
827
+ "loss": 0.056,
828
+ "step": 1170
829
+ },
830
+ {
831
+ "epoch": 0.04710766896882111,
832
+ "grad_norm": 0.7889381357026073,
833
+ "learning_rate": 8.302747944400921e-05,
834
+ "loss": 0.0944,
835
+ "step": 1180
836
+ },
837
+ {
838
+ "epoch": 0.04750688650245519,
839
+ "grad_norm": 0.6639073315567026,
840
+ "learning_rate": 8.312653660159161e-05,
841
+ "loss": 0.0649,
842
+ "step": 1190
843
+ },
844
+ {
845
+ "epoch": 0.04790610403608927,
846
+ "grad_norm": 1.1182816591396891,
847
+ "learning_rate": 8.322476481926952e-05,
848
+ "loss": 0.0546,
849
+ "step": 1200
850
+ },
851
+ {
852
+ "epoch": 0.04830532156972334,
853
+ "grad_norm": 1.2459681517250556,
854
+ "learning_rate": 8.332217785562046e-05,
855
+ "loss": 0.0781,
856
+ "step": 1210
857
+ },
858
+ {
859
+ "epoch": 0.04870453910335742,
860
+ "grad_norm": 1.019130956290865,
861
+ "learning_rate": 8.341878912949629e-05,
862
+ "loss": 0.0518,
863
+ "step": 1220
864
+ },
865
+ {
866
+ "epoch": 0.0491037566369915,
867
+ "grad_norm": 1.220775853388589,
868
+ "learning_rate": 8.351461173111643e-05,
869
+ "loss": 0.1007,
870
+ "step": 1230
871
+ },
872
+ {
873
+ "epoch": 0.049502974170625574,
874
+ "grad_norm": 1.7989207118679498,
875
+ "learning_rate": 8.360965843271233e-05,
876
+ "loss": 0.089,
877
+ "step": 1240
878
+ },
879
+ {
880
+ "epoch": 0.04990219170425965,
881
+ "grad_norm": 0.2744753836119675,
882
+ "learning_rate": 8.370394169874404e-05,
883
+ "loss": 0.1105,
884
+ "step": 1250
885
+ },
886
+ {
887
+ "epoch": 0.05030140923789373,
888
+ "grad_norm": 0.9289260996715916,
889
+ "learning_rate": 8.379747369571008e-05,
890
+ "loss": 0.0745,
891
+ "step": 1260
892
+ },
893
+ {
894
+ "epoch": 0.050700626771527806,
895
+ "grad_norm": 0.854436318996496,
896
+ "learning_rate": 8.389026630156928e-05,
897
+ "loss": 0.0633,
898
+ "step": 1270
899
+ },
900
+ {
901
+ "epoch": 0.05109984430516188,
902
+ "grad_norm": 0.8800673144054446,
903
+ "learning_rate": 8.39823311147933e-05,
904
+ "loss": 0.0578,
905
+ "step": 1280
906
+ },
907
+ {
908
+ "epoch": 0.05149906183879596,
909
+ "grad_norm": 0.6454065116002895,
910
+ "learning_rate": 8.407367946306682e-05,
911
+ "loss": 0.1094,
912
+ "step": 1290
913
+ },
914
+ {
915
+ "epoch": 0.05189827937243004,
916
+ "grad_norm": 1.5118193034152674,
917
+ "learning_rate": 8.416432241165186e-05,
918
+ "loss": 0.0722,
919
+ "step": 1300
920
+ },
921
+ {
922
+ "epoch": 0.05229749690606411,
923
+ "grad_norm": 2.0322104479409444,
924
+ "learning_rate": 8.425427077143155e-05,
925
+ "loss": 0.045,
926
+ "step": 1310
927
+ },
928
+ {
929
+ "epoch": 0.05269671443969819,
930
+ "grad_norm": 1.213509221133465,
931
+ "learning_rate": 8.434353510664808e-05,
932
+ "loss": 0.0488,
933
+ "step": 1320
934
+ },
935
+ {
936
+ "epoch": 0.05309593197333227,
937
+ "grad_norm": 0.3212234416060924,
938
+ "learning_rate": 8.443212574234867e-05,
939
+ "loss": 0.0541,
940
+ "step": 1330
941
+ },
942
+ {
943
+ "epoch": 0.05349514950696634,
944
+ "grad_norm": 2.459961563897475,
945
+ "learning_rate": 8.45200527715526e-05,
946
+ "loss": 0.0612,
947
+ "step": 1340
948
+ },
949
+ {
950
+ "epoch": 0.053894367040600424,
951
+ "grad_norm": 0.15501293377743686,
952
+ "learning_rate": 8.46073260621519e-05,
953
+ "loss": 0.0361,
954
+ "step": 1350
955
+ },
956
+ {
957
+ "epoch": 0.0542935845742345,
958
+ "grad_norm": 2.549676142822002,
959
+ "learning_rate": 8.469395526355722e-05,
960
+ "loss": 0.0678,
961
+ "step": 1360
962
+ },
963
+ {
964
+ "epoch": 0.05469280210786858,
965
+ "grad_norm": 1.1347192124455705,
966
+ "learning_rate": 8.477994981310044e-05,
967
+ "loss": 0.0403,
968
+ "step": 1370
969
+ },
970
+ {
971
+ "epoch": 0.055092019641502656,
972
+ "grad_norm": 1.3093319730595865,
973
+ "learning_rate": 8.486531894220431e-05,
974
+ "loss": 0.0498,
975
+ "step": 1380
976
+ },
977
+ {
978
+ "epoch": 0.05549123717513673,
979
+ "grad_norm": 1.4536529924377708,
980
+ "learning_rate": 8.495007168232951e-05,
981
+ "loss": 0.0701,
982
+ "step": 1390
983
+ },
984
+ {
985
+ "epoch": 0.05589045470877081,
986
+ "grad_norm": 0.531916156069153,
987
+ "learning_rate": 8.503421687070836e-05,
988
+ "loss": 0.0622,
989
+ "step": 1400
990
+ },
991
+ {
992
+ "epoch": 0.05628967224240489,
993
+ "grad_norm": 1.1571467808278315,
994
+ "learning_rate": 8.511776315587463e-05,
995
+ "loss": 0.0585,
996
+ "step": 1410
997
+ },
998
+ {
999
+ "epoch": 0.05668888977603896,
1000
+ "grad_norm": 1.7541520882599784,
1001
+ "learning_rate": 8.520071900299782e-05,
1002
+ "loss": 0.1239,
1003
+ "step": 1420
1004
+ },
1005
+ {
1006
+ "epoch": 0.05708810730967304,
1007
+ "grad_norm": 0.26434833237037547,
1008
+ "learning_rate": 8.528309269903042e-05,
1009
+ "loss": 0.06,
1010
+ "step": 1430
1011
+ },
1012
+ {
1013
+ "epoch": 0.05748732484330712,
1014
+ "grad_norm": 2.6666324283388474,
1015
+ "learning_rate": 8.536489235767567e-05,
1016
+ "loss": 0.0718,
1017
+ "step": 1440
1018
+ },
1019
+ {
1020
+ "epoch": 0.05788654237694119,
1021
+ "grad_norm": 0.3400922948582512,
1022
+ "learning_rate": 8.544612592418352e-05,
1023
+ "loss": 0.0447,
1024
+ "step": 1450
1025
+ },
1026
+ {
1027
+ "epoch": 0.058285759910575274,
1028
+ "grad_norm": 1.3999499096361745,
1029
+ "learning_rate": 8.552680117998175e-05,
1030
+ "loss": 0.0742,
1031
+ "step": 1460
1032
+ },
1033
+ {
1034
+ "epoch": 0.05868497744420935,
1035
+ "grad_norm": 0.7611622338959521,
1036
+ "learning_rate": 8.560692574714892e-05,
1037
+ "loss": 0.0468,
1038
+ "step": 1470
1039
+ },
1040
+ {
1041
+ "epoch": 0.059084194977843424,
1042
+ "grad_norm": 2.470390247344094,
1043
+ "learning_rate": 8.568650709273593e-05,
1044
+ "loss": 0.0904,
1045
+ "step": 1480
1046
+ },
1047
+ {
1048
+ "epoch": 0.059483412511477506,
1049
+ "grad_norm": 0.7772869206537453,
1050
+ "learning_rate": 8.576555253294187e-05,
1051
+ "loss": 0.0799,
1052
+ "step": 1490
1053
+ },
1054
+ {
1055
+ "epoch": 0.05988263004511158,
1056
+ "grad_norm": 0.20457141929145445,
1057
+ "learning_rate": 8.584406923715019e-05,
1058
+ "loss": 0.0378,
1059
+ "step": 1500
1060
+ },
1061
+ {
1062
+ "epoch": 0.060281847578745655,
1063
+ "grad_norm": 0.6068750755678066,
1064
+ "learning_rate": 8.592206423183099e-05,
1065
+ "loss": 0.0771,
1066
+ "step": 1510
1067
+ },
1068
+ {
1069
+ "epoch": 0.06068106511237974,
1070
+ "grad_norm": 1.452516141509269,
1071
+ "learning_rate": 8.599954440431426e-05,
1072
+ "loss": 0.0554,
1073
+ "step": 1520
1074
+ },
1075
+ {
1076
+ "epoch": 0.06108028264601381,
1077
+ "grad_norm": 0.569251980814837,
1078
+ "learning_rate": 8.60765165064396e-05,
1079
+ "loss": 0.0705,
1080
+ "step": 1530
1081
+ },
1082
+ {
1083
+ "epoch": 0.06147950017964789,
1084
+ "grad_norm": 0.2294559534793912,
1085
+ "learning_rate": 8.615298715808692e-05,
1086
+ "loss": 0.0574,
1087
+ "step": 1540
1088
+ },
1089
+ {
1090
+ "epoch": 0.06187871771328197,
1091
+ "grad_norm": 0.8496052440834033,
1092
+ "learning_rate": 8.622896285059299e-05,
1093
+ "loss": 0.0627,
1094
+ "step": 1550
1095
+ },
1096
+ {
1097
+ "epoch": 0.06227793524691604,
1098
+ "grad_norm": 0.7136223062428863,
1099
+ "learning_rate": 8.630444995005803e-05,
1100
+ "loss": 0.0515,
1101
+ "step": 1560
1102
+ },
1103
+ {
1104
+ "epoch": 0.06267715278055012,
1105
+ "grad_norm": 1.1106546858635593,
1106
+ "learning_rate": 8.637945470054669e-05,
1107
+ "loss": 0.0632,
1108
+ "step": 1570
1109
+ },
1110
+ {
1111
+ "epoch": 0.0630763703141842,
1112
+ "grad_norm": 0.3371801019234766,
1113
+ "learning_rate": 8.645398322718747e-05,
1114
+ "loss": 0.0788,
1115
+ "step": 1580
1116
+ },
1117
+ {
1118
+ "epoch": 0.06347558784781827,
1119
+ "grad_norm": 0.236169068255148,
1120
+ "learning_rate": 8.652804153917429e-05,
1121
+ "loss": 0.0744,
1122
+ "step": 1590
1123
+ },
1124
+ {
1125
+ "epoch": 0.06387480538145235,
1126
+ "grad_norm": 0.2064526314096174,
1127
+ "learning_rate": 8.660163553267397e-05,
1128
+ "loss": 0.0443,
1129
+ "step": 1600
1130
+ },
1131
+ {
1132
+ "epoch": 0.06427402291508644,
1133
+ "grad_norm": 2.459251200063712,
1134
+ "learning_rate": 8.667477099364316e-05,
1135
+ "loss": 0.0661,
1136
+ "step": 1610
1137
+ },
1138
+ {
1139
+ "epoch": 0.06467324044872051,
1140
+ "grad_norm": 1.0043243991341193,
1141
+ "learning_rate": 8.674745360055805e-05,
1142
+ "loss": 0.077,
1143
+ "step": 1620
1144
+ },
1145
+ {
1146
+ "epoch": 0.06507245798235459,
1147
+ "grad_norm": 0.6077242408276503,
1148
+ "learning_rate": 8.681968892706012e-05,
1149
+ "loss": 0.0758,
1150
+ "step": 1630
1151
+ },
1152
+ {
1153
+ "epoch": 0.06547167551598866,
1154
+ "grad_norm": 0.5112997355430338,
1155
+ "learning_rate": 8.689148244452086e-05,
1156
+ "loss": 0.0508,
1157
+ "step": 1640
1158
+ },
1159
+ {
1160
+ "epoch": 0.06587089304962274,
1161
+ "grad_norm": 0.6811161312428127,
1162
+ "learning_rate": 8.696283952452874e-05,
1163
+ "loss": 0.1102,
1164
+ "step": 1650
1165
+ },
1166
+ {
1167
+ "epoch": 0.06627011058325681,
1168
+ "grad_norm": 0.08253020817352436,
1169
+ "learning_rate": 8.703376544130078e-05,
1170
+ "loss": 0.0609,
1171
+ "step": 1660
1172
+ },
1173
+ {
1174
+ "epoch": 0.0666693281168909,
1175
+ "grad_norm": 0.14002755044269277,
1176
+ "learning_rate": 8.710426537402192e-05,
1177
+ "loss": 0.0687,
1178
+ "step": 1670
1179
+ },
1180
+ {
1181
+ "epoch": 0.06706854565052497,
1182
+ "grad_norm": 2.7372240520461677,
1183
+ "learning_rate": 8.717434440911451e-05,
1184
+ "loss": 0.0472,
1185
+ "step": 1680
1186
+ },
1187
+ {
1188
+ "epoch": 0.06746776318415905,
1189
+ "grad_norm": 0.33623563408312035,
1190
+ "learning_rate": 8.724400754244037e-05,
1191
+ "loss": 0.0588,
1192
+ "step": 1690
1193
+ },
1194
+ {
1195
+ "epoch": 0.06786698071779312,
1196
+ "grad_norm": 0.5698564181062653,
1197
+ "learning_rate": 8.731325968143787e-05,
1198
+ "loss": 0.0769,
1199
+ "step": 1700
1200
+ },
1201
+ {
1202
+ "epoch": 0.0682661982514272,
1203
+ "grad_norm": 0.4525305014011975,
1204
+ "learning_rate": 8.738210564719664e-05,
1205
+ "loss": 0.0828,
1206
+ "step": 1710
1207
+ },
1208
+ {
1209
+ "epoch": 0.06866541578506127,
1210
+ "grad_norm": 0.28785918919805653,
1211
+ "learning_rate": 8.745055017647127e-05,
1212
+ "loss": 0.0762,
1213
+ "step": 1720
1214
+ },
1215
+ {
1216
+ "epoch": 0.06906463331869536,
1217
+ "grad_norm": 0.4232601140844762,
1218
+ "learning_rate": 8.751859792363715e-05,
1219
+ "loss": 0.071,
1220
+ "step": 1730
1221
+ },
1222
+ {
1223
+ "epoch": 0.06946385085232944,
1224
+ "grad_norm": 0.4192502611164337,
1225
+ "learning_rate": 8.758625346258969e-05,
1226
+ "loss": 0.0839,
1227
+ "step": 1740
1228
+ },
1229
+ {
1230
+ "epoch": 0.06986306838596351,
1231
+ "grad_norm": 0.6623452619836312,
1232
+ "learning_rate": 8.765352128858903e-05,
1233
+ "loss": 0.089,
1234
+ "step": 1750
1235
+ },
1236
+ {
1237
+ "epoch": 0.07026228591959759,
1238
+ "grad_norm": 1.857160256591451,
1239
+ "learning_rate": 8.772040582005251e-05,
1240
+ "loss": 0.0792,
1241
+ "step": 1760
1242
+ },
1243
+ {
1244
+ "epoch": 0.07066150345323166,
1245
+ "grad_norm": 0.6388474695897128,
1246
+ "learning_rate": 8.778691140029603e-05,
1247
+ "loss": 0.0625,
1248
+ "step": 1770
1249
+ },
1250
+ {
1251
+ "epoch": 0.07106072098686575,
1252
+ "grad_norm": 0.6978578032659793,
1253
+ "learning_rate": 8.785304229922658e-05,
1254
+ "loss": 0.1292,
1255
+ "step": 1780
1256
+ },
1257
+ {
1258
+ "epoch": 0.07145993852049982,
1259
+ "grad_norm": 0.3741360039984228,
1260
+ "learning_rate": 8.791880271498722e-05,
1261
+ "loss": 0.0714,
1262
+ "step": 1790
1263
+ },
1264
+ {
1265
+ "epoch": 0.0718591560541339,
1266
+ "grad_norm": 1.6505423990273078,
1267
+ "learning_rate": 8.798419677555634e-05,
1268
+ "loss": 0.0911,
1269
+ "step": 1800
1270
+ },
1271
+ {
1272
+ "epoch": 0.07225837358776797,
1273
+ "grad_norm": 0.6050083970686374,
1274
+ "learning_rate": 8.804922854030258e-05,
1275
+ "loss": 0.0882,
1276
+ "step": 1810
1277
+ },
1278
+ {
1279
+ "epoch": 0.07265759112140205,
1280
+ "grad_norm": 1.2408152482478003,
1281
+ "learning_rate": 8.811390200149687e-05,
1282
+ "loss": 0.0373,
1283
+ "step": 1820
1284
+ },
1285
+ {
1286
+ "epoch": 0.07305680865503612,
1287
+ "grad_norm": 0.21333985745568887,
1288
+ "learning_rate": 8.81782210857831e-05,
1289
+ "loss": 0.0449,
1290
+ "step": 1830
1291
+ },
1292
+ {
1293
+ "epoch": 0.07345602618867021,
1294
+ "grad_norm": 0.27016768497953925,
1295
+ "learning_rate": 8.824218965560874e-05,
1296
+ "loss": 0.069,
1297
+ "step": 1840
1298
+ },
1299
+ {
1300
+ "epoch": 0.07385524372230429,
1301
+ "grad_norm": 0.38626281679917784,
1302
+ "learning_rate": 8.830581151061661e-05,
1303
+ "loss": 0.0392,
1304
+ "step": 1850
1305
+ },
1306
+ {
1307
+ "epoch": 0.07425446125593836,
1308
+ "grad_norm": 4.446410260564538,
1309
+ "learning_rate": 8.836909038899914e-05,
1310
+ "loss": 0.0814,
1311
+ "step": 1860
1312
+ },
1313
+ {
1314
+ "epoch": 0.07465367878957244,
1315
+ "grad_norm": 0.5932527992254153,
1316
+ "learning_rate": 8.843202996881645e-05,
1317
+ "loss": 0.0618,
1318
+ "step": 1870
1319
+ },
1320
+ {
1321
+ "epoch": 0.07505289632320651,
1322
+ "grad_norm": 0.5026914124479029,
1323
+ "learning_rate": 8.849463386927907e-05,
1324
+ "loss": 0.0847,
1325
+ "step": 1880
1326
+ },
1327
+ {
1328
+ "epoch": 0.07545211385684059,
1329
+ "grad_norm": 1.8341657629391837,
1330
+ "learning_rate": 8.85569056519969e-05,
1331
+ "loss": 0.0818,
1332
+ "step": 1890
1333
+ },
1334
+ {
1335
+ "epoch": 0.07585133139047467,
1336
+ "grad_norm": 0.9574408278201783,
1337
+ "learning_rate": 8.861884882219493e-05,
1338
+ "loss": 0.0791,
1339
+ "step": 1900
1340
+ },
1341
+ {
1342
+ "epoch": 0.07625054892410875,
1343
+ "grad_norm": 1.1904224506814953,
1344
+ "learning_rate": 8.868046682989726e-05,
1345
+ "loss": 0.0836,
1346
+ "step": 1910
1347
+ },
1348
+ {
1349
+ "epoch": 0.07664976645774282,
1350
+ "grad_norm": 0.523474266581198,
1351
+ "learning_rate": 8.874176307108011e-05,
1352
+ "loss": 0.0628,
1353
+ "step": 1920
1354
+ },
1355
+ {
1356
+ "epoch": 0.0770489839913769,
1357
+ "grad_norm": 1.559277136342286,
1358
+ "learning_rate": 8.880274088879496e-05,
1359
+ "loss": 0.0609,
1360
+ "step": 1930
1361
+ },
1362
+ {
1363
+ "epoch": 0.07744820152501097,
1364
+ "grad_norm": 0.5127563922641422,
1365
+ "learning_rate": 8.88634035742625e-05,
1366
+ "loss": 0.0925,
1367
+ "step": 1940
1368
+ },
1369
+ {
1370
+ "epoch": 0.07784741905864506,
1371
+ "grad_norm": 0.22233699013087163,
1372
+ "learning_rate": 8.89237543679387e-05,
1373
+ "loss": 0.0523,
1374
+ "step": 1950
1375
+ },
1376
+ {
1377
+ "epoch": 0.07824663659227914,
1378
+ "grad_norm": 0.12044996122785652,
1379
+ "learning_rate": 8.898379646055336e-05,
1380
+ "loss": 0.0619,
1381
+ "step": 1960
1382
+ },
1383
+ {
1384
+ "epoch": 0.07864585412591321,
1385
+ "grad_norm": 0.23180408798314633,
1386
+ "learning_rate": 8.904353299412252e-05,
1387
+ "loss": 0.0602,
1388
+ "step": 1970
1389
+ },
1390
+ {
1391
+ "epoch": 0.07904507165954729,
1392
+ "grad_norm": 3.283710519006683,
1393
+ "learning_rate": 8.91029670629349e-05,
1394
+ "loss": 0.055,
1395
+ "step": 1980
1396
+ },
1397
+ {
1398
+ "epoch": 0.07944428919318136,
1399
+ "grad_norm": 0.15911681428276372,
1400
+ "learning_rate": 8.916210171451378e-05,
1401
+ "loss": 0.0587,
1402
+ "step": 1990
1403
+ },
1404
+ {
1405
+ "epoch": 0.07984350672681544,
1406
+ "grad_norm": 0.17469923550169483,
1407
+ "learning_rate": 8.922093995055462e-05,
1408
+ "loss": 0.0639,
1409
+ "step": 2000
1410
+ },
1411
+ {
1412
+ "epoch": 0.08024272426044952,
1413
+ "grad_norm": 5.417640775721216,
1414
+ "learning_rate": 8.927948472783942e-05,
1415
+ "loss": 0.0887,
1416
+ "step": 2010
1417
+ },
1418
+ {
1419
+ "epoch": 0.0806419417940836,
1420
+ "grad_norm": 1.4044429457986154,
1421
+ "learning_rate": 8.933773895912817e-05,
1422
+ "loss": 0.1035,
1423
+ "step": 2020
1424
+ },
1425
+ {
1426
+ "epoch": 0.08104115932771767,
1427
+ "grad_norm": 0.3385160644949956,
1428
+ "learning_rate": 8.939570551402853e-05,
1429
+ "loss": 0.0743,
1430
+ "step": 2030
1431
+ },
1432
+ {
1433
+ "epoch": 0.08144037686135175,
1434
+ "grad_norm": 0.21983666639913327,
1435
+ "learning_rate": 8.945338721984404e-05,
1436
+ "loss": 0.0601,
1437
+ "step": 2040
1438
+ },
1439
+ {
1440
+ "epoch": 0.08183959439498582,
1441
+ "grad_norm": 2.553192791387149,
1442
+ "learning_rate": 8.951078686240155e-05,
1443
+ "loss": 0.0806,
1444
+ "step": 2050
1445
+ },
1446
+ {
1447
+ "epoch": 0.0822388119286199,
1448
+ "grad_norm": 1.1619542999670456,
1449
+ "learning_rate": 8.956790718685858e-05,
1450
+ "loss": 0.0894,
1451
+ "step": 2060
1452
+ },
1453
+ {
1454
+ "epoch": 0.08263802946225399,
1455
+ "grad_norm": 0.38923286712392613,
1456
+ "learning_rate": 8.962475089849113e-05,
1457
+ "loss": 0.1112,
1458
+ "step": 2070
1459
+ },
1460
+ {
1461
+ "epoch": 0.08303724699588806,
1462
+ "grad_norm": 4.726496524970905,
1463
+ "learning_rate": 8.968132066346247e-05,
1464
+ "loss": 0.1007,
1465
+ "step": 2080
1466
+ },
1467
+ {
1468
+ "epoch": 0.08343646452952214,
1469
+ "grad_norm": 1.1094373526242203,
1470
+ "learning_rate": 8.973761910957347e-05,
1471
+ "loss": 0.0969,
1472
+ "step": 2090
1473
+ },
1474
+ {
1475
+ "epoch": 0.08383568206315621,
1476
+ "grad_norm": 0.2895375239530162,
1477
+ "learning_rate": 8.97936488269952e-05,
1478
+ "loss": 0.0379,
1479
+ "step": 2100
1480
+ },
1481
+ {
1482
+ "epoch": 0.08423489959679029,
1483
+ "grad_norm": 1.7543701496251414,
1484
+ "learning_rate": 8.984941236898378e-05,
1485
+ "loss": 0.0675,
1486
+ "step": 2110
1487
+ },
1488
+ {
1489
+ "epoch": 0.08463411713042437,
1490
+ "grad_norm": 1.005491300629812,
1491
+ "learning_rate": 8.990491225257873e-05,
1492
+ "loss": 0.0824,
1493
+ "step": 2120
1494
+ },
1495
+ {
1496
+ "epoch": 0.08503333466405845,
1497
+ "grad_norm": 0.09759666962453613,
1498
+ "learning_rate": 8.996015095928464e-05,
1499
+ "loss": 0.0559,
1500
+ "step": 2130
1501
+ },
1502
+ {
1503
+ "epoch": 0.08543255219769252,
1504
+ "grad_norm": 0.39268330721556804,
1505
+ "learning_rate": 9.001513093573687e-05,
1506
+ "loss": 0.1009,
1507
+ "step": 2140
1508
+ },
1509
+ {
1510
+ "epoch": 0.0858317697313266,
1511
+ "grad_norm": 1.1297782557560954,
1512
+ "learning_rate": 9.006985459435193e-05,
1513
+ "loss": 0.0483,
1514
+ "step": 2150
1515
+ },
1516
+ {
1517
+ "epoch": 0.08623098726496067,
1518
+ "grad_norm": 1.0829945030987893,
1519
+ "learning_rate": 9.012432431396249e-05,
1520
+ "loss": 0.1053,
1521
+ "step": 2160
1522
+ },
1523
+ {
1524
+ "epoch": 0.08663020479859475,
1525
+ "grad_norm": 1.078088393046469,
1526
+ "learning_rate": 9.017854244043799e-05,
1527
+ "loss": 0.0785,
1528
+ "step": 2170
1529
+ },
1530
+ {
1531
+ "epoch": 0.08702942233222884,
1532
+ "grad_norm": 0.5133010653273858,
1533
+ "learning_rate": 9.023251128729078e-05,
1534
+ "loss": 0.065,
1535
+ "step": 2180
1536
+ },
1537
+ {
1538
+ "epoch": 0.08742863986586291,
1539
+ "grad_norm": 0.4491579268972776,
1540
+ "learning_rate": 9.028623313626857e-05,
1541
+ "loss": 0.1013,
1542
+ "step": 2190
1543
+ },
1544
+ {
1545
+ "epoch": 0.08782785739949699,
1546
+ "grad_norm": 2.6647367358509193,
1547
+ "learning_rate": 9.033971023793318e-05,
1548
+ "loss": 0.1198,
1549
+ "step": 2200
1550
+ },
1551
+ {
1552
+ "epoch": 0.08822707493313106,
1553
+ "grad_norm": 0.24347726763609193,
1554
+ "learning_rate": 9.039294481222638e-05,
1555
+ "loss": 0.0705,
1556
+ "step": 2210
1557
+ },
1558
+ {
1559
+ "epoch": 0.08862629246676514,
1560
+ "grad_norm": 0.8291531898362153,
1561
+ "learning_rate": 9.044593904902276e-05,
1562
+ "loss": 0.0832,
1563
+ "step": 2220
1564
+ },
1565
+ {
1566
+ "epoch": 0.08902551000039922,
1567
+ "grad_norm": 0.3659855529329366,
1568
+ "learning_rate": 9.049869510867034e-05,
1569
+ "loss": 0.1044,
1570
+ "step": 2230
1571
+ },
1572
+ {
1573
+ "epoch": 0.0894247275340333,
1574
+ "grad_norm": 5.902300509051372,
1575
+ "learning_rate": 9.055121512251897e-05,
1576
+ "loss": 0.0769,
1577
+ "step": 2240
1578
+ },
1579
+ {
1580
+ "epoch": 0.08982394506766737,
1581
+ "grad_norm": 1.350743023531939,
1582
+ "learning_rate": 9.060350119343701e-05,
1583
+ "loss": 0.0639,
1584
+ "step": 2250
1585
+ },
1586
+ {
1587
+ "epoch": 0.09022316260130145,
1588
+ "grad_norm": 0.6596569601933064,
1589
+ "learning_rate": 9.065555539631659e-05,
1590
+ "loss": 0.0776,
1591
+ "step": 2260
1592
+ },
1593
+ {
1594
+ "epoch": 0.09062238013493552,
1595
+ "grad_norm": 0.42640773750176064,
1596
+ "learning_rate": 9.070737977856764e-05,
1597
+ "loss": 0.1034,
1598
+ "step": 2270
1599
+ },
1600
+ {
1601
+ "epoch": 0.0910215976685696,
1602
+ "grad_norm": 0.1479642989294186,
1603
+ "learning_rate": 9.075897636060108e-05,
1604
+ "loss": 0.1035,
1605
+ "step": 2280
1606
+ },
1607
+ {
1608
+ "epoch": 0.09142081520220369,
1609
+ "grad_norm": 0.8493088400026695,
1610
+ "learning_rate": 9.081034713630148e-05,
1611
+ "loss": 0.0973,
1612
+ "step": 2290
1613
+ },
1614
+ {
1615
+ "epoch": 0.09182003273583776,
1616
+ "grad_norm": 1.5886948837208812,
1617
+ "learning_rate": 9.086149407348943e-05,
1618
+ "loss": 0.0432,
1619
+ "step": 2300
1620
+ },
1621
+ {
1622
+ "epoch": 0.09221925026947184,
1623
+ "grad_norm": 3.079194389194246,
1624
+ "learning_rate": 9.091241911437374e-05,
1625
+ "loss": 0.0701,
1626
+ "step": 2310
1627
+ },
1628
+ {
1629
+ "epoch": 0.09261846780310591,
1630
+ "grad_norm": 0.9720536038970082,
1631
+ "learning_rate": 9.096312417599413e-05,
1632
+ "loss": 0.0652,
1633
+ "step": 2320
1634
+ },
1635
+ {
1636
+ "epoch": 0.09301768533673999,
1637
+ "grad_norm": 2.554630977671387,
1638
+ "learning_rate": 9.101361115065411e-05,
1639
+ "loss": 0.1155,
1640
+ "step": 2330
1641
+ },
1642
+ {
1643
+ "epoch": 0.09341690287037406,
1644
+ "grad_norm": 1.4123609822496237,
1645
+ "learning_rate": 9.106388190634483e-05,
1646
+ "loss": 0.0795,
1647
+ "step": 2340
1648
+ },
1649
+ {
1650
+ "epoch": 0.09381612040400815,
1651
+ "grad_norm": 0.4540060979609624,
1652
+ "learning_rate": 9.111393828715974e-05,
1653
+ "loss": 0.0732,
1654
+ "step": 2350
1655
+ },
1656
+ {
1657
+ "epoch": 0.09421533793764222,
1658
+ "grad_norm": 0.4075911828916644,
1659
+ "learning_rate": 9.116378211370049e-05,
1660
+ "loss": 0.0619,
1661
+ "step": 2360
1662
+ },
1663
+ {
1664
+ "epoch": 0.0946145554712763,
1665
+ "grad_norm": 1.4701997003972624,
1666
+ "learning_rate": 9.121341518347428e-05,
1667
+ "loss": 0.0605,
1668
+ "step": 2370
1669
+ },
1670
+ {
1671
+ "epoch": 0.09501377300491037,
1672
+ "grad_norm": 2.6283239651453347,
1673
+ "learning_rate": 9.126283927128288e-05,
1674
+ "loss": 0.071,
1675
+ "step": 2380
1676
+ },
1677
+ {
1678
+ "epoch": 0.09541299053854445,
1679
+ "grad_norm": 0.3425418898573907,
1680
+ "learning_rate": 9.131205612960331e-05,
1681
+ "loss": 0.0625,
1682
+ "step": 2390
1683
+ },
1684
+ {
1685
+ "epoch": 0.09581220807217854,
1686
+ "grad_norm": 1.2410360487953485,
1687
+ "learning_rate": 9.136106748896079e-05,
1688
+ "loss": 0.1062,
1689
+ "step": 2400
1690
+ },
1691
+ {
1692
+ "epoch": 0.09621142560581261,
1693
+ "grad_norm": 0.19271858886467114,
1694
+ "learning_rate": 9.140987505829382e-05,
1695
+ "loss": 0.0743,
1696
+ "step": 2410
1697
+ },
1698
+ {
1699
+ "epoch": 0.09661064313944669,
1700
+ "grad_norm": 1.9714280272805322,
1701
+ "learning_rate": 9.145848052531172e-05,
1702
+ "loss": 0.0613,
1703
+ "step": 2420
1704
+ },
1705
+ {
1706
+ "epoch": 0.09700986067308076,
1707
+ "grad_norm": 0.3896419401090849,
1708
+ "learning_rate": 9.150688555684487e-05,
1709
+ "loss": 0.0634,
1710
+ "step": 2430
1711
+ },
1712
+ {
1713
+ "epoch": 0.09740907820671484,
1714
+ "grad_norm": 0.46048251433669973,
1715
+ "learning_rate": 9.155509179918754e-05,
1716
+ "loss": 0.0472,
1717
+ "step": 2440
1718
+ },
1719
+ {
1720
+ "epoch": 0.09780829574034891,
1721
+ "grad_norm": 1.5138256061910327,
1722
+ "learning_rate": 9.160310087843403e-05,
1723
+ "loss": 0.0765,
1724
+ "step": 2450
1725
+ },
1726
+ {
1727
+ "epoch": 0.098207513273983,
1728
+ "grad_norm": 1.1248799093801287,
1729
+ "learning_rate": 9.16509144008077e-05,
1730
+ "loss": 0.0664,
1731
+ "step": 2460
1732
+ },
1733
+ {
1734
+ "epoch": 0.09860673080761707,
1735
+ "grad_norm": 1.1166878766617159,
1736
+ "learning_rate": 9.16985339529834e-05,
1737
+ "loss": 0.1108,
1738
+ "step": 2470
1739
+ },
1740
+ {
1741
+ "epoch": 0.09900594834125115,
1742
+ "grad_norm": 0.729913896764391,
1743
+ "learning_rate": 9.174596110240358e-05,
1744
+ "loss": 0.0759,
1745
+ "step": 2480
1746
+ },
1747
+ {
1748
+ "epoch": 0.09940516587488522,
1749
+ "grad_norm": 0.46379569663470255,
1750
+ "learning_rate": 9.17931973975876e-05,
1751
+ "loss": 0.0787,
1752
+ "step": 2490
1753
+ },
1754
+ {
1755
+ "epoch": 0.0998043834085193,
1756
+ "grad_norm": 0.558357527018331,
1757
+ "learning_rate": 9.18402443684353e-05,
1758
+ "loss": 0.0794,
1759
+ "step": 2500
1760
+ },
1761
+ {
1762
+ "epoch": 0.10020360094215337,
1763
+ "grad_norm": 0.7685779629415801,
1764
+ "learning_rate": 9.18871035265243e-05,
1765
+ "loss": 0.038,
1766
+ "step": 2510
1767
+ },
1768
+ {
1769
+ "epoch": 0.10060281847578746,
1770
+ "grad_norm": 1.0882487656136575,
1771
+ "learning_rate": 9.193377636540134e-05,
1772
+ "loss": 0.0556,
1773
+ "step": 2520
1774
+ },
1775
+ {
1776
+ "epoch": 0.10100203600942154,
1777
+ "grad_norm": 0.5229247040180375,
1778
+ "learning_rate": 9.198026436086798e-05,
1779
+ "loss": 0.0448,
1780
+ "step": 2530
1781
+ },
1782
+ {
1783
+ "epoch": 0.10140125354305561,
1784
+ "grad_norm": 0.6391101299703439,
1785
+ "learning_rate": 9.202656897126054e-05,
1786
+ "loss": 0.0815,
1787
+ "step": 2540
1788
+ },
1789
+ {
1790
+ "epoch": 0.10180047107668969,
1791
+ "grad_norm": 0.3895503015517989,
1792
+ "learning_rate": 9.207269163772471e-05,
1793
+ "loss": 0.114,
1794
+ "step": 2550
1795
+ },
1796
+ {
1797
+ "epoch": 0.10219968861032376,
1798
+ "grad_norm": 0.3269743813110824,
1799
+ "learning_rate": 9.211863378448457e-05,
1800
+ "loss": 0.0706,
1801
+ "step": 2560
1802
+ },
1803
+ {
1804
+ "epoch": 0.10259890614395785,
1805
+ "grad_norm": 1.1652042050729956,
1806
+ "learning_rate": 9.216439681910651e-05,
1807
+ "loss": 0.0684,
1808
+ "step": 2570
1809
+ },
1810
+ {
1811
+ "epoch": 0.10299812367759192,
1812
+ "grad_norm": 1.6153664622217794,
1813
+ "learning_rate": 9.22099821327581e-05,
1814
+ "loss": 0.1103,
1815
+ "step": 2580
1816
+ },
1817
+ {
1818
+ "epoch": 0.103397341211226,
1819
+ "grad_norm": 0.16509593039498946,
1820
+ "learning_rate": 9.22553911004616e-05,
1821
+ "loss": 0.0734,
1822
+ "step": 2590
1823
+ },
1824
+ {
1825
+ "epoch": 0.10379655874486007,
1826
+ "grad_norm": 0.8302580796273004,
1827
+ "learning_rate": 9.230062508134313e-05,
1828
+ "loss": 0.0864,
1829
+ "step": 2600
1830
+ },
1831
+ {
1832
+ "epoch": 0.10419577627849415,
1833
+ "grad_norm": 0.5909787133808202,
1834
+ "learning_rate": 9.23456854188765e-05,
1835
+ "loss": 0.108,
1836
+ "step": 2610
1837
+ },
1838
+ {
1839
+ "epoch": 0.10459499381212822,
1840
+ "grad_norm": 1.4761953161125028,
1841
+ "learning_rate": 9.239057344112281e-05,
1842
+ "loss": 0.0537,
1843
+ "step": 2620
1844
+ },
1845
+ {
1846
+ "epoch": 0.10499421134576231,
1847
+ "grad_norm": 0.7445714330320076,
1848
+ "learning_rate": 9.243529046096521e-05,
1849
+ "loss": 0.0872,
1850
+ "step": 2630
1851
+ },
1852
+ {
1853
+ "epoch": 0.10539342887939639,
1854
+ "grad_norm": 0.5239333743685901,
1855
+ "learning_rate": 9.247983777633934e-05,
1856
+ "loss": 0.0667,
1857
+ "step": 2640
1858
+ },
1859
+ {
1860
+ "epoch": 0.10579264641303046,
1861
+ "grad_norm": 2.6708744396415383,
1862
+ "learning_rate": 9.252421667045939e-05,
1863
+ "loss": 0.0507,
1864
+ "step": 2650
1865
+ },
1866
+ {
1867
+ "epoch": 0.10619186394666454,
1868
+ "grad_norm": 1.6660882584362238,
1869
+ "learning_rate": 9.256842841203993e-05,
1870
+ "loss": 0.1077,
1871
+ "step": 2660
1872
+ },
1873
+ {
1874
+ "epoch": 0.10659108148029861,
1875
+ "grad_norm": 0.6395774684596824,
1876
+ "learning_rate": 9.26124742555134e-05,
1877
+ "loss": 0.0881,
1878
+ "step": 2670
1879
+ },
1880
+ {
1881
+ "epoch": 0.10699029901393269,
1882
+ "grad_norm": 0.15837676083751612,
1883
+ "learning_rate": 9.265635544124386e-05,
1884
+ "loss": 0.0854,
1885
+ "step": 2680
1886
+ },
1887
+ {
1888
+ "epoch": 0.10738951654756677,
1889
+ "grad_norm": 1.1860026106628092,
1890
+ "learning_rate": 9.270007319573648e-05,
1891
+ "loss": 0.0931,
1892
+ "step": 2690
1893
+ },
1894
+ {
1895
+ "epoch": 0.10778873408120085,
1896
+ "grad_norm": 0.6988001228875755,
1897
+ "learning_rate": 9.274362873184316e-05,
1898
+ "loss": 0.0718,
1899
+ "step": 2700
1900
+ },
1901
+ {
1902
+ "epoch": 0.10818795161483492,
1903
+ "grad_norm": 1.7627547758724476,
1904
+ "learning_rate": 9.278702324896465e-05,
1905
+ "loss": 0.0932,
1906
+ "step": 2710
1907
+ },
1908
+ {
1909
+ "epoch": 0.108587169148469,
1910
+ "grad_norm": 4.563162695809214,
1911
+ "learning_rate": 9.283025793324848e-05,
1912
+ "loss": 0.1143,
1913
+ "step": 2720
1914
+ },
1915
+ {
1916
+ "epoch": 0.10898638668210307,
1917
+ "grad_norm": 0.501400842221048,
1918
+ "learning_rate": 9.287333395778369e-05,
1919
+ "loss": 0.0331,
1920
+ "step": 2730
1921
+ },
1922
+ {
1923
+ "epoch": 0.10938560421573716,
1924
+ "grad_norm": 2.1431330678009024,
1925
+ "learning_rate": 9.291625248279169e-05,
1926
+ "loss": 0.0753,
1927
+ "step": 2740
1928
+ },
1929
+ {
1930
+ "epoch": 0.10978482174937124,
1931
+ "grad_norm": 1.5022132701371798,
1932
+ "learning_rate": 9.295901465581385e-05,
1933
+ "loss": 0.066,
1934
+ "step": 2750
1935
+ },
1936
+ {
1937
+ "epoch": 0.11018403928300531,
1938
+ "grad_norm": 1.3014789034361454,
1939
+ "learning_rate": 9.300162161189557e-05,
1940
+ "loss": 0.0803,
1941
+ "step": 2760
1942
+ },
1943
+ {
1944
+ "epoch": 0.11058325681663939,
1945
+ "grad_norm": 0.5743639067193882,
1946
+ "learning_rate": 9.304407447376704e-05,
1947
+ "loss": 0.0581,
1948
+ "step": 2770
1949
+ },
1950
+ {
1951
+ "epoch": 0.11098247435027346,
1952
+ "grad_norm": 2.296274245959155,
1953
+ "learning_rate": 9.308637435202077e-05,
1954
+ "loss": 0.0775,
1955
+ "step": 2780
1956
+ },
1957
+ {
1958
+ "epoch": 0.11138169188390754,
1959
+ "grad_norm": 2.3308610052506897,
1960
+ "learning_rate": 9.312852234528597e-05,
1961
+ "loss": 0.068,
1962
+ "step": 2790
1963
+ },
1964
+ {
1965
+ "epoch": 0.11178090941754162,
1966
+ "grad_norm": 2.256240916675447,
1967
+ "learning_rate": 9.317051954039962e-05,
1968
+ "loss": 0.0962,
1969
+ "step": 2800
1970
+ },
1971
+ {
1972
+ "epoch": 0.1121801269511757,
1973
+ "grad_norm": 1.7856796465922442,
1974
+ "learning_rate": 9.321236701257485e-05,
1975
+ "loss": 0.0685,
1976
+ "step": 2810
1977
+ },
1978
+ {
1979
+ "epoch": 0.11257934448480977,
1980
+ "grad_norm": 0.19735101163543742,
1981
+ "learning_rate": 9.325406582556589e-05,
1982
+ "loss": 0.0292,
1983
+ "step": 2820
1984
+ },
1985
+ {
1986
+ "epoch": 0.11297856201844385,
1987
+ "grad_norm": 1.3621946498396835,
1988
+ "learning_rate": 9.329561703183045e-05,
1989
+ "loss": 0.0945,
1990
+ "step": 2830
1991
+ },
1992
+ {
1993
+ "epoch": 0.11337777955207792,
1994
+ "grad_norm": 0.40134959605889914,
1995
+ "learning_rate": 9.333702167268908e-05,
1996
+ "loss": 0.1014,
1997
+ "step": 2840
1998
+ },
1999
+ {
2000
+ "epoch": 0.113776997085712,
2001
+ "grad_norm": 0.1841107116358454,
2002
+ "learning_rate": 9.337828077848174e-05,
2003
+ "loss": 0.0863,
2004
+ "step": 2850
2005
+ },
2006
+ {
2007
+ "epoch": 0.11417621461934609,
2008
+ "grad_norm": 0.49244833047293635,
2009
+ "learning_rate": 9.341939536872168e-05,
2010
+ "loss": 0.0499,
2011
+ "step": 2860
2012
+ },
2013
+ {
2014
+ "epoch": 0.11457543215298016,
2015
+ "grad_norm": 0.23498749839126715,
2016
+ "learning_rate": 9.346036645224653e-05,
2017
+ "loss": 0.0862,
2018
+ "step": 2870
2019
+ },
2020
+ {
2021
+ "epoch": 0.11497464968661424,
2022
+ "grad_norm": 3.916038906890403,
2023
+ "learning_rate": 9.350119502736693e-05,
2024
+ "loss": 0.0941,
2025
+ "step": 2880
2026
+ },
2027
+ {
2028
+ "epoch": 0.11537386722024831,
2029
+ "grad_norm": 0.1948321889118585,
2030
+ "learning_rate": 9.35418820820124e-05,
2031
+ "loss": 0.0793,
2032
+ "step": 2890
2033
+ },
2034
+ {
2035
+ "epoch": 0.11577308475388239,
2036
+ "grad_norm": 1.6380239320871577,
2037
+ "learning_rate": 9.35824285938748e-05,
2038
+ "loss": 0.0958,
2039
+ "step": 2900
2040
+ },
2041
+ {
2042
+ "epoch": 0.11617230228751647,
2043
+ "grad_norm": 1.011966587551444,
2044
+ "learning_rate": 9.362283553054931e-05,
2045
+ "loss": 0.0988,
2046
+ "step": 2910
2047
+ },
2048
+ {
2049
+ "epoch": 0.11657151982115055,
2050
+ "grad_norm": 0.9595034384482286,
2051
+ "learning_rate": 9.366310384967302e-05,
2052
+ "loss": 0.0585,
2053
+ "step": 2920
2054
+ },
2055
+ {
2056
+ "epoch": 0.11697073735478462,
2057
+ "grad_norm": 1.6309704442029718,
2058
+ "learning_rate": 9.370323449906096e-05,
2059
+ "loss": 0.0875,
2060
+ "step": 2930
2061
+ },
2062
+ {
2063
+ "epoch": 0.1173699548884187,
2064
+ "grad_norm": 0.8516664642667977,
2065
+ "learning_rate": 9.374322841684018e-05,
2066
+ "loss": 0.0583,
2067
+ "step": 2940
2068
+ },
2069
+ {
2070
+ "epoch": 0.11776917242205277,
2071
+ "grad_norm": 0.2052752626955342,
2072
+ "learning_rate": 9.378308653158114e-05,
2073
+ "loss": 0.0723,
2074
+ "step": 2950
2075
+ },
2076
+ {
2077
+ "epoch": 0.11816838995568685,
2078
+ "grad_norm": 0.5262234856148836,
2079
+ "learning_rate": 9.38228097624272e-05,
2080
+ "loss": 0.0704,
2081
+ "step": 2960
2082
+ },
2083
+ {
2084
+ "epoch": 0.11856760748932094,
2085
+ "grad_norm": 0.3928351586460673,
2086
+ "learning_rate": 9.386239901922172e-05,
2087
+ "loss": 0.1023,
2088
+ "step": 2970
2089
+ },
2090
+ {
2091
+ "epoch": 0.11896682502295501,
2092
+ "grad_norm": 0.8807251297741585,
2093
+ "learning_rate": 9.390185520263314e-05,
2094
+ "loss": 0.0499,
2095
+ "step": 2980
2096
+ },
2097
+ {
2098
+ "epoch": 0.11936604255658909,
2099
+ "grad_norm": 1.3492280712823417,
2100
+ "learning_rate": 9.394117920427792e-05,
2101
+ "loss": 0.0494,
2102
+ "step": 2990
2103
+ },
2104
+ {
2105
+ "epoch": 0.11976526009022316,
2106
+ "grad_norm": 2.7357178664397077,
2107
+ "learning_rate": 9.398037190684144e-05,
2108
+ "loss": 0.0816,
2109
+ "step": 3000
2110
+ }
2111
+ ],
2112
+ "logging_steps": 10,
2113
+ "max_steps": 50098,
2114
+ "num_input_tokens_seen": 0,
2115
+ "num_train_epochs": 2,
2116
+ "save_steps": 200,
2117
+ "stateful_callbacks": {
2118
+ "TrainerControl": {
2119
+ "args": {
2120
+ "should_epoch_stop": false,
2121
+ "should_evaluate": false,
2122
+ "should_log": false,
2123
+ "should_save": true,
2124
+ "should_training_stop": false
2125
+ },
2126
+ "attributes": {}
2127
+ }
2128
+ },
2129
+ "total_flos": 0.0,
2130
+ "train_batch_size": 2,
2131
+ "trial_name": null,
2132
+ "trial_params": null
2133
+ }
zero_to_fp32.py ADDED
@@ -0,0 +1,674 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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:
14
+ # python zero_to_fp32.py . output_dir/
15
+ # or
16
+ # python zero_to_fp32.py . output_dir/ --safe_serialization
17
+
18
+ import argparse
19
+ import torch
20
+ import glob
21
+ import math
22
+ import os
23
+ import re
24
+ import json
25
+ from tqdm import tqdm
26
+ from collections import OrderedDict
27
+ from dataclasses import dataclass
28
+
29
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
30
+ # DeepSpeed data structures it has to be available in the current python environment.
31
+ from deepspeed.utils import logger
32
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
33
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
34
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
35
+
36
+
37
+ @dataclass
38
+ class zero_model_state:
39
+ buffers: dict()
40
+ param_shapes: dict()
41
+ shared_params: list
42
+ ds_version: int
43
+ frozen_param_shapes: dict()
44
+ frozen_param_fragments: dict()
45
+
46
+
47
+ debug = 0
48
+
49
+ # load to cpu
50
+ device = torch.device('cpu')
51
+
52
+
53
+ def atoi(text):
54
+ return int(text) if text.isdigit() else text
55
+
56
+
57
+ def natural_keys(text):
58
+ '''
59
+ alist.sort(key=natural_keys) sorts in human order
60
+ http://nedbatchelder.com/blog/200712/human_sorting.html
61
+ (See Toothy's implementation in the comments)
62
+ '''
63
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
64
+
65
+
66
+ def get_model_state_file(checkpoint_dir, zero_stage):
67
+ if not os.path.isdir(checkpoint_dir):
68
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
69
+
70
+ # there should be only one file
71
+ if zero_stage <= 2:
72
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
73
+ elif zero_stage == 3:
74
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
75
+
76
+ if not os.path.exists(file):
77
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
78
+
79
+ return file
80
+
81
+
82
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
83
+ # XXX: need to test that this simple glob rule works for multi-node setup too
84
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
85
+
86
+ if len(ckpt_files) == 0:
87
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
88
+
89
+ return ckpt_files
90
+
91
+
92
+ def get_optim_files(checkpoint_dir):
93
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
94
+
95
+
96
+ def get_model_state_files(checkpoint_dir):
97
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
98
+
99
+
100
+ def parse_model_states(files):
101
+ zero_model_states = []
102
+ for file in files:
103
+ state_dict = torch.load(file, map_location=device)
104
+
105
+ if BUFFER_NAMES not in state_dict:
106
+ raise ValueError(f"{file} is not a model state checkpoint")
107
+ buffer_names = state_dict[BUFFER_NAMES]
108
+ if debug:
109
+ print("Found buffers:", buffer_names)
110
+
111
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
112
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
113
+ param_shapes = state_dict[PARAM_SHAPES]
114
+
115
+ # collect parameters that are included in param_shapes
116
+ param_names = []
117
+ for s in param_shapes:
118
+ for name in s.keys():
119
+ param_names.append(name)
120
+
121
+ # update with frozen parameters
122
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
123
+ if frozen_param_shapes is not None:
124
+ if debug:
125
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
126
+ param_names += list(frozen_param_shapes.keys())
127
+
128
+ # handle shared params
129
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
130
+
131
+ ds_version = state_dict.get(DS_VERSION, None)
132
+
133
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
134
+
135
+ z_model_state = zero_model_state(buffers=buffers,
136
+ param_shapes=param_shapes,
137
+ shared_params=shared_params,
138
+ ds_version=ds_version,
139
+ frozen_param_shapes=frozen_param_shapes,
140
+ frozen_param_fragments=frozen_param_fragments)
141
+ zero_model_states.append(z_model_state)
142
+
143
+ return zero_model_states
144
+
145
+
146
+ def parse_optim_states(files, ds_checkpoint_dir):
147
+ total_files = len(files)
148
+ state_dicts = []
149
+ for f in files:
150
+ state_dict = torch.load(f, map_location=device)
151
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
152
+ # and also handle the case where it was already removed by another helper script
153
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
154
+ state_dicts.append(state_dict)
155
+
156
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
157
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
158
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
159
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
160
+
161
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
162
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
163
+ # use the max of the partition_count to get the dp world_size.
164
+
165
+ if type(world_size) is list:
166
+ world_size = max(world_size)
167
+
168
+ if world_size != total_files:
169
+ raise ValueError(
170
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
171
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
172
+ )
173
+
174
+ # the groups are named differently in each stage
175
+ if zero_stage <= 2:
176
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
177
+ elif zero_stage == 3:
178
+ fp32_groups_key = FP32_FLAT_GROUPS
179
+ else:
180
+ raise ValueError(f"unknown zero stage {zero_stage}")
181
+
182
+ if zero_stage <= 2:
183
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
184
+ elif zero_stage == 3:
185
+ # if there is more than one param group, there will be multiple flattened tensors - one
186
+ # flattened tensor per group - for simplicity merge them into a single tensor
187
+ #
188
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
189
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
190
+
191
+ fp32_flat_groups = [
192
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
193
+ ]
194
+
195
+ return zero_stage, world_size, fp32_flat_groups
196
+
197
+
198
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
199
+ """
200
+ Returns fp32 state_dict reconstructed from ds checkpoint
201
+
202
+ Args:
203
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
204
+
205
+ """
206
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
207
+
208
+ optim_files = get_optim_files(ds_checkpoint_dir)
209
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
210
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
211
+
212
+ model_files = get_model_state_files(ds_checkpoint_dir)
213
+
214
+ zero_model_states = parse_model_states(model_files)
215
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
216
+
217
+ if zero_stage <= 2:
218
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
219
+ exclude_frozen_parameters)
220
+ elif zero_stage == 3:
221
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
222
+ exclude_frozen_parameters)
223
+
224
+
225
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
226
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
227
+ return
228
+
229
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
230
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
231
+
232
+ if debug:
233
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
234
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
235
+
236
+ wanted_params = len(frozen_param_shapes)
237
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
238
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
239
+ print(f'Frozen params: Have {avail_numel} numels to process.')
240
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
241
+
242
+ total_params = 0
243
+ total_numel = 0
244
+ for name, shape in frozen_param_shapes.items():
245
+ total_params += 1
246
+ unpartitioned_numel = shape.numel()
247
+ total_numel += unpartitioned_numel
248
+
249
+ state_dict[name] = frozen_param_fragments[name]
250
+
251
+ if debug:
252
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
253
+
254
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
255
+
256
+
257
+ def _has_callable(obj, fn):
258
+ attr = getattr(obj, fn, None)
259
+ return callable(attr)
260
+
261
+
262
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
263
+ param_shapes = zero_model_states[0].param_shapes
264
+
265
+ # Reconstruction protocol:
266
+ #
267
+ # XXX: document this
268
+
269
+ if debug:
270
+ for i in range(world_size):
271
+ for j in range(len(fp32_flat_groups[0])):
272
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
273
+
274
+ # XXX: memory usage doubles here (zero2)
275
+ num_param_groups = len(fp32_flat_groups[0])
276
+ merged_single_partition_of_fp32_groups = []
277
+ for i in range(num_param_groups):
278
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
279
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
280
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
281
+ avail_numel = sum(
282
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
283
+
284
+ if debug:
285
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
286
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
287
+ # not asserting if there is a mismatch due to possible padding
288
+ print(f"Have {avail_numel} numels to process.")
289
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
290
+
291
+ # params
292
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
293
+ # out-of-core computing solution
294
+ total_numel = 0
295
+ total_params = 0
296
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
297
+ offset = 0
298
+ avail_numel = full_single_fp32_vector.numel()
299
+ for name, shape in shapes.items():
300
+
301
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
302
+ total_numel += unpartitioned_numel
303
+ total_params += 1
304
+
305
+ if debug:
306
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
307
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
308
+ offset += unpartitioned_numel
309
+
310
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
311
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
312
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
313
+ # live optimizer object, so we are checking that the numbers are within the right range
314
+ align_to = 2 * world_size
315
+
316
+ def zero2_align(x):
317
+ return align_to * math.ceil(x / align_to)
318
+
319
+ if debug:
320
+ print(f"original offset={offset}, avail_numel={avail_numel}")
321
+
322
+ offset = zero2_align(offset)
323
+ avail_numel = zero2_align(avail_numel)
324
+
325
+ if debug:
326
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
327
+
328
+ # Sanity check
329
+ if offset != avail_numel:
330
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
331
+
332
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
333
+
334
+
335
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
336
+ exclude_frozen_parameters):
337
+ state_dict = OrderedDict()
338
+
339
+ # buffers
340
+ buffers = zero_model_states[0].buffers
341
+ state_dict.update(buffers)
342
+ if debug:
343
+ print(f"added {len(buffers)} buffers")
344
+
345
+ if not exclude_frozen_parameters:
346
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
347
+
348
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
349
+
350
+ # recover shared parameters
351
+ for pair in zero_model_states[0].shared_params:
352
+ if pair[1] in state_dict:
353
+ state_dict[pair[0]] = state_dict[pair[1]]
354
+
355
+ return state_dict
356
+
357
+
358
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
359
+ remainder = unpartitioned_numel % world_size
360
+ padding_numel = (world_size - remainder) if remainder else 0
361
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
362
+ return partitioned_numel, padding_numel
363
+
364
+
365
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
366
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
367
+ return
368
+
369
+ if debug:
370
+ for i in range(world_size):
371
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
372
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
373
+
374
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
375
+ wanted_params = len(frozen_param_shapes)
376
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
377
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
378
+ print(f'Frozen params: Have {avail_numel} numels to process.')
379
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
380
+
381
+ total_params = 0
382
+ total_numel = 0
383
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
384
+ total_params += 1
385
+ unpartitioned_numel = shape.numel()
386
+ total_numel += unpartitioned_numel
387
+
388
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
389
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
390
+
391
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
392
+
393
+ if debug:
394
+ print(
395
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
396
+ )
397
+
398
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
399
+
400
+
401
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
402
+ param_shapes = zero_model_states[0].param_shapes
403
+ avail_numel = fp32_flat_groups[0].numel() * world_size
404
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
405
+ # param, re-consolidating each param, while dealing with padding if any
406
+
407
+ # merge list of dicts, preserving order
408
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
409
+
410
+ if debug:
411
+ for i in range(world_size):
412
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
413
+
414
+ wanted_params = len(param_shapes)
415
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
416
+ # not asserting if there is a mismatch due to possible padding
417
+ avail_numel = fp32_flat_groups[0].numel() * world_size
418
+ print(f"Trainable params: Have {avail_numel} numels to process.")
419
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
420
+
421
+ # params
422
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
423
+ # out-of-core computing solution
424
+ offset = 0
425
+ total_numel = 0
426
+ total_params = 0
427
+ for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
428
+ unpartitioned_numel = shape.numel()
429
+ total_numel += unpartitioned_numel
430
+ total_params += 1
431
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
432
+
433
+ if debug:
434
+ print(
435
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
436
+ )
437
+
438
+ # XXX: memory usage doubles here
439
+ state_dict[name] = torch.cat(
440
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
441
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
442
+ offset += partitioned_numel
443
+
444
+ offset *= world_size
445
+
446
+ # Sanity check
447
+ if offset != avail_numel:
448
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
449
+
450
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
451
+
452
+
453
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
454
+ exclude_frozen_parameters):
455
+ state_dict = OrderedDict()
456
+
457
+ # buffers
458
+ buffers = zero_model_states[0].buffers
459
+ state_dict.update(buffers)
460
+ if debug:
461
+ print(f"added {len(buffers)} buffers")
462
+
463
+ if not exclude_frozen_parameters:
464
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
465
+
466
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
467
+
468
+ # recover shared parameters
469
+ for pair in zero_model_states[0].shared_params:
470
+ if pair[1] in state_dict:
471
+ state_dict[pair[0]] = state_dict[pair[1]]
472
+
473
+ return state_dict
474
+
475
+
476
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
477
+ """
478
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
479
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
480
+ via a model hub.
481
+
482
+ Args:
483
+ - ``checkpoint_dir``: path to the desired checkpoint folder
484
+ - ``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``
485
+ - ``exclude_frozen_parameters``: exclude frozen parameters
486
+
487
+ Returns:
488
+ - pytorch ``state_dict``
489
+
490
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
491
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
492
+ the checkpoint.
493
+
494
+ A typical usage might be ::
495
+
496
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
497
+ # do the training and checkpoint saving
498
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
499
+ model = model.cpu() # move to cpu
500
+ model.load_state_dict(state_dict)
501
+ # submit to model hub or save the model to share with others
502
+
503
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
504
+ application. i.e. you will need to re-initialize the deepspeed engine, since
505
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
506
+
507
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
508
+
509
+ """
510
+ if tag is None:
511
+ latest_path = os.path.join(checkpoint_dir, 'latest')
512
+ if os.path.isfile(latest_path):
513
+ with open(latest_path, 'r') as fd:
514
+ tag = fd.read().strip()
515
+ else:
516
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
517
+
518
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
519
+
520
+ if not os.path.isdir(ds_checkpoint_dir):
521
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
522
+
523
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
524
+
525
+
526
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
527
+ output_dir,
528
+ max_shard_size="5GB",
529
+ safe_serialization=False,
530
+ tag=None,
531
+ exclude_frozen_parameters=False):
532
+ """
533
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
534
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
535
+
536
+ Args:
537
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
538
+ - ``output_dir``: directory to the pytorch fp32 state_dict output files
539
+ - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
540
+ - ``safe_serialization``: whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
541
+ - ``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``
542
+ - ``exclude_frozen_parameters``: exclude frozen parameters
543
+ """
544
+ # Dependency pre-check
545
+ if safe_serialization:
546
+ try:
547
+ from safetensors.torch import save_file
548
+ except ImportError:
549
+ print('If you want to use `safe_serialization`, please `pip install safetensors`')
550
+ raise
551
+ if max_shard_size is not None:
552
+ try:
553
+ from huggingface_hub import split_torch_state_dict_into_shards
554
+ except ImportError:
555
+ print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
556
+ raise
557
+
558
+ # Convert zero checkpoint to state_dict
559
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
560
+
561
+ # Shard the model if it is too big.
562
+ weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
563
+ if max_shard_size is not None:
564
+ filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
565
+ state_dict_split = split_torch_state_dict_into_shards(state_dict,
566
+ filename_pattern=filename_pattern,
567
+ max_shard_size=max_shard_size)
568
+ else:
569
+ from collections import namedtuple
570
+ StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
571
+ state_dict_split = StateDictSplit(is_sharded=False,
572
+ filename_to_tensors={weights_name: list(state_dict.keys())})
573
+
574
+ # Save the model
575
+ filename_to_tensors = state_dict_split.filename_to_tensors.items()
576
+ for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
577
+ shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
578
+ output_path = os.path.join(output_dir, shard_file)
579
+ if safe_serialization:
580
+ save_file(shard, output_path, metadata={"format": "pt"})
581
+ else:
582
+ torch.save(shard, output_path)
583
+
584
+ # Save index if sharded
585
+ if state_dict_split.is_sharded:
586
+ index = {
587
+ "metadata": state_dict_split.metadata,
588
+ "weight_map": state_dict_split.tensor_to_filename,
589
+ }
590
+ save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
591
+ save_index_file = os.path.join(output_dir, save_index_file)
592
+ with open(save_index_file, "w", encoding="utf-8") as f:
593
+ content = json.dumps(index, indent=2, sort_keys=True) + "\n"
594
+ f.write(content)
595
+
596
+
597
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
598
+ """
599
+ 1. Put the provided model to cpu
600
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
601
+ 3. Load it into the provided model
602
+
603
+ Args:
604
+ - ``model``: the model object to update
605
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
606
+ - ``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``
607
+
608
+ Returns:
609
+ - ``model`: modified model
610
+
611
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
612
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
613
+ conveniently placed for you in the checkpoint folder.
614
+
615
+ A typical usage might be ::
616
+
617
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
618
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
619
+ # submit to model hub or save the model to share with others
620
+
621
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
622
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
623
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
624
+
625
+ """
626
+ logger.info(f"Extracting fp32 weights")
627
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
628
+
629
+ logger.info(f"Overwriting model with fp32 weights")
630
+ model = model.cpu()
631
+ model.load_state_dict(state_dict, strict=False)
632
+
633
+ return model
634
+
635
+
636
+ if __name__ == "__main__":
637
+ parser = argparse.ArgumentParser()
638
+ parser.add_argument("checkpoint_dir",
639
+ type=str,
640
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
641
+ parser.add_argument("output_dir",
642
+ type=str,
643
+ help="directory to the pytorch fp32 state_dict output files"
644
+ "(e.g. path/checkpoint-12-output/)")
645
+ parser.add_argument(
646
+ "--max_shard_size",
647
+ type=str,
648
+ default="5GB",
649
+ help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
650
+ "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
651
+ "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
652
+ "without CPU OOM issues.")
653
+ parser.add_argument(
654
+ "--safe_serialization",
655
+ default=False,
656
+ action='store_true',
657
+ help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
658
+ parser.add_argument("-t",
659
+ "--tag",
660
+ type=str,
661
+ default=None,
662
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
663
+ parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
664
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
665
+ args = parser.parse_args()
666
+
667
+ debug = args.debug
668
+
669
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
670
+ args.output_dir,
671
+ max_shard_size=args.max_shard_size,
672
+ safe_serialization=args.safe_serialization,
673
+ tag=args.tag,
674
+ exclude_frozen_parameters=args.exclude_frozen_parameters)